Google Git
Sign in
android / platform / external / pytorch / 5f6735ea97c90913f54c00b5eb2fe782b65b257b / . / test / test_jit.py
blob: a4f535921e5588ac335adc13c53033432b39ce98 [file] [log] [blame]
# -*- coding: utf-8 -*-
# Owner(s): ["oncall: jit"]
import torch
# This is how we include tests located in test/jit/...
# They are included here so that they are invoked when you call `test_jit.py`,
# do not run these test files directly.
from jit.test_tracer import TestTracer, TestMixTracingScripting # noqa: F401
from jit.test_recursive_script import TestRecursiveScript # noqa: F401
from jit.test_type_sharing import TestTypeSharing # noqa: F401
from jit.test_logging import TestLogging # noqa: F401
from jit.test_backends import TestBackends, TestBackendsWithCompiler # noqa: F401
from jit.test_backend_nnapi import TestNnapiBackend # noqa: F401
from jit.test_list_dict import TestList, TestDict, TestNamedTuple, TestScriptDict, TestScriptList # noqa: F401
from jit.test_async import TestAsync # noqa: F401
from jit.test_data_parallel import TestDataParallel # noqa: F401
from jit.test_models import TestModels # noqa: F401
from jit.test_modules import TestModules # noqa: F401
from jit.test_autodiff import TestAutodiffJit # noqa: F401
from jit.test_autodiff_subgraph_slicing import TestAutodiffSubgraphSlicing # noqa: F401
from jit.test_custom_operators import TestCustomOperators # noqa: F401
from jit.test_export_modes import TestExportModes # noqa: F401
from jit.test_graph_rewrite_passes import TestGraphRewritePasses # noqa: F401
from jit.test_class_type import TestClassType # noqa: F401
from jit.test_builtins import TestBuiltins, TestTensorBuiltins # noqa: F401
from jit.test_ignore_context_manager import TestIgnoreContextManager # noqa: F401
from jit.test_symbolic_shape_analysis import TestSymbolicShapeAnalysis # noqa: F401
from jit.test_op_decompositions import TestOpDecompositions # noqa: F401
from jit.test_unsupported_ops import TestUnsupportedOps # noqa: F401
from jit.test_freezing import TestFreezing, TestFrozenOptimizations, TestMKLDNNReinplacing # noqa: F401
from jit.test_peephole import TestPeephole # noqa: F401
from jit.test_alias_analysis import TestAliasAnalysis # noqa: F401
from jit.test_save_load import TestSaveLoad, TestSaveLoadFlatbuffer # noqa: F401
from jit.test_save_load_for_op_version import TestSaveLoadForOpVersion # noqa: F401
from jit.test_module_containers import TestModuleContainers # noqa: F401
from jit.test_python_bindings import TestPythonBindings # noqa: F401
from jit.test_python_ir import TestPythonIr # noqa: F401
from jit.test_functional_blocks import TestFunctionalBlocks # noqa: F401
from jit.test_remove_mutation import TestRemoveMutation # noqa: F401
from jit.test_torchbind import TestTorchbind # noqa: F401
from jit.test_module_interface import TestModuleInterface # noqa: F401 # noqa: F401
from jit.test_with import TestWith # noqa: F401
from jit.test_enum import TestEnum # noqa: F401
from jit.test_string_formatting import TestStringFormatting # noqa: F401
from jit.test_profiler import TestProfiler # noqa: F401
from jit.test_slice import TestSlice # noqa: F401
from jit.test_ignorable_args import TestIgnorableArgs # noqa: F401
from jit.test_hooks import TestHooks # noqa: F401
from jit.test_warn import TestWarn # noqa: F401
from jit.test_isinstance import TestIsinstance # noqa: F401
from jit.test_cuda import TestCUDA # noqa: F401
from jit.test_python_builtins import TestPythonBuiltinOP # noqa: F401
from jit.test_typing import TestTyping # noqa: F401
from jit.test_hash import TestHash # noqa: F401
from jit.test_complex import TestComplex # noqa: F401
from jit.test_jit_utils import TestJitUtils # noqa: F401
from jit.test_scriptmod_ann import TestScriptModuleInstanceAttributeTypeAnnotation # noqa: F401
from jit.test_types import TestTypesAndAnnotation # noqa: F401
from jit.test_misc import TestMisc # noqa: F401
from jit.test_upgraders import TestUpgraders # noqa: F401
from jit.test_pdt import TestPDT # noqa: F401
from jit.test_tensor_creation_ops import TestTensorCreationOps # noqa: F401
from jit.test_module_apis import TestModuleAPIs # noqa: F401
from jit.test_script_profile import TestScriptProfile # noqa: F401
from jit.test_convert_activation import TestFunctionalToInplaceActivation, TestInplaceToFunctionalActivation # noqa: F401
from jit.test_parametrization import TestParametrization # noqa: F401
from jit.test_attr import TestGetDefaultAttr # noqa: F401
from jit.test_aten_pow import TestAtenPow # noqa: F401
from jit.test_optimize_for_mobile_preserve_debug_info import TestOptimizeForMobilePreserveDebugInfo # noqa: F401
from jit.test_union import TestUnion # noqa: F401
from jit.test_batch_mm import TestBatchMM # noqa: F401
from jit.test_dtype_analysis import TestDtypeAnalysis, TestDtypeCustomRulesCPU # noqa: F401
from jit.test_device_analysis import TestDeviceAnalysis # noqa: F401
from jit.test_dce import TestDCE # noqa: F401
from jit.test_sparse import TestSparse # noqa: F401
from jit.test_tensor_methods import TestTensorMethods # noqa: F401
from jit.test_dataclasses import TestDataclasses # noqa: F401
# Torch
from torch import Tensor
from torch._C import TensorType, BoolType, parse_ir, _propagate_shapes
from torch.autograd import Variable
from torch.jit.annotations import BroadcastingList2, BroadcastingList3, Any # noqa: F401
from torch.nn.utils.rnn import PackedSequence
from torch.testing import FileCheck, make_tensor
import torch.autograd.profiler
import torch.cuda
import torch.jit
import torch.jit._logging
import torch.jit.frontend
import torch.nn as nn
import torch.nn.functional as F
# Testing utils
from torch.testing._internal import jit_utils
from torch.testing._internal.common_jit import check_against_reference
from torch.testing._internal.common_utils import run_tests, IS_WINDOWS, TEST_WITH_UBSAN, \
suppress_warnings, BUILD_WITH_CAFFE2, IS_SANDCASTLE, GRAPH_EXECUTOR, ProfilingMode, TestCase, \
freeze_rng_state, slowTest, TemporaryFileName, skipIfCompiledWithoutNumpy, \
enable_profiling_mode_for_profiling_tests, TEST_MKL, set_default_dtype, num_profiled_runs, \
skipIfCrossRef, IS_MACOS, skipIfTorchDynamo
from torch.testing._internal.jit_utils import JitTestCase, enable_cpu_fuser, disable_autodiff_subgraph_inlining, \
_trace, do_input_map, get_execution_plan, make_global, \
execWrapper, _inline_everything, _tmp_donotuse_dont_inline_everything, \
RUN_CUDA
from torch.testing._internal.jit_metaprogramming_utils import (
get_script_args,
create_input, unpack_variables,
additional_module_tests, EXCLUDE_SCRIPT_MODULES,
get_nn_module_name_from_kwargs, get_nn_mod_test_name, script_method_template)
from torch.testing._internal.common_nn import module_tests, new_module_tests, criterion_tests
# For testing truediv in python 2
from torch.testing._internal.test_module.future_div import div_int_future, div_float_future
from torch.testing._internal.test_module.no_future_div import div_int_nofuture, div_float_nofuture
# Standard library
from collections import defaultdict, namedtuple, OrderedDict
from copy import deepcopy
from itertools import product
from textwrap import dedent
from typing import List, Dict, NamedTuple, Optional, Tuple, Union
import copy
import functools
import inspect
import io
import itertools
import math
import numpy as np
import os
import pickle
import pickletools
import random
import re
import shutil
import string
import sys
import tempfile
import types
import typing
import unittest
import warnings
import zipfile
def canonical(graph):
return torch._C._jit_pass_canonicalize(graph).str(False)
def LSTMCellF(input, hx, cx, *params):
return LSTMCell(input, (hx, cx), *params)
def doAutodiffCheck(testname):
# TODO: setting false on test itself is not working
if "test_t_" in testname or testname == "test_t":
return False
if GRAPH_EXECUTOR == ProfilingMode.SIMPLE:
return False
if GRAPH_EXECUTOR == ProfilingMode.LEGACY:
return True
# these tests are disabled because BailOut nodes
# inserted by ProfilingExecutor interfere with
# subgraph slicing of Differentiable Graphs
test_exceptions = [
# functional
'test_nn_dropout',
'test_nn_log_softmax',
'test_nn_relu',
'test_nn_softmax',
'test_nn_threshold',
'test_nn_lp_pool2d',
'test_nn_lp_pool1d',
'test_nn_gumbel_softmax_hard',
'test_nn_gumbel_softmax',
'test_nn_multilabel_soft_margin_loss',
'test_nn_batch_norm',
'test_nn_max_pool2d_with_indices',
# AutogradJitGenerated
'test___rdiv___constant',
'test___rdiv___scalar_constant',
'test_split',
'test_split_dim',
'test_split_dim_neg0',
'test_split_size_list',
'test_split_size_list_dim',
'test_split_size_list_dim_neg0',
'test_split_with_sizes',
'test_split_with_sizes_dim',
'test_split_with_sizes_dim_neg0',
'test_split_with_sizes_size_0',
'test_nn_max_pool2d_with_indices',
]
if testname in test_exceptions:
return False
return True
# TODO: enable TE in PE when all tests are fixed
torch._C._jit_set_texpr_fuser_enabled(GRAPH_EXECUTOR == ProfilingMode.PROFILING)
torch._C._jit_set_profiling_executor(GRAPH_EXECUTOR != ProfilingMode.LEGACY)
def LSTMCell(input, hidden, w_ih, w_hh, b_ih=None, b_hh=None):
hx, cx = hidden
gates = F.linear(input, w_ih, b_ih) + F.linear(hx, w_hh, b_hh)
ingate, forgetgate, cellgate, outgate = gates.chunk(4, 1)
ingate = torch.sigmoid(ingate)
forgetgate = torch.sigmoid(forgetgate)
cellgate = torch.tanh(cellgate)
outgate = torch.sigmoid(outgate)
cy = (forgetgate * cx) + (ingate * cellgate)
hy = outgate * torch.tanh(cy)
return hy, cy
def LSTMCellC(*args, **kwargs):
hy, cy = LSTMCellF(*args, **kwargs)
return torch.cat((hy, cy))
def LSTMCellS(x, hx, cx, w_ih, w_hh, b_ih, b_hh):
gates = x.mm(w_ih.t()) + hx.mm(w_hh.t()) + b_ih + b_hh
ingate, forgetgate, cellgate, outgate = gates.chunk(4, 1)
ingate = torch.sigmoid(ingate)
forgetgate = torch.sigmoid(forgetgate)
cellgate = torch.tanh(cellgate)
outgate = torch.sigmoid(outgate)
cy = (forgetgate * cx) + (ingate * cellgate)
hy = outgate * torch.tanh(cy)
return hy, cy
# Code reference: https://github.com/pytorch/translate/blob/master/pytorch_translate/rnn_cell.py#L27:44
def MiLSTMCell(x, hx, cx, w_ih, w_hh, alpha, beta_i, beta_h, bias):
Wx = x.mm(w_ih.t())
Uz = hx.mm(w_hh.t())
# Section 2.1 in https://arxiv.org/pdf/1606.06630.pdf
gates = alpha * Wx * Uz + beta_i * Wx + beta_h * Uz + bias
# Same as LSTMCell after this point
ingate, forgetgate, cellgate, outgate = gates.chunk(4, 1)
ingate = ingate.sigmoid()
forgetgate = forgetgate.sigmoid()
cellgate = cellgate.tanh()
outgate = outgate.sigmoid()
cy = (forgetgate * cx) + (ingate * cellgate)
hy = outgate * cy.tanh()
return hy, cy
def get_lstm_inputs(device, training=False, seq_length=None):
input_shape = (3, 10) if seq_length is None else (seq_length, 3, 10)
input = torch.randn(*input_shape, dtype=torch.float, device=device, requires_grad=training)
hx = torch.randn(3, 20, dtype=torch.float, device=device, requires_grad=training)
cx = torch.randn(3, 20, dtype=torch.float, device=device, requires_grad=training)
module = nn.LSTMCell(10, 20).to(device, torch.float) # Just to allocate weights with correct sizes
if training:
params = tuple(module.parameters())
else:
params = tuple(p.requires_grad_(False) for p in module.parameters())
return (input, hx, cx) + params
def get_milstm_inputs(device, training=False):
minibatch = 3
input_size = 10
hidden_size = 20
x = torch.randn(minibatch, input_size, device=device, dtype=torch.float)
hx = torch.randn(minibatch, hidden_size, device=device, dtype=torch.float)
cx = torch.randn(minibatch, hidden_size, device=device, dtype=torch.float)
ih = torch.randn(4 * hidden_size, input_size, device=device, dtype=torch.float, requires_grad=training)
hh = torch.randn(4 * hidden_size, hidden_size, device=device, dtype=torch.float, requires_grad=training)
alpha = torch.randn(4 * hidden_size, dtype=torch.float, device=device, requires_grad=training)
ibeta = torch.randn(4 * hidden_size, dtype=torch.float, device=device, requires_grad=training)
hbeta = torch.randn(4 * hidden_size, dtype=torch.float, device=device, requires_grad=training)
bias = torch.randn(4 * hidden_size, dtype=torch.float, device=device, requires_grad=training)
return x, hx, cx, ih, hh, alpha, ibeta, hbeta, bias
def get_fn(file_name, script_path):
import importlib.util
spec = importlib.util.spec_from_file_location(file_name, script_path)
module = importlib.util.module_from_spec(spec)
spec.loader.exec_module(module)
fn = module.fn
return fn
def get_grad_executor(plan_state, diff_graph_idx=None, skip_check=False):
if diff_graph_idx is None:
nodes = list(plan_state.graph.nodes())
if not skip_check:
nodes = list(filter(lambda n : n.kind() != "prim::BailOut" and n.kind() != "prim::BailoutTemplate", nodes))
if len(nodes) == 1 or (len(nodes) == 2 and nodes[1].kind() == "prim::TupleConstruct"):
pass
elif len(nodes) == 2 and nodes[0].kind() == "prim::RequiresGradCheck" and nodes[1].kind() == "prim::If":
pass
else:
raise RuntimeError("Can't get a grad_executor for a non-differentiable graph")
grad_executors = list(plan_state.code.grad_executor_states())
return grad_executors[diff_graph_idx or 0]
def all_backward_graphs(script_module, diff_graph_idx=None):
# Note: for Python 2 the order seems to be unstable
ge_state = script_module.get_debug_state()
fwd_plan = get_execution_plan(ge_state)
grad_executor_state = get_grad_executor(fwd_plan, diff_graph_idx=diff_graph_idx)
bwd_plans = list(grad_executor_state.execution_plans.values())
return [p.graph.copy() for p in bwd_plans]
def backward_graph(script_module, diff_graph_idx=None, skip_check=False):
ge_state = script_module.get_debug_state()
fwd_plan = get_execution_plan(ge_state)
grad_executor_state = get_grad_executor(fwd_plan, diff_graph_idx=diff_graph_idx, skip_check=skip_check)
bwd_plan = get_execution_plan(grad_executor_state)
# Running JIT passes requires that we own the graph (with a shared_ptr).
# The debug state struct does not own its graph so we make a copy of it.
return bwd_plan.graph.copy()
# helper function to get sum of List[Tensor]
def _sum_of_list(tensorlist):
s = 0
for t in tensorlist:
s += t.sum()
return s
# has to be at top level or Pickle complains
class FooToPickle(torch.nn.Module):
def __init__(self):
super(FooToPickle, self).__init__()
self.bar = torch.jit.ScriptModule()
class TestJit(JitTestCase):
@unittest.skip("Requires a lot of RAM")
def test_big(self):
m = torch.jit.ScriptModule()
gig = int(1024 * 1024 * 1024 / 4)
# a small tensor in the first 4GB
m.v0 = nn.Parameter(torch.full((2,), 1, dtype=torch.float))
# a large tensor in the first 4GB that ends outside of it
m.v1 = nn.Parameter(torch.full((5, gig), 2, dtype=torch.float))
# a small tensor in >4GB space
m.v2 = nn.Parameter(torch.full((2,), 3, dtype=torch.float))
# s large tensor in the > 4GB space
m.v3 = nn.Parameter(torch.full((5, gig), 4, dtype=torch.float))
m2 = self.getExportImportCopy(m)
self.assertEqual(tuple(m.parameters()), tuple(m2.parameters()))
def test_inferred_as_tensor(self):
with self.assertRaisesRegex(RuntimeError, "Inferred the value for argument 'dim' to be of type 'Tensor' "
"because it was not annotated with an explicit type"):
@torch.jit.script
def dot(points, query, dim):
return (points * query).sum(dim)
def test_constants_pkl(self):
# This test asserts that the serialization archive includes a `constants.pkl`
# file. This file is used by `torch.load` to determine whether a zip file
# is a normal eager-mode serialization zip or a jit serialization zip. If
# you are deleting `constants.pkl`, make sure to update `torch.serialization.load`
# so it is still able to figure out which is which.
@torch.jit.script
def fn(x):
return x
buf = io.BytesIO()
torch.jit.save(fn, buf)
buf.seek(0)
files = zipfile.ZipFile(buf).filelist
self.assertTrue(any(['archive/constants.pkl' == f.filename for f in files]))
def test_script_fn_pkl(self):
with self.assertRaisesRegex(pickle.PickleError, "ScriptFunction cannot be pickled"):
@torch.jit.script
def fn(x: torch.Tensor) -> torch.Tensor:
return x
pkl_fn = pickle.dumps(fn, protocol=0)
def test_restore_device(self):
class M(torch.jit.ScriptModule):
def __init__(self, cpu_device_str):
super(M, self).__init__()
self.p0 = nn.Parameter(torch.tensor([0.3], dtype=torch.float,
device=cpu_device_str))
self.b0 = torch.tensor([0.9], dtype=torch.float,
device=cpu_device_str)
# main purpose is checking map_location works
m = M("cpu")
m2 = self.getExportImportCopy(m)
self.assertEqual(tuple(m.parameters()), tuple(m2.parameters()))
self.assertEqual(tuple(m.buffers()), tuple(m2.buffers()))
self.assertFalse(m2.p0.is_cuda)
self.assertFalse(m2.b0.is_cuda)
@unittest.skipIf(not RUN_CUDA, "restore device requires CUDA")
def test_restore_device_cuda(self):
class MyModule(torch.jit.ScriptModule):
def __init__(self):
super(MyModule, self).__init__()
self.register_buffer('b0', torch.randn(1, 3))
self.p0 = nn.Parameter(torch.randn(2, 3))
@torch.jit.script_method
def forward(self, x):
return x + self.b0 + self.p0
m = MyModule()
m.cuda(torch.cuda.device_count() - 1)
cuda_device_str = 'cuda:' + str(torch.cuda.device_count() - 1)
self.assertTrue(m.p0.is_cuda)
self.assertTrue(m.b0.is_cuda)
# restore to the saved devices
m2 = self.getExportImportCopy(m)
self.assertEqual(tuple(m.parameters()), tuple(m2.parameters()))
self.assertEqual(tuple(m.buffers()), tuple(m2.buffers()))
self.assertEqual(str(m2.p0.device), cuda_device_str)
self.assertEqual(str(m2.b0.device), cuda_device_str)
# restore all to cpu using string
cpu_device_str = 'cpu'
m3 = self.getExportImportCopy(m, map_location=cpu_device_str)
self.assertEqual(str(m3.p0.device), cpu_device_str)
self.assertEqual(str(m3.b0.device), cpu_device_str)
# restore all to first gpu using device
m4 = self.getExportImportCopy(
m3, map_location=torch.device('cuda:0'))
self.assertEqual(str(m4.p0.device), 'cuda:0')
self.assertEqual(str(m4.b0.device), 'cuda:0')
# compute and compare the results
input = torch.rand(2, 3).cuda(torch.cuda.device_count() - 1)
origin_result = m(input)
self.assertEqual(origin_result, m2(input))
self.assertEqual(origin_result, m3(input.cpu()))
self.assertEqual(origin_result, m4(input.cuda(0)))
def test_trace_retains_train(self):
class M(torch.nn.Module):
def forward(self, x):
return x
m = M()
m.eval()
tm = torch.jit.trace(m, (torch.rand(3)))
self.assertEqual(tm.training, m.training)
@unittest.skipIf(not RUN_CUDA, "restore device requires CUDA")
def test_restore_shared_storage_on_cuda(self):
class Foo(torch.jit.ScriptModule):
def __init__(self):
super(Foo, self).__init__()
whole_tensor = torch.randn(4, 5, dtype=torch.float, device='cpu')
self.p0 = nn.Parameter(whole_tensor.narrow(0, 0, 1))
self.register_buffer('b0', whole_tensor.narrow(0, 3, 1))
m = Foo()
m2 = self.getExportImportCopy(m, map_location=torch.device('cuda:0'))
self.assertEqual(tuple(m.parameters()), tuple(m2.parameters()))
self.assertEqual(tuple(m.buffers()), tuple(m2.buffers()))
self.assertTrue(m2.p0.is_cuda)
self.assertTrue(m2.b0.is_cuda)
self.assertTrue(m2.p0.is_shared())
self.assertTrue(m2.b0.is_shared())
self.assertEqual(m2.b0.storage().data_ptr(), m2.p0.storage().data_ptr())
def test_add_relu_fusion(self):
class M(torch.nn.Module):
def __init__(self, relu_op):
super(M, self).__init__()
self.relu_op = relu_op
def forward(self, a, b, c):
tmp = torch.add(a, b)
x = self.relu_op(tmp)
d = torch.add(a, c)
return x + d
a = torch.rand((7, 11))
a = a * -10
a = a + 5
b = torch.rand((7, 11))
c = torch.rand((7, 11))
m = torch.jit.script(M(torch.relu))
orig_res = m(a, b, c)
torch._C._jit_pass_fuse_add_relu(m.graph)
buffer = io.BytesIO()
torch.jit.save(m, buffer)
buffer.seek(0)
m = torch.jit.load(buffer)
new_res = m(a, b, c)
FileCheck().check_not("aten::relu(") \
.check("aten::_add_relu(") \
.run(m.graph)
torch.testing.assert_close(orig_res, new_res)
# add, relu_
a = torch.rand((7, 11))
a = a * -10
a = a + 5
b = torch.rand((7, 11))
c = torch.rand((7, 11))
m = torch.jit.script(M(torch.relu_))
orig_res = m(a, b, c)
torch._C._jit_pass_fuse_add_relu(m.graph)
buffer = io.BytesIO()
torch.jit.save(m, buffer)
buffer.seek(0)
m = torch.jit.load(buffer)
new_res = m(a, b, c)
FileCheck().check_not("aten::relu_(") \
.check("aten::_add_relu(") \
.run(m.graph)
torch.testing.assert_close(orig_res, new_res)
class Madd_(torch.nn.Module):
def __init__(self, relu_op):
super(Madd_, self).__init__()
self.relu_op = relu_op
def forward(self, a, b):
x = a.add_(b)
x = self.relu_op(x)
return x
# add_, relu_
a = torch.rand((7, 11))
a = a * -10
a = a + 5
b = torch.rand((7, 11))
# Because in place add_ will overwrite a
a_copy = a.clone()
m = torch.jit.script(Madd_(torch.relu_))
orig_res = m(a, b)
torch._C._jit_pass_fuse_add_relu(m.graph)
buffer = io.BytesIO()
torch.jit.save(m, buffer)
buffer.seek(0)
m = torch.jit.load(buffer)
new_res = m(a_copy, b)
FileCheck().check_not("aten::add_(") \
.check_not("aten::relu_(") \
.check("aten::_add_relu_(") \
.run(m.graph)
torch.testing.assert_close(orig_res, new_res)
# Since _add_relu_ does inplace mutation ensure
# a_copy is modified
torch.testing.assert_close(orig_res, a_copy)
class Madd_out(torch.nn.Module):
def __init__(self, relu_op):
super(Madd_out, self).__init__()
self.relu_op = relu_op
def forward(self, a, b):
x = torch.add(a, b, out=a)
x = self.relu_op(x)
return x
a = torch.rand((7, 11))
a = a * -10
a = a + 5
b = torch.rand((7, 11))
# add_out, relu_
a = torch.rand((7, 11))
a = a * -10
a = a + 5
b = torch.rand((7, 11))
# Because in place add_ will overwrite a
a_copy = a.clone()
m = torch.jit.script(Madd_out(torch.relu_))
orig_res = m(a, b)
torch._C._jit_pass_fuse_add_relu(m.graph)
buffer = io.BytesIO()
torch.jit.save(m, buffer)
buffer.seek(0)
m = torch.jit.load(buffer)
new_res = m(a_copy, b)
FileCheck().check_not("aten::add(") \
.check_not("aten::relu_(") \
.check("aten::_add_relu(") \
.run(m.graph)
torch.testing.assert_close(orig_res, new_res)
# Since _add_relu_ with out=a does inplace mutation ensure
# a_copy is modified
torch.testing.assert_close(orig_res, a_copy)
def test_repeat_interleave_script(self):
def fn(input: torch.Tensor, repeats: torch.Tensor) -> torch.Tensor:
output = input.repeat_interleave(repeats)
return output
fn_scripted = torch.jit.script(fn)
input = torch.tensor([5, 7], dtype=torch.int64)
repeats = torch.tensor([3, 6], dtype=torch.int64)
output = fn(input, repeats)
output_scripted = fn_scripted(input, repeats)
self.assertEqual(output_scripted, output)
@unittest.skipIf(GRAPH_EXECUTOR != ProfilingMode.LEGACY, "Simple executor doesn't have shape information")
def test_peephole_optimize_shape_ops(self):
def test_input(func, input, result):
# if result == 2 we will trigger a bailout and
# the unprofiled graph should return the correct result
self.assertEqual(func(input, profile_and_replay=True), result)
gre = func.graph_for(input)
FileCheck().check_not("prim::If").run(gre)
def test_dim():
@torch.jit.script
def func(x):
if x.dim() == 1:
return 1
else:
return 2
test_input(func, torch.tensor([0.5]), 1)
test_input(func, torch.tensor([[0.5]]), 2)
test_dim()
def test_size_index():
@torch.jit.script
def func(x):
if x.size(0) == 1:
return 1
else:
return 2
test_input(func, torch.rand([1, 2]), 1)
test_input(func, torch.rand([1, 3]), 1)
@torch.jit.script
def neg_index(x):
if x.size(-2) == 1:
return 1
else:
return 2
test_input(neg_index, torch.rand([1, 2]), 1)
test_input(neg_index, torch.rand([1, 3]), 1)
if GRAPH_EXECUTOR == ProfilingMode.PROFILING:
test_size_index()
def test_dtype():
@torch.jit.script
def func(x):
if x.dtype == torch.float32:
return 1
else:
return 2
test_input(func, torch.tensor(0.5, dtype=torch.float32), 1)
test_input(func, torch.tensor(0.5, dtype=torch.int64), 2)
test_dtype()
def test_is_floating_poiint():
@torch.jit.script
def func(x):
if x.is_floating_point():
return 1
else:
return 2
test_input(func, torch.tensor(0.5, dtype=torch.float32), 1)
test_input(func, torch.tensor(0.5, dtype=torch.int64), 2)
test_is_floating_poiint()
def test_device():
@torch.jit.script
def func_1(x):
if x.device == torch.device('cuda:0'):
a = 0
else:
a = 1
return a
@torch.jit.script
def func_2(x):
if x.is_cuda:
a = 0
else:
a = 1
return a
test_input(func_1, torch.tensor(0.5), 1)
test_input(func_2, torch.tensor(0.5), 1)
if RUN_CUDA:
test_input(func_1, torch.tensor(0.5, device="cuda:0"), 0)
test_input(func_2, torch.tensor(0.5, device="cuda:0"), 0)
test_device()
def test_attrs(self):
def foo(x):
return (
# x.dtype, TODO: dtype long -> instance conversion
x.device,
x.shape,
x.is_cuda,
x.is_mkldnn,
x.is_quantized,
x.requires_grad,
x.T,
x.mT,
x.H,
x.mH
# x.layout TODO: layout long -> instance conversion
)
scripted = torch.jit.script(foo)
x = torch.rand(3, 4)
self.assertEqual(scripted(x), foo(x))
def test_layout(self):
@torch.jit.script
def check(x, y):
return x.layout == y.layout
x = torch.rand(3, 4)
y = torch.rand(3, 4)
self.assertTrue(check(x, y))
def test_matrix_transpose(self):
@torch.jit.script
def check(x):
return torch.equal(x.mT, x.transpose(-2, -1))
x = torch.rand(3, 4)
self.assertTrue(check(x))
def test_transpose(self):
@torch.jit.script
def check(x):
return torch.equal(x.T, x.t())
x = torch.rand(3, 4)
self.assertTrue(check(x))
def test_matrix_conj_transpose(self):
@torch.jit.script
def check(x):
return torch.equal(x.mH, x.transpose(-2, -1).conj())
x = torch.rand(3, 4)
self.assertTrue(check(x))
x = make_tensor((3, 4), device="cpu", dtype=torch.complex64)
self.assertTrue(check(x))
def test_conj_transpose(self):
@torch.jit.script
def check(x):
return torch.equal(x.H, x.t().conj())
x = torch.rand(3, 4)
self.assertTrue(check(x))
x = make_tensor((3, 4), device="cpu", dtype=torch.complex64)
self.assertTrue(check(x))
def test_T_mT_H_mH(self):
def T(x):
return x.mT
def mT(x):
return x.mT
def H(x):
return x.H
def mH(x):
return x.mH
x = torch.rand(3, 4)
y = make_tensor((3, 4), device="cpu", dtype=torch.complex64)
self.checkScript(T, (x, ))
self.checkScript(mT, (x, ))
self.checkScript(H, (x, ))
self.checkScript(mH, (x, ))
self.checkScript(T, (y, ))
self.checkScript(mT, (y, ))
self.checkScript(H, (y, ))
self.checkScript(mH, (y, ))
def test_nn_conv(self):
class Mod(nn.Module):
def __init__(self, conv):
super().__init__()
self.conv = conv
def forward(self, input):
return self.conv(input)
inputs = [
# Conv
(Mod(nn.Conv1d(16, 33, 3, stride=2)), torch.randn(20, 16, 5)),
(Mod(nn.Conv2d(16, 33, 3, stride=2)), torch.randn(20, 16, 5, 10)),
(Mod(nn.Conv3d(16, 33, 3, stride=2)), torch.randn(20, 16, 3, 5, 4)),
# ConvTransposed
(Mod(nn.ConvTranspose1d(16, 33, 3, stride=2)), torch.randn(20, 16, 5)),
(Mod(nn.ConvTranspose2d(16, 33, 3, stride=2)), torch.randn(20, 16, 5, 10)),
(Mod(nn.ConvTranspose3d(16, 33, 3, stride=2)), torch.randn(20, 16, 3, 5, 4)),
]
for m, inp in inputs:
self.checkModule(m, (inp,))
@unittest.skipIf(GRAPH_EXECUTOR != ProfilingMode.PROFILING, 'Not implemented for Simple or Legacy')
def test_debug_flush_compilation_cache(self):
def foo(x):
return x + 2
class Mod(nn.Module):
def __init__(self):
super(Mod, self).__init__()
def forward(self, t):
return t + 2
m = torch.jit.script(Mod())
x = torch.rand(1, 10)
with enable_profiling_mode_for_profiling_tests():
jitted = self.checkScript(foo, (x,))
# shouldn't throw
states = jitted.get_debug_state()
# after flushing there shouldn't be
# no opt plan
jitted._debug_flush_compilation_cache()
with self.assertRaisesRegex(RuntimeError, "INTERNAL ASSERT FAILED"):
states = jitted.get_debug_state()
NUM_RUNS = 1
with num_profiled_runs(NUM_RUNS):
m(x)
m(x)
fwd = m._c._get_method("forward")
states = m.get_debug_state()
# after flushing there shouldn't be
# no opt plan
fwd._debug_flush_compilation_cache()
with self.assertRaisesRegex(RuntimeError, "INTERNAL ASSERT FAILED"):
states = m.get_debug_state()
def test_numel(self):
@torch.jit.script
def get_numel_script(x):
return x.numel()
x = torch.rand(3, 4)
numel = get_numel_script(x)
self.assertEqual(numel, x.numel())
def test_element_size(self):
@torch.jit.script
def get_element_size_script(x):
return x.element_size()
x = torch.rand(3, 4)
element_size = get_element_size_script(x)
self.assertEqual(element_size, x.element_size())
def test_Sequential(self):
class Seq(nn.Module):
def __init__(self):
super(Seq, self).__init__()
self.seq = nn.Sequential(nn.Linear(10, 20), nn.Linear(20, 30))
@torch.jit.script_method
def forward(self, x):
for l in self.seq:
x = l(x)
return x
m = torch.jit.script(Seq())
assert m.graph # ensure jit was able to compile
def test_ModuleList(self):
class Mod(nn.Module):
def __init__(self):
super(Mod, self).__init__()
self.model = nn.ModuleList([nn.Linear(10, 10) for _ in range(10)])
self.model += (nn.Linear(10, 20),)
self.model.append(nn.Linear(20, 30))
self.model.extend([nn.Linear(30, 40), nn.Linear(40, 50)])
def forward(self, v):
for m in self.model:
v = m(v)
return v
m = torch.jit.script(Mod())
assert m.graph # ensure jit was able to compile
def test_disabled(self):
torch.jit._state.disable()
try:
def f(x, y):
return x + y
self.assertIs(torch.jit.trace(f, (torch.randn(2, 2), torch.randn(2, 2))), f)
self.assertIs(torch.jit.script(f), f)
class MyModule(torch.jit.ScriptModule):
@torch.jit.script_method
def method(self, x):
return x
# XXX: Unfortunately ScriptModule won't simply become Module now,
# because that requires disabling the JIT at startup time, which
# we can't do in here.
# We need to or those two conditions to make it work with all versions of Python
self.assertTrue(inspect.ismethod(MyModule.method) or inspect.isfunction(MyModule.method))
finally:
torch.jit._state.enable()
def test_train_eval(self):
class Sub(nn.Module):
def forward(self, input):
if self.training:
return input
else:
return -input
class MyModule(torch.jit.ScriptModule):
def __init__(self, module):
super(MyModule, self).__init__()
self.module = module
@torch.jit.script_method
def forward(self, input):
return self.module(input) + 1
m = MyModule(Sub())
input = torch.rand(3, 4)
self.assertEqual(input + 1, m(input))
m.eval()
self.assertEqual(-input + 1, m(input))
# test batchnorm and dropout train/eval
input = torch.randn(6, 10)
batchnorm = nn.BatchNorm1d(10)
dropout = nn.Dropout(p=0.2)
m_batchnorm = MyModule(batchnorm)
self.assertEqual(batchnorm(input) + 1, m_batchnorm(input))
batchnorm.eval()
m_batchnorm.eval()
self.assertEqual(batchnorm(input) + 1, m_batchnorm(input))
m_dropout = MyModule(dropout)
dropout.eval()
m_dropout.eval()
self.assertEqual(dropout(input) + 1, m_dropout(input))
def test_nn_lp_pool2d(self):
class Mod(torch.nn.Module):
def __init__(self):
super().__init__()
self.l = torch.nn.LPPool2d(2, 3)
self.n = torch.nn.LPPool2d(2, (7, 1))
def forward(self, x):
return (self.l(x),
self.n(x),
torch.nn.functional.lp_pool2d(x, float(2), 3),
torch.nn.functional.lp_pool2d(x, 2, 3),
torch.nn.functional.lp_pool2d(x, float(2), (7, 1)))
self.checkModule(Mod(), (torch.rand(1, 3, 7, 7),))
def test_nn_lp_pool1d(self):
class Mod(torch.nn.Module):
def __init__(self):
super().__init__()
self.l = torch.nn.LPPool1d(2, 3)
self.n = torch.nn.LPPool1d(2, 7)
def forward(self, x):
return (self.l(x),
self.n(x),
torch.nn.functional.lp_pool1d(x, float(2), 3),
torch.nn.functional.lp_pool1d(x, 2, 3),
torch.nn.functional.lp_pool1d(x, float(2), 7))
self.checkModule(Mod(), (torch.rand(1, 3, 7),))
def test_nn_padding_functional(self):
class Mod(nn.Module):
def __init__(self, *pad):
super().__init__()
self.pad = pad
def forward(self, x):
return F.pad(x, self.pad, mode='constant', value=3.5)
inputs = [
(Mod(1, 2), torch.randn(1, 3, 4)), # 1D
(Mod(1, 2, 3, 4), torch.randn(1, 3, 4)), # 2D
(Mod(1, 2, 3, 4, 5, 6), torch.randn(1, 3, 4)), # 3D
]
for m, inp in inputs:
self.checkModule(m, (inp,))
def test_nn_padding(self):
class Mod(nn.Module):
def __init__(self, padding):
super().__init__()
self.padding = padding
def forward(self, input):
return self.padding(input)
inputs = [
(Mod(nn.ConstantPad1d(2, 3.5)), torch.randn(1, 2, 4)),
(Mod(nn.ConstantPad2d(2, 3.5)), torch.randn(1, 2, 2)),
(Mod(nn.ConstantPad3d(3, 3.5)), torch.randn(16, 3, 10, 20, 30)),
(Mod(nn.ReflectionPad1d(2)), torch.arange(8, dtype=torch.float).reshape(1, 2, 4)),
(Mod(nn.ReflectionPad2d(2)), torch.arange(9, dtype=torch.float).reshape(1, 1, 3, 3)),
(Mod(nn.ReflectionPad3d(3)), torch.randn(16, 3, 8, 32, 48)),
(Mod(nn.ReplicationPad1d(2)), torch.arange(8, dtype=torch.float).reshape(1, 2, 4)),
(Mod(nn.ReplicationPad2d(2)), torch.arange(9, dtype=torch.float).reshape(1, 1, 3, 3)),
(Mod(nn.ReplicationPad3d(3)), torch.randn(16, 3, 8, 32, 48)),
(Mod(nn.ZeroPad2d(2)), torch.randn(1, 1, 3, 3))
]
for m, inp in inputs:
self.checkModule(m, (inp,))
def test_script_autograd_grad(self):
def test_simple_grad(x, y):
# type: (Tensor, Tensor) -> List[Optional[Tensor]]
z = x + 2 * y + x * y
return torch.autograd.grad((z.sum(), ), (x, y))
def test_simple_grad_with_grad_outputs(x, y):
# type: (Tensor, Tensor) -> List[Optional[Tensor]]
z = x + 2 * y + x * y
grad_outputs = torch.jit.annotate(List[Optional[torch.Tensor]], [torch.ones((2, 2)), ])
return torch.autograd.grad((z, ), (x, y), grad_outputs)
def test_one_output_not_requires_grad(x, y):
# type: (Tensor, Tensor) -> List[Optional[Tensor]]
z = 2 * y + y
return torch.autograd.grad((z.sum(),), (x, y), allow_unused=True)
def test_retain_graph(x, y):
# type: (Tensor, Tensor) -> None
z = x + 2 * y + x * y
torch.autograd.grad((z.sum(), ), (x, y), retain_graph=True)
torch.autograd.grad((z.sum(), ), (x, y))
x = torch.randn(2, 2, requires_grad=True)
y = torch.randn(2, 2, requires_grad=True)
self.checkScript(test_simple_grad, (x, y), inputs_requires_grad=True)
self.checkScript(test_simple_grad_with_grad_outputs, (x, y), inputs_requires_grad=True)
self.checkScript(test_one_output_not_requires_grad, (x, y), inputs_requires_grad=True)
self.checkScript(test_retain_graph, (x, y), inputs_requires_grad=True)
def test_script_backward(self):
def checkBackwardScript(fn, inputs):
scripted_fn = torch.jit.script(fn)
FileCheck().check("torch.autograd.backward").run(scripted_fn.code)
recording_inputs = do_input_map(lambda t: t.detach().requires_grad_(), inputs)
fn(*inputs)
scripted_fn(*recording_inputs)
for inp1, inp2 in zip(inputs, recording_inputs):
self.assertEqual(inp1.grad, inp2.grad)
def test_tensor_backward(input):
# type: (Tensor) -> None
output = torch.relu(input)
output = output.softmax(0)
sum_out = output.sum()
sum_out.backward()
def test_torch_autograd_backward(input):
# type: (Tensor) -> None
output = torch.relu(input)
output = output.softmax(0)
torch.autograd.backward(output.sum())
def test_torch_autograd_backward_with_grad_tensors(input):
# type: (Tensor) -> None
output = torch.relu(input)
output = output.softmax(0)
grad_outputs = torch.jit.annotate(List[Optional[torch.Tensor]], [torch.ones((2, 2)), ])
torch.autograd.backward((output,), grad_outputs)
inp = torch.randn(2, 2, requires_grad=True)
checkBackwardScript(test_tensor_backward, (inp,))
checkBackwardScript(test_torch_autograd_backward, (inp,))
checkBackwardScript(test_torch_autograd_backward_with_grad_tensors, (inp,))
def test_script_backward_twice(self):
def checkBackwardTwiceScript(fn, inputs, retain_graph_=False):
torch._C._jit_set_profiling_executor(False)
with torch.jit.optimized_execution(True):
scripted_fn = torch.jit.script(fn, inputs)
FileCheck().check("prim::DifferentiableGraph").run(scripted_fn.graph_for(*inputs))
result = scripted_fn(*inputs)
result.sum().backward(retain_graph=retain_graph_)
if not retain_graph_:
self.assertRaisesRegex(RuntimeError, 'Specify retain_graph=True',
lambda: result.sum().backward())
else:
result.sum().backward()
def test_script_backward_twice_with_saved_values(input1, input2):
# type: (Tensor, Tensor) -> Tensor
tmp1 = torch.mul(input1, input2)
tmp2 = torch.abs(tmp1)
if torch.equal(input1, input2):
tmp2 = torch.acos(tmp2)
else:
tmp2 = torch.atan(tmp2)
result = torch.add(tmp2, input2)
return result
inp1 = torch.randn(2, 2, requires_grad=True)
inp2 = torch.randn(2, 2, requires_grad=True)
checkBackwardTwiceScript(test_script_backward_twice_with_saved_values, (inp1, inp2), False)
checkBackwardTwiceScript(test_script_backward_twice_with_saved_values, (inp1, inp2), True)
def test_diff_subgraph_clones_constants(self):
@torch.jit.script
def f(x, y):
return x + x + y + x + y + x + y + x + y + x
def count_constants(graph):
return sum(node.kind() == 'prim::Constant' for node in graph.nodes())
graph = f.graph.copy()
self.run_pass('cse', graph)
self.run_pass('create_autodiff_subgraphs', graph)
nodes = list(graph.nodes())
self.assertEqual(count_constants(graph), 1)
self.assertEqual(count_constants(nodes[1].g('Subgraph')), 1)
# TODO: adapt this test to check that GraphExecutor treats them differently
@unittest.skip("Need to be adjusted to Graph Executor")
def test_arg_configurations(self):
"""Different arg configurations should trigger different traces"""
x = Variable(torch.FloatTensor(4, 4).uniform_())
x_double = Variable(x.data.double())
x_grad = Variable(x.data.clone(), requires_grad=True)
y = Variable(torch.randn(4))
configurations = [
(x,),
(x_double,),
(x_grad,),
(y,),
([x, x],),
([x, y],),
]
if torch.cuda.is_available():
x_cuda = Variable(x.data.cuda())
configurations += [
(x_cuda,),
([x, x_cuda],),
([x_cuda, x],),
([[x_cuda, x]],),
]
if torch.cuda.device_count() > 1:
x_cuda_1 = Variable(x.data.cuda(1))
configurations += [
(x_cuda_1,),
([x_cuda, x_cuda_1],),
]
@torch.jit.compile(nderivs=0)
def fn(*args):
in_vars, _ = torch._C._jit_flatten(args)
return in_vars[0] + 1
for i, config in enumerate(configurations):
self.assertFalse(fn.has_trace_for(*config))
fn(*config)
self.assertTrue(fn.has_trace_for(*config))
for unk_config in configurations[i + 1:]:
self.assertFalse(fn.has_trace_for(*unk_config))
self.assertEqual(fn.hits, 0)
def test_torch_sum(self):
def fn(x):
return torch.sum(x)
def fn1(x, dim: int):
return torch.sum(x, dim)
x = torch.randn(3, 4)
self.checkScript(fn, (x, ))
self.checkScript(fn1, (x, 1, ))
self.checkScript(fn1, (x, 0, ))
def test_cse(self):
x = torch.tensor([0.4, 0.3], requires_grad=True)
y = torch.tensor([0.7, 0.5], requires_grad=True)
def fn(x, y):
w = (x + y) * (x + y) * (x + y)
t = torch.tanh(w) + torch.tanh(w)
z = (x + y) * (x + y) * (x + y) + t
return z
g, _ = torch.jit._get_trace_graph(fn, (x, y))
self.run_pass('cse', g)
do_exactly = True
FileCheck().check_count("add", 1).check_count("mul", 2, do_exactly) \
.check_count("tanh", 1, do_exactly).check_count("add", 2, do_exactly).check_next("return") \
.run(str(g))
self.assertExportImport(g, (x, y))
def test_cse_not_introduce_aliasing(self):
@torch.jit.script
def tensor_alias_outputs(x):
return x + x, x + x
self.run_pass('cse', tensor_alias_outputs.graph)
FileCheck().check_count("aten::add", 2).run(tensor_alias_outputs.graph)
@torch.jit.script
def ints_alias_outputs(x):
# type: (int) -> Tuple[int, int]
return x + x, x + x
# non-aliasing types can be CSEd
self.run_pass('cse', ints_alias_outputs.graph)
FileCheck().check_count("aten::add", 1, exactly=True).run(ints_alias_outputs.graph)
def test_recursive_cse(self):
input_str = """
graph(%x : Tensor,
%y : Tensor,
%20 : int):
%2 : int = prim::Constant[value=1]()
%3 : Tensor = aten::add(%x, %y, %2)
%4 : int = aten::add(%2, %20)
%5 : bool = aten::Bool(%4)
%z : int = prim::If(%5)
# CHECK: block
block0():
# CHECK-NOT: aten::add
%z.1 : int = aten::add(%2, %20)
-> (%z.1)
block1():
-> (%2)
return (%z)
"""
graph = parse_ir(input_str)
self.run_pass('cse', graph)
FileCheck().run(input_str, graph)
def test_pattern_based_rewrite(self):
# mul(mul(mul(mul(x,y),z),x),y) --> mul(mul(mulmul(x,y,z), x), y) -->
# --> mulmul(mulmul(x,y,z), x, y)
input_str = """
graph(%x, %y, %z):
# CHECK-NOT: aten::mul
# CHECK: my::fused_mulmul
%t = aten::mul(%x, %y)
%p = aten::mul(%t, %z)
# CHECK: my::fused_mulmul
%u = aten::mul(%p, %x)
%o = aten::mul(%u, %y)
return (%o)"""
graph = parse_ir(input_str)
torch._C._jit_pass_custom_pattern_based_rewrite_graph("""
graph(%a, %b, %c):
%q = aten::mul(%a, %b)
%r = aten::mul(%q, %c)
return (%r)""", """
graph(%a, %b, %c):
%r = my::fused_mulmul(%a, %b, %c)
return (%r)""", graph)
FileCheck().run(input_str, graph)
# Check that overlapping matches are handled correctly
# mul(mul(mul(x,y),z),x) --> mul(mulmul(x,y,z), x)
input_str = """
graph(%x, %y, %z):
# CHECK-NOT: aten::mul
# CHECK: my::fused_mulmul
%t = aten::mul(%x, %y)
%p = aten::mul(%t, %z)
# CHECK-NEXT: aten::mul
%u = aten::mul(%p, %x)
return (%u)"""
graph = parse_ir(input_str)
torch._C._jit_pass_custom_pattern_based_rewrite_graph("""
graph(%a, %b, %c):
%q = aten::mul(%a, %b)
%r = aten::mul(%q, %c)
return (%r)""", """
graph(%a, %b, %c):
%r = my::fused_mulmul(%a, %b, %c)
return (%r)""", graph)
FileCheck().run(input_str, graph)
# Check add(mul(x,y),z) --> muladd(x,y,z) replacement
input_str = """
graph(%x, %y, %z):
# CHECK-NOT: aten::mul
# CHECK-NOT: aten::add
%c = prim::Const[value=1]()
%t = aten::mul(%x, %y)
%p = aten::add(%t, %z, %c)
# CHECK: my::muladd
# CHECK-NEXT: return
return (%p)"""
graph = parse_ir(input_str)
torch._C._jit_pass_custom_pattern_based_rewrite_graph("""
graph(%a, %b, %c, %d):
%q = aten::mul(%a, %b)
%r = aten::add(%q, %c, %d)
return (%r)""", """
graph(%a, %b, %c, %d):
%r = my::muladd(%a, %b, %c, %d)
return (%r)""", graph)
FileCheck().run(input_str, graph)
# Check add(mul(x,y),z) --> sub(add(x,y),z) replacement
input_str = """
graph(%x, %y, %z):
# CHECK-NOT: aten::mul
%c = prim::Const[value=1]()
# CHECK: aten::add
%t = aten::mul(%x, %y)
# CHECK-NEXT: aten::sub
%p = aten::add(%t, %z, %c)
# CHECK-NOT: aten::add
# CHECK-NEXT: return
return (%p)"""
graph = parse_ir(input_str)
torch._C._jit_pass_custom_pattern_based_rewrite_graph("""
graph(%a, %b, %c, %d):
%q = aten::mul(%a, %b)
%r = aten::add(%q, %c, %d)
return (%r)""", """
graph(%a, %b, %c, %d):
%q = aten::add(%a, %b, %d)
%r = aten::sub(%q, %c, %d)
return (%r)""", graph)
FileCheck().run(input_str, graph)
# Check mul(x,y) --> x replacement
input_str = """
graph(%x, %y, %z):
%c = prim::Const[value=1]()
# CHECK-NOT: aten::mul
%t = aten::mul(%x, %y)
# CHECK: aten::add(%x, %z
%p = aten::add(%t, %z, %c)
# CHECK-NEXT: return
return (%p)"""
graph = parse_ir(input_str)
torch._C._jit_pass_custom_pattern_based_rewrite_graph("""
graph(%Pa, %Pb):
%Pq = aten::mul(%Pa, %Pb)
return (%Pq)""", """
graph(%Ra, %Rb):
return (%Ra)""", graph)
FileCheck().run(input_str, graph)
@_tmp_donotuse_dont_inline_everything
def test_pattern_based_module_rewrite(self):
# Check match::module behavior
class Test(torch.nn.Module):
def __init__(self):
super(Test, self).__init__()
self.conv = torch.nn.Conv2d(1, 20, 5, 1)
self.bn = torch.nn.BatchNorm2d(num_features=20)
def forward(self, x):
x = self.conv(x)
x = self.bn(x)
return x
m = torch.jit.script(Test())
torch._C._jit_pass_custom_pattern_based_rewrite_graph("""
graph(%self, %x):
%conv = match::module[name="Conv2d"](%self)
%y = prim::CallMethod[name="forward"](%conv, %x)
%bn = match::module[name="BatchNorm2d"](%self)
%z = prim::CallMethod[name="forward"](%bn, %y)
return (%z)""", """
graph(%self, %x):
%z = my::matched_conv_bn(%self, %x)
return (%z)""", m._c._get_method("forward").graph)
FileCheck().check("my::matched_conv_bn").run(m._c._get_method("forward").graph)
def test_pattern_based_rewrite_with_source_range_preserved(self):
class TestModule1(torch.nn.Module):
def __init__(self):
super(TestModule1, self).__init__()
def forward(self, x, y, z, w):
x = x + y
x = x * z
return w - x
input_pattern = """
graph(%x, %y, %z, %const):
%t = aten::add(%x, %y, %const)
%o = aten::mul(%t, %z)
return (%o)"""
replacement_pattern = """
graph(%x, %y, %z, %const):
%o = my::add_mul(%x, %y, %z, %const)
return (%o)"""
scripted_model = torch.jit.script(TestModule1())
graph = scripted_model.graph
value_mappings = [("o", "t")]
for node in graph.nodes():
if node.kind() == "aten::add":
source_range_1 = node.sourceRange()
torch._C._jit_pass_custom_pattern_based_rewrite_graph(
input_pattern, replacement_pattern, scripted_model.graph, value_name_pairs=value_mappings)
graph = scripted_model.graph
for node in graph.nodes():
if node.kind() == "my::add_mul":
source_range_2 = node.sourceRange()
self.assertTrue(source_range_1 == source_range_2)
class TestModule2(torch.nn.Module):
def __init__(self):
super(TestModule2, self).__init__()
def forward(self, x, y, z, w):
x = x + y
x = x + z
x = x * z
x = x * w
return x - 2
# Check source range preservation for two node transforms add -> my_add
input_pattern = """
graph(%x, %y, %const):
%o = aten::add(%x, %y, %const)
return (%o)"""
replacement_pattern = """
graph(%x, %y, %const):
%o = my::add(%x, %y, %const)
return (%o)"""
scripted_model = copy.deepcopy(torch.jit.script(TestModule2()))
graph_copy = scripted_model.graph.copy()
value_mappings = [("o", "o")]
source_range_add_1 = None
for node in graph_copy.nodes():
if source_range_add_1 is None and node.kind() == "aten::add":
source_range_add_1 = node.sourceRange()
if source_range_add_1 is not None and node.kind() == "aten::add":
source_range_add_2 = node.sourceRange()
torch._C._jit_pass_custom_pattern_based_rewrite_graph(
input_pattern, replacement_pattern, graph_copy, value_name_pairs=value_mappings)
source_range_my_add_1 = None
for node in graph_copy.nodes():
if source_range_my_add_1 is None and node.kind() == "my::add":
source_range_my_add_1 = node.sourceRange()
if source_range_my_add_1 is not None and node.kind() == "my::add":
source_range_my_add_2 = node.sourceRange()
self.assertTrue(source_range_add_1 == source_range_my_add_1)
self.assertTrue(source_range_add_2 == source_range_my_add_2)
# Check source range preservation for add-add -> double_add transform
# fuse nodes
input_pattern = """
graph(%x, %y, %z, %const):
%t = aten::add(%x, %y, %const)
%o = aten::add(%t, %z, %const)
return (%o)"""
replacement_pattern = """
graph(%x, %y, %z, %const):
%o = my::double_add(%x, %y, %z, %const)
return (%o)"""
scripted_model = torch.jit.script(TestModule2())
graph_copy = scripted_model.graph.copy()
value_mappings = [("o", "t")]
source_range_1 = None
source_range_2 = None
for node in graph_copy.nodes():
if node.kind() == "aten::add":
source_range_1 = node.sourceRange()
break
torch._C._jit_pass_custom_pattern_based_rewrite_graph(
input_pattern, replacement_pattern, graph_copy, value_name_pairs=value_mappings)
for node in graph_copy.nodes():
if node.kind() == "my::double_add":
source_range_2 = node.sourceRange()
self.assertTrue(source_range_1 == source_range_2)
# Check source range preservation for mul -> add + add transform
# split node
input_pattern = """
graph(%x, %y):
%t = aten::mul(%x, %y)
return (%t)"""
replacement_pattern = """
graph(%x, %y):
%t = my::add(%x, %y)
%o = my::add(%t, %y)
return (%o)"""
scripted_model = torch.jit.script(TestModule2())
graph_copy = scripted_model.graph.copy()
value_mappings = [("t", "t"), ("o", "t")]
source_range_mul_1 = None
for node in graph_copy.nodes():
if source_range_mul_1 is None and node.kind() == "aten::mul":
source_range_mul_1 = node.sourceRange()
if source_range_mul_1 is not None and node.kind() == "aten::mul":
source_range_mul_2 = node.sourceRange()
torch._C._jit_pass_custom_pattern_based_rewrite_graph(
input_pattern, replacement_pattern, graph_copy, value_name_pairs=value_mappings)
source_range_add_1 = None
for node in graph_copy.nodes():
if source_range_add_1 is None and node.kind() == "my::add":
source_range_add_1 = node.sourceRange()
if source_range_add_1 is not None and node.kind() == "my::add":
source_range_add_2 = node.sourceRange()
self.assertTrue(source_range_mul_1 == source_range_add_1)
self.assertTrue(source_range_mul_2 == source_range_add_2)
# Check lack of source range preservation for mul-mul-> double_mul transform
input_pattern = """
graph(%x, %y, %z):
%t = aten::mul(%x, %y)
%o = aten::mul(%t, %z)
return (%o)"""
replacement_pattern = """
graph(%x, %y, %z):
%o = my::double_mul(%x, %y, %z)
return (%o)"""
scripted_model = torch.jit.script(TestModule2())
graph_copy = scripted_model.graph.copy()
for node in graph_copy.nodes():
if node.kind() == "aten::mul":
source_range_1 = node.sourceRange()
torch._C._jit_pass_custom_pattern_based_rewrite_graph(input_pattern, replacement_pattern, graph_copy)
for node in graph_copy.nodes():
if node.kind() == "my::double_mul":
source_range_2 = node.sourceRange()
self.assertFalse(source_range_1 == source_range_2)
def test_expand_quantlint(self):
pass
def test_expand_fold_quant_inputs(self):
pass
def test_shape_analysis_broadcast(self):
def broadcast(a, b):
return a + b
x = torch.randn(3, 1, 5, requires_grad=True)
y = torch.randn(4, 1, 8, 5, requires_grad=True)
graph = torch.jit.script(broadcast).graph
torch._C._jit_pass_complete_shape_analysis(graph, (x, y), False)
FileCheck().check("Double(4, 3, 8, 5, strides=[120, 40, 5, 1], device=cpu)").run(str(graph))
def test_shape_analysis_unsqueeze_in_loop(self):
input_str = """graph(%x.1 : Tensor):
%4 : bool = prim::Constant[value=1]()
%1 : int = prim::Constant[value=2]()
%7 : int = prim::Constant[value=0]()
# CHECK: FloatTensor(requires_grad=0, device=cpu) = prim::Loop
%x : Tensor = prim::Loop(%1, %4, %x.1)
# CHECK: : FloatTensor(requires_grad=0, device=cpu)):
block0(%i : int, %x.6 : Tensor):
# CHECK: FloatTensor(requires_grad=0, device=cpu) = aten::unsqueeze
%x.3 : Tensor = aten::unsqueeze(%x.6, %7)
-> (%4, %x.3)
return (%x)"""
graph = parse_ir(input_str)
torch._C._jit_pass_complete_shape_analysis(graph, (torch.zeros(2, 2, dtype=torch.float32),), False)
FileCheck().run(input_str, graph)
def test_script_tensor_type(self):
def foo(x, t: torch.dtype):
return x.type(t)
scr = torch.jit.script(foo)
x = torch.rand(3, 4)
for t in [torch.int8, torch.float64, torch.float32,
torch.bfloat16, torch.complex64, torch.complex128, torch.bool]:
self.assertEqual(scr(x, t), foo(x, t))
def test_shape_analysis_masked_select(self):
input_str = """graph(%0 : Float(),
%1 : Bool()):
# CHECK: Float(*, requires_grad=0, device=cpu) = aten::masked_select
%2 : Tensor = aten::masked_select(%0, %1) # test/test_jit.py:15261:0
return (%2)"""
graph = parse_ir(input_str)
x = torch.ones(1, dtype=torch.float32)[0]
mask = x.ge(0.5)
torch._C._jit_pass_complete_shape_analysis(graph, (x, mask), False)
FileCheck().run(input_str, graph)
# TODO: update verify to work with GraphExecutors
@unittest.skip("verify needs to be updated to work with GraphExecutors")
def test_verify(self):
x = torch.tensor([0.4], requires_grad=True)
y = torch.tensor([0.7], requires_grad=True)
@torch.jit.compile
def f(x, y):
z = torch.sigmoid(x * (x + y))
w = torch.abs(x * x * x + y) + Variable(torch.ones(1))
return z, w
torch.jit.verify(f, (x, y), loss_fn=lambda z, w: z * w, devices=[])
# TODO: adapt to a GraphExecutor test
@unittest.skip("Need to instrument GraphExecutors a bit more")
def test_flags(self):
x, y = torch.randn(2, 2)
y = Variable(torch.randn(2, 2))
@torch.jit.compile
def fn(x, y):
return (x * x + y * y + x * y).sum()
grads = {}
for rx, ry in product((True, False), repeat=2):
x.requires_grad = rx
y.requires_grad = ry
self.assertFalse(fn.has_trace_for(x, y))
out = fn(x, y)
self.assertFalse(fn.has_trace_for(x, y))
for v, name, compute in [(x, 'x', rx), (y, 'y', ry)]:
if not compute:
continue
grad_v, = torch.autograd.grad(out, v, retain_graph=True)
expected_grad = grads.setdefault(name, grad_v)
self.assertEqual(grad_v, expected_grad)
self.assertEqual(fn.has_trace_for(x, y), rx or ry)
def test_python_ir(self):
x = torch.tensor([0.4], requires_grad=True)
y = torch.tensor([0.7], requires_grad=True)
def doit(x, y):
return torch.sigmoid(torch.tanh(x * (x + y)))
g, _ = torch.jit._get_trace_graph(doit, (x, y))
self.run_pass('dce', g)
self.run_pass('canonicalize', g)
g2 = torch._C.Graph()
g_to_g2 = {}
for node in g.inputs():
g_to_g2[node] = g2.addInput()
for node in g.nodes():
n_ = g2.createClone(node, lambda x: g_to_g2[x])
g2.appendNode(n_)
for o, no in zip(node.outputs(), n_.outputs()):
g_to_g2[o] = no
for node in g.outputs():
g2.registerOutput(g_to_g2[node])
t_node = g2.create("prim::TensorTest").t_("a", torch.ones([2, 2]))
self.assertEqual(t_node.attributeNames(), ["a"])
g2.appendNode(t_node)
self.assertTrue(torch.equal(torch.ones(2, 2), t_node.t("a")))
for node in g.nodes():
self.assertTrue(g2.findNode(node.kind()) is not None)
def test_permute_inputs_binding(self):
@torch.jit.script
def foo(i, j, k):
pass
g = foo.graph
idxs = []
for i, inp in enumerate(g.inputs()):
inp.setDebugName(f"inp{i}")
idxs.append(i)
permuted_idxs = list(np.random.permutation(idxs))
g.permuteInputs(permuted_idxs)
for i, inp in enumerate(g.inputs()):
self.assertEqual(f"inp{permuted_idxs[i]}", inp.debugName())
@unittest.skipIf(IS_MACOS, "Failing on MacOS only")
def test_python_ir_utils(self):
@torch.jit.script
def foo(inp):
x = inp + 1
y = x / 2
z = y * y
return z
add_node = foo.graph.findNode("aten::add")
div_node = foo.graph.findNode("aten::div")
with foo.graph.insert_point_guard(add_node):
with foo.graph.insert_point_guard(div_node):
foo.graph.insertConstant("goodbye")
foo.graph.insertConstant("hello")
with foo.graph.insert_point_guard(foo.graph.findNode("aten::mul")):
foo.graph.insertConstant("hello")
FileCheck().check("hello").check("goodbye").check("hello").run(foo.graph)
self.assertTrue(add_node.matches(add_node.schema()))
self.assertFalse(add_node.matches(div_node.schema()))
def test_python_ir_utils_graph(self):
@torch.jit.script
def unrolled_mul(x: torch.Tensor, y: int):
out = x
for _ in range(y - 1):
out = out + x
return out
@torch.jit.script
def foo(x):
return x * 4
g = foo.graph
muls = g.findAllNodes("aten::mul")
scalar_muls = filter(lambda x: x.matches("aten::mul(Tensor self, Scalar other) -> Tensor"), muls)
mul_constant_int = filter(lambda x: isinstance(list(x.inputs())[1].toIValue(), int), scalar_muls)
for mul in mul_constant_int:
with g.insert_point_guard(mul):
outputs = g.insertGraph(unrolled_mul.graph, list(mul.inputs()))
assert len(outputs) == len(list(mul.outputs()))
for new_out, old_out in zip(outputs, g.outputs()):
old_out.replaceAllUsesWith(new_out)
mul.destroy()
FileCheck().check_not("aten::mul").check("aten::add").run(foo.graph)
self.assertEqual(foo(torch.ones([2, 2])), torch.ones([2, 2]) * 4)
@unittest.skipIf(IS_SANDCASTLE, "gtest runs these in sandcastle")
@unittest.skipIf(RUN_CUDA, "covered by test_cpp_cuda")
@unittest.skipIf(not torch._C._jit_has_cpp_tests(), "Tests were not built, use BUILD_TEST=1")
def test_cpp(self):
from cpp.jit import tests_setup
tests_setup.setup()
torch._C._jit_run_cpp_tests()
tests_setup.shutdown()
def test_batchnorm(self):
x = torch.ones(2, 2, 2, 2)
g, outputs, inputs = torch.jit._get_trace_graph(nn.BatchNorm2d(2), x,
_force_outplace=True, return_inputs=True)
m = self.createFunctionFromGraph(g)
self.assertEqual(outputs, m(*inputs))
def test_dropout(self):
x = torch.ones(2, 2)
with torch.random.fork_rng(devices=[]):
g, outputs, inputs = torch.jit._get_trace_graph(nn.Dropout(0.6), x, return_inputs=True)
with torch.random.fork_rng(devices=[]):
m = self.createFunctionFromGraph(g)
self.assertEqual(outputs, m(*inputs))
@unittest.skipIf(not RUN_CUDA, "test requires CUDA")
@unittest.skipIf(GRAPH_EXECUTOR != ProfilingMode.PROFILING, "skip if profiling isn't enabled")
def test_native_dropout_corner_case(self):
with disable_autodiff_subgraph_inlining():
def t(x, p: float, t: bool):
o = torch.dropout(x, p, t)
return o
jit_t = torch.jit.script(t)
x = torch.randn(5).requires_grad_()
FileCheck().check("prim::DifferentiableGraph").run(jit_t.graph_for(x, 1.0, True, profile_and_replay=True))
for train in [True, False]:
for p in [0.0, 1.0]:
for device in ["cuda", "cpu"]:
x = torch.randn(5).to(device=device).requires_grad_()
x_ref = x.detach().requires_grad_()
o = jit_t(x, p, train)
o_ref = t(x_ref, p, train)
o.sum().backward()
o_ref.sum().backward()
assert(o.equal(o_ref))
assert(x.grad.equal(x_ref.grad))
@slowTest
@unittest.skipIf(GRAPH_EXECUTOR != ProfilingMode.LEGACY, 'Testing differentiable graph')
def test_dropout_module_requires_grad(self):
with enable_profiling_mode_for_profiling_tests():
class MyModule(torch.nn.Module):
def __init__(self, M):
super(MyModule, self).__init__()
self.dropout = torch.nn.Dropout(0.5)
self.linear = torch.nn.Linear(M, M)
def forward(self, input):
input = self.dropout(input)
output = self.linear(input)
return output
def profile(func, X):
with torch.autograd.profiler.profile() as prof:
func(X)
return [e.name for e in prof.function_events]
M = 1000
scripted = torch.jit.script(MyModule(M))
# To reduce confusion about expected behaviors:
# requires_grad controls whether dropout is symbolically differentiated.
# training controls whether bernoulli_ is called inside symbolic differentiation of dropout.
# * When requires_grad == training, the expected behaviors are obvious.
# * When requires_grad=True and training=False, bernoulli_ might still show up in the graph.
# But it's in a branch that's not called. That's why we have separate checks for autograd
# profiler to make sure it's not run.
# * When requires_grad=False and training=True, bernoulli_ must be run since it's the expected
# behavior for the dropout layer in training mode. It's independent of whether graph requires
# gradient. In fact bernoulli_ comes from autograd instead of autodiff in this case.
for training in (True, False):
if training:
scripted.train()
else:
scripted.eval()
for requires_grad in (True, False):
X = torch.randn(M, M, requires_grad=requires_grad)
if requires_grad:
FileCheck().check("aten::native_dropout").run(scripted.graph_for(X, profile_and_replay=True))
self.assertEqual(training, 'aten::bernoulli_' in profile(scripted, X))
@unittest.skipIf(GRAPH_EXECUTOR == ProfilingMode.SIMPLE, 'Testing differentiable graph')
@skipIfTorchDynamo("Torchdynamo cannot correctly handle profiler.profile calls")
def test_dropout_func_requires_grad(self):
def dropout_training(input):
return F.dropout(input, 0.5, training=True)
def dropout_eval(input):
return F.dropout(input, 0.5, training=False)
def profile(func, X):
with torch.autograd.profiler.profile() as prof:
func(X)
return [e.name for e in prof.function_events]
M = 1000
scripted_training = torch.jit.script(dropout_training)
scripted_eval = torch.jit.script(dropout_eval)
# See comments in test_dropout_module_requires_grad.
with disable_autodiff_subgraph_inlining():
for requires_grad in (True, False):
X = torch.randn(M, M, requires_grad=requires_grad)
if requires_grad:
FileCheck().check("aten::native_dropout").run(scripted_training.graph_for(X, profile_and_replay=True))
self.assertIn('aten::bernoulli_', profile(scripted_training, X))
self.assertNotIn('aten::bernoulli_', profile(scripted_eval, X))
@unittest.skipIf(not RUN_CUDA, "test_dropout_cuda require CUDA")
def test_dropout_cuda(self):
# Dropout AD is dispatched to _fused_dropout in CUDA case,
# which is not included in TestJitGeneratedFunctional
def _zero_rate(t):
return torch.true_divide((t == 0).sum(), t.numel())
x = torch.ones(1000, 1000).cuda().requires_grad_()
with enable_profiling_mode_for_profiling_tests():
@torch.jit.script
def func(x):
return torch.nn.functional.dropout(x)
with freeze_rng_state():
out_ref = torch.nn.functional.dropout(x)
grad_ref = torch.autograd.grad(out_ref.sum(), x)
with freeze_rng_state():
out = func(x)
grad = torch.autograd.grad(out.sum(), x)
# TODO(#40882): previously we assert exact matches between eager and JIT result:
# self.assertEqual(out, out_ref)
# self.assertEqual(grad, grad_ref)
# This test was disabled during legacy -> profiling executor transition.
# Currently JIT fused results doesn't match eager result exactly due to some changes merged in between.
# We temporarily only check statstical difference but it should be reverted once the issue is fixed.
self.assertEqual(_zero_rate(out), _zero_rate(out_ref), rtol=1e-3, atol=1e-4)
self.assertEqual(_zero_rate(grad[0]), _zero_rate(grad_ref[0]), rtol=1e-3, atol=1e-4)
def test_torch_ops_overloaded(self):
with self.assertRaisesRegex(RuntimeError, "failed to many any schema"):
torch.ops.aten.add("a", 1)
self.assertEqual("ab", torch.ops.aten.add("a", "b"))
a, b = torch.rand(3, 4), torch.rand(3, 4)
self.assertEqual(a + b, torch.ops.aten.add(a, b))
self.assertEqual(a + 1, torch.ops.aten.add(a, 1))
def test_torch_ops_kwonly(self):
a, b = torch.rand(3, 4), torch.rand(3, 4)
with self.assertRaisesRegex(RuntimeError, "positional argument"):
torch.ops.aten.add(a, b, 2)
# h/t Chillee for this ambiguous case
self.assertEqual(a.prod(1), torch.ops.aten.prod(a, 1))
def test_torch_complex(self):
def fn(real, img):
return torch.complex(real, img)
def fn_out(real, img, out):
return torch.complex(real, img, out=out)
self.checkScript(fn, (torch.rand(3, 4), torch.rand(3, 4), ))
self.checkScript(fn, (torch.ones(5, 1, 4), torch.ones(5, 1, 4), ))
self.checkScript(fn, (torch.zeros(1, 6), torch.ones(6, 1), ))
self.checkScript(fn, (torch.zeros(1, 6), torch.zeros(6, 1), ))
self.checkScript(fn, (torch.empty(3, 4), torch.empty(3, 4), ))
real = torch.tensor([1, 2], dtype=torch.float32)
img = torch.tensor([3, 4], dtype=torch.float32)
out = torch.empty([3, 4], dtype=torch.complex64)
self.checkScript(fn_out, (real, img, out, ))
real = torch.tensor([5, 2], dtype=torch.float64)
img = torch.tensor([3, 4], dtype=torch.float64)
out = torch.empty([5, 2], dtype=torch.complex128)
self.checkScript(fn_out, (real, img, out, ))
real = torch.ones([1, 2])
img = torch.ones([1, 2])
out = torch.empty([1, 2], dtype=torch.complex128)
self.checkScript(fn_out, (real, img, out, ))
real = torch.ones([3, 8, 7])
img = torch.ones([3, 8, 7])
out = torch.empty([3, 8, 7], dtype=torch.complex128)
self.checkScript(fn_out, (real, img, out, ))
real = torch.empty([3, 2, 6])
img = torch.empty([3, 2, 6])
out = torch.empty([3, 2, 6], dtype=torch.complex128)
self.checkScript(fn_out, (real, img, out, ))
real = torch.zeros([1, 3])
img = torch.empty([3, 1])
out = torch.empty([3, 3], dtype=torch.complex128)
self.checkScript(fn_out, (real, img, out, ))
real = torch.ones([2, 5])
img = torch.empty([2, 1])
out = torch.empty([2, 5], dtype=torch.complex128)
self.checkScript(fn_out, (real, img, out, ))
real = torch.ones([2, 5])
img = torch.zeros([2, 1])
out = torch.empty([2, 5], dtype=torch.complex128)
self.checkScript(fn_out, (real, img, out, ))
def test_einsum(self):
def check(fn, jitted, *args):
self.assertGraphContains(jitted.graph, kind='aten::einsum')
self.assertEqual(fn(*args), jitted(*args))
def equation_format(x, y):
return torch.einsum('i,j->ij', (x, y))
def equation_format_varargs(x, y):
return torch.einsum('i,j->ij', x, y)
def sublist_format(x, y):
return torch.einsum(x, [0], y, [1], [0, 1])
x = make_tensor((5,), dtype=torch.float32, device="cpu")
y = make_tensor((10,), dtype=torch.float32, device="cpu")
for fn in [equation_format, equation_format_varargs, sublist_format]:
check(fn, torch.jit.script(fn), x, y)
check(fn, torch.jit.trace(fn, (x, y)), x, y)
@skipIfTorchDynamo("TorchDynamo fails with unknown reason")
def test_python_ivalue(self):
# Test if pure python object can be hold as IValue and conversion
# between IValue and PyObject are correct
# test for numpy object
py_array = np.arange(15)
ret_py_obj = torch._C._ivalue_debug_python_object(py_array)
self.assertEqual(py_array, ret_py_obj)
# test for function object
ret_py_obj = torch._C._ivalue_debug_python_object(F.relu)
self.assertEqual(F.relu, ret_py_obj)
# test for memory management
# we need to ensure IValue correctly call incref/decref to avoid
# dangling behavior and potential memory leaks during conversions
def test_func_scope_helper(inp):
# create a scope and do the conversion -> ivalue -> pyobject
# this func return a new pyobject that refcount + 1
inp_refcount = sys.getrefcount(inp)
ivalue_holder = torch._C._ivalue_debug_python_object(inp)
self.assertEqual(inp_refcount + 1, sys.getrefcount(ivalue_holder))
return ivalue_holder + 1
test_input = 2200
before_count = sys.getrefcount(test_input)
test_func_scope_helper(test_input)
after_count = sys.getrefcount(test_input)
# after the test_func_scope_helper_call, the refcount of
# test_input should be equal to the original refcount
# otherwise we get either dangling pointer or memory leak!
self.assertEqual(before_count, after_count)
def test_decompose_addmm(self):
def does_decompose():
@torch.jit.script
def addmm(mat, mat1, mat2):
a = mat.addmm(mat1, mat2)
b = mat.addmm(mat1, mat2, alpha=1.0, beta=1.0)
return a + b
mat = torch.randn(2, 2)
mat1 = torch.randn(2, 4)
mat2 = torch.randn(4, 2)
out_ref = addmm(mat, mat1, mat2)
self.run_pass('decompose_ops', addmm.graph)
out_test = addmm(mat, mat1, mat2)
self.assertEqual(out_ref, out_test)
FileCheck().check_not("addmm").run(str(addmm.graph))
def doesnt_decompose():
@torch.jit.script
def addmm(mat, mat1, mat2, alpha, beta):
a = mat.addmm(mat1, mat2, alpha=4.20, beta=2.0)
b = mat.addmm(mat1, mat2, alpha=int(alpha), beta=int(beta))
return a + b
orig = str(addmm.graph)
self.run_pass('decompose_ops', addmm.graph)
self.assertTrue(orig == str(addmm.graph))
does_decompose()
doesnt_decompose()
@suppress_warnings
def test_sparse_tensors(self):
@torch.jit.ignore
def get_sparse():
return torch.sparse.FloatTensor(2, 3)
@torch.jit.script
def test_is_sparse(input):
# type: (Tensor) -> bool
return input.is_sparse
script_out_is_sparse = test_is_sparse(get_sparse())
script_out_is_dense = test_is_sparse(torch.randn(2, 3))
self.assertEqual(script_out_is_sparse, True)
self.assertEqual(script_out_is_dense, False)
def test_basic_sparse(input):
output = get_sparse()
return output, input
self.checkScript(test_basic_sparse, (get_sparse(),))
self.checkScript(test_basic_sparse, (torch.tensor([1]),))
def test_sparse_sum(input):
return torch.sparse.sum(input)
self.checkScript(test_sparse_sum, (get_sparse(),))
def test_sparse_mm(input1, input2):
return torch.sparse.mm(input1, input2)
self.checkScript(test_sparse_mm, (get_sparse(), torch.randn(3, 4)))
def test_sparse_addmm(input, input1, input2):
return torch.sparse.addmm(input, input1, input2)
def test_sparse_addmm_alpha_beta(input, input1, input2):
return torch.sparse.addmm(input, input1, input2, alpha=1.3, beta=1.5)
self.checkScript(test_sparse_addmm, (torch.randn(2, 4), get_sparse(), torch.randn(3, 4)))
self.checkScript(test_sparse_addmm_alpha_beta, (torch.randn(2, 4), get_sparse(), torch.randn(3, 4)))
@suppress_warnings
def test_sparse_csr_tensors(self):
@torch.jit.ignore
def get_sparse_csr():
return torch.randn(3, 3).to_sparse_csr()
@torch.jit.script
def test_is_sparse_csr(input):
# type: (Tensor) -> bool
return input.is_sparse_csr
script_out_is_sparse_csr = test_is_sparse_csr(get_sparse_csr())
script_out_is_dense_csr = test_is_sparse_csr(torch.randn(3, 3))
self.assertEqual(script_out_is_sparse_csr, True)
self.assertEqual(script_out_is_dense_csr, False)
@unittest.skipIf(not RUN_CUDA, "requires CUDA")
def test_device_not_equal(self):
def compare_device(x: torch.device):
return x != torch.device("cuda:0")
def compare_two_device(x: torch.device, y: torch.device):
return x != y
self.checkScript(compare_device, (torch.device("cuda:0"),))
self.checkScript(compare_two_device, (torch.device("cuda:0"), torch.device("cuda:1"), ))
def test_constant_prop_simple(self):
@torch.jit.script
def constant_prop(input_int):
# type: (int) -> int
a = 2 * 3
b = a + 2
return b - input_int
out_ref = constant_prop(2)
self.run_pass('constant_propagation', constant_prop.graph)
out_test = constant_prop(2)
self.assertEqual(out_ref, out_test)
graph_str = str(constant_prop.graph)
self.assertTrue("aten::add" not in graph_str and "aten::mul" not in graph_str)
const = constant_prop.graph.findNode("prim::Constant").output().toIValue()
self.assertEqual(const, 8)
def test_constant_prop_nested(self):
@torch.jit.script
def constant_prop(a):
b = 2 + 1
if bool(a < 2):
c = b + 2
else:
c = b - 2
return c
out_ref = constant_prop(torch.tensor(2))
self.run_pass('constant_propagation', constant_prop.graph)
out_test = constant_prop(torch.tensor(2))
self.assertEqual(out_ref, out_test)
if_node = constant_prop.graph.findNode("prim::If")
for block in if_node.blocks():
for node in block.nodes():
self.assertTrue(node.kind() == "prim::Constant")
def test_constant_prop_print(self):
@torch.jit.script
def constant_prop(input_tensor):
a = 2 * 3
print(a)
b = a + 2
return b + input_tensor
self.run_pass('constant_propagation', constant_prop.graph)
graph = constant_prop.graph
print_node = graph.findNode("prim::Print")
self.assertTrue(print_node.input().toIValue() == 6)
def test_constant_prop_rand(self):
@torch.jit.script
def constant_prop():
a = torch.randn([3])
b = a + 2
return b
self.run_pass('constant_propagation', constant_prop.graph)
self.assertTrue("aten::randn" in str(constant_prop.graph))
def test_constant_prop_none(self):
@torch.jit.script
def typed_none():
# type: () -> Optional[int]
return None
@torch.jit.script
def constant_prop():
a = typed_none()
b = typed_none()
if (a is None and b is None):
a = 2
else:
a = 1
return a
self.run_pass('constant_propagation', constant_prop.graph)
FileCheck().check("prim::Constant").run(constant_prop.graph)
def test_constant_prop_if_inline(self):
@torch.jit.script
def constant_prop():
cond = True
a = 1
if cond:
a = 1 * 2
else:
a = 1 // 0
return a
# testing that 1 // 0 error is not thrownn
self.run_pass('constant_propagation', constant_prop.graph)
def test_constant_prop_exception(self):
# checking y = a[4] does not error in constant propagation
def bad_index(x):
# type: (bool)
y = 0
if x:
a = [1, 2, 3]
y = a[4]
return y
self.checkScript(bad_index, (False,))
def test_constant_prop_aliasing_type(self):
@torch.jit.script
def foo():
return len([1]), len(torch.tensor([2]))
FileCheck().check_dag("aten::tensor").check_dag("aten::len").run(foo.graph)
@torch.jit.script
def fn():
if 1 == 1:
return 1
else:
return 2
FileCheck().check_not("prim::If").run(fn.graph)
def test_unchecked_cast(self):
def test(cond):
# type: (bool)
a = torch.tensor([10])
if cond:
b = None
else:
b = a
if b is not None:
b[0] = 5
return a.int()
self.checkScript(test, (True,))
self.checkScript(test, (False,))
def test_constant_prop_if_constant(self):
@torch.jit.script
def constant_prop(a, b):
c0 = 1
c1 = 1
c2 = 1
if bool(a): # -> c0, c1
if bool(b): # -> c0
if 1 == 1: # -> c0
c0 = c0 + 1
if 1 == 2:
c1 = c1 + 1
c2 = c2 + 1
else: # -> c0, c1
c1 = c1 + 1
if 1 == 1: # inlined
c0 = c0 + 1 # dynamic
c2 = c2 + 4 # set to 5
return a + c0 + c1 + c2
graph = constant_prop.graph
self.run_pass('constant_propagation', graph)
ifs = graph.findAllNodes("prim::If", recurse=False)
snd_if_inlined = len(ifs) == 1
self.assertTrue(snd_if_inlined)
first_if = ifs[0]
self.assertTrue(first_if.outputsSize() == 2)
second_if = first_if.findNode("prim::If", recurse=False)
self.assertTrue(second_if.outputsSize() == 1)
self.assertTrue(second_if.findNode("prim::If") is None)
def test_constant_prop_loop_constant(self):
@torch.jit.script
def constant_prop(cond, iter):
# type: (bool, int) -> int
b = 0
while True:
print("stays")
for _ in range(2):
print("stays")
for _ in range(iter):
print("stays")
while cond:
print("stays")
while False:
print("removed")
for _i in range(0):
print("removed")
for _i in range(-4):
print("removed")
return b
self.run_pass('constant_propagation', constant_prop.graph)
graph = canonical(constant_prop.graph)
self.assertTrue(graph.count("removed") == 0)
self.assertTrue(graph.count("stays") == 1) # constant gets pooled
self.assertTrue(graph.count("prim::Print") == 4)
def test_constant_prop_remove_output(self):
@torch.jit.script
def constant_prop(iter):
# type: (int) -> None
a = 1
b = 1
c = 1
for i in range(iter):
if 1 == 2:
a = 10
if i == 5:
b = 2
c = 3
print(a, b, c)
graph = constant_prop.graph
self.run_pass('constant_propagation', graph)
self.assertTrue(graph.findNode("prim::Loop").outputsSize() == 2)
# TODO(gmagogsfm): Refactor this test to reduce complexity.
def test_constant_insertion(self):
funcs_template = dedent('''
def func():
return {constant_constructor}
''')
# constants: primitives: int, double, bool, str, lists of primitives,
# and tuples
def check_constant(constant_constructor):
scope = {}
funcs_str = funcs_template.format(constant_constructor=constant_constructor)
execWrapper(funcs_str, globals(), scope)
cu = torch.jit.CompilationUnit(funcs_str)
f_script = cu.func
self.run_pass('constant_propagation', f_script.graph)
FileCheck().check_count("prim::Constant", 1, exactly=True).run(f_script.graph)
self.assertEqual(scope['func'](), f_script())
imported = self.getExportImportCopy(f_script)
self.assertEqual(imported(), f_script())
constants = ["None", "-.5", "0", "1", "True", "False", "''", "'a'", "'b'", "torch.tensor(1)",
"[True, False]", "[0., .5]", "[torch.tensor(4), torch.tensor(2)]", "[0, 1]", "['0', '1']",
"[True, None]", "[.5, None, .2]"]
for type in ["Tensor", "str", "int", "float", "bool"]:
constants.append("torch.jit.annotate(List[ " + type + "], [])")
for constant in constants:
check_constant(constant)
for key_type in ["str", "int", "float"]:
for value_type in ["Tensor", "bool", "str", "int", "float"]:
check_constant("torch.jit.annotate(Dict[ " + key_type + ", " + value_type + "], {})")
check_constant("torch.jit.annotate(Dict[ " + key_type + ", Optional[" + value_type + "]], {})")
for i in range(len(constants)):
for j in range(i + 1, len(constants)):
tup_constant = constants[i] + ", " + constants[j]
check_constant(tup_constant)
dict_constants = []
for i in range(len(constants)):
# check_constant constructs the second dict with another Tensor
# which fails the comparison
if not isinstance(eval(constants[i]), (str, int, float)):
continue
for j in range(len(constants)):
dict_constant = "{ " + constants[i] + ": " + constants[j] + "}"
check_constant(dict_constant)
dict_constants.append(dict_constant)
constants = constants + dict_constants
# testing node hashing
funcs_template = dedent('''
def func():
print({constant_constructor})
''')
single_elem_tuples = ("(" + x + ",)" for x in constants)
input_arg = ", ".join(single_elem_tuples)
scope = {}
funcs_str = funcs_template.format(constant_constructor=input_arg)
execWrapper(funcs_str, globals(), scope)
cu = torch.jit.CompilationUnit(funcs_str)
f_script = cu.func
self.run_pass('constant_propagation', f_script.graph)
# prim::None return adds one constant
self.assertEqual(len(constants) + 1, str(f_script.graph).count("prim::Constant"))
self.run_pass('cse', f_script.graph)
# node hashing correctly working, no CSE occurs
self.assertEqual(len(constants) + 1, str(f_script.graph).count("prim::Constant"))
funcs_template = dedent('''
def func():
a = {constant_constructor}
print(a)
b = {constant_constructor}
print(b)
''')
# generate dicts with built-in types (excluding torch.Tensor)
xprod = itertools.product(constants, constants)
# test that equal tuples and dicts correctly work with node hashing
for tup in ("(" + x + ",)" for x in constants):
funcs_str = funcs_template.format(constant_constructor=tup)
scope = {}
execWrapper(funcs_str, globals(), scope)
cu = torch.jit.CompilationUnit(funcs_str)
f_script = cu.func
self.run_pass('constant_propagation_immutable_types', f_script.graph)
num_constants = str(f_script.graph).count("prim::Constant")
self.run_pass('cse', f_script.graph)
FileCheck().check_count("prim::Constant", num_constants, exactly=True).run(f_script.graph)
@unittest.skipIf(not RUN_CUDA, "requires CUDA")
def test_cuda_export_restore(self):
class Sub(torch.jit.ScriptModule):
def __init__(self):
super(Sub, self).__init__()
self.weight = nn.Parameter(torch.randn(3, 4))
@torch.jit.script_method
def forward(self, thing):
return self.weight + thing
class M(torch.jit.ScriptModule):
def __init__(self):
super(M, self).__init__()
self.mod = Sub()
@torch.jit.script_method
def forward(self, v):
return self.mod(v)
m = M()
m.cuda()
m2 = self.getExportImportCopy(m)
m2.cuda()
input = torch.rand(3, 4).cuda()
self.assertEqual(m(input), m2(input))
@slowTest
def test_export_batchnorm(self):
for mode in ['eval', 'train']:
for clazz in [
torch.nn.BatchNorm1d(100),
torch.nn.BatchNorm1d(100, affine=False),
torch.nn.BatchNorm2d(100),
torch.nn.BatchNorm2d(100, affine=False)]:
getattr(clazz, mode)()
input = torch.randn(20, 100) if isinstance(clazz, torch.nn.BatchNorm1d) else \
torch.randn(20, 100, 35, 45)
traced = torch.jit.trace(clazz, (input,))
imported = self.getExportImportCopy(traced)
x = torch.randn(20, 100) if isinstance(clazz, torch.nn.BatchNorm1d) else \
torch.randn(20, 100, 35, 45)
self.assertEqual(traced(x), imported(x))
def test_export_rnn(self):
for clazz in [nn.RNN(10, 20, 2), nn.GRU(10, 20, 2)]:
class RNNTest(torch.nn.Module):
def __init__(self):
super(RNNTest, self).__init__()
self.rnn = clazz
def forward(self, x, lengths, h0):
packed = torch.nn.utils.rnn.pack_padded_sequence(x, lengths)
out, h = self.rnn(packed, h0)
padded_outs, _ = torch.nn.utils.rnn.pad_packed_sequence(out)
return padded_outs
test = RNNTest()
traced = torch.jit.trace(test, (torch.randn(5, 3, 10), torch.LongTensor([3, 2, 1]), torch.randn(2, 3, 20)))
imported = self.getExportImportCopy(traced)
# NB: We make sure to pass in a batch with a different max sequence
# length to ensure that the argument stashing for pad_packed works
# properly.
x, lengths, h0 = torch.randn(7, 4, 10), torch.LongTensor([7, 3, 2, 1]), torch.randn(2, 4, 20)
self.assertEqual(traced(x, lengths, h0), imported(x, lengths, h0))
def test_export_lstm(self):
class LSTMTest(torch.nn.Module):
def __init__(self):
super(LSTMTest, self).__init__()
self.rnn = nn.LSTM(10, 20, 2)
def forward(self, x, lengths, hiddens):
h0, c0 = hiddens
packed = torch.nn.utils.rnn.pack_padded_sequence(x, lengths)
out, (h, c) = self.rnn(packed, (h0, c0))
padded_outs, _ = torch.nn.utils.rnn.pad_packed_sequence(out)
return padded_outs
test = LSTMTest()
traced = torch.jit.trace(test, (torch.randn(5, 3, 10),
torch.LongTensor([3, 2, 1]),
(torch.randn(2, 3, 20), torch.randn(2, 3, 20))))
imported = self.getExportImportCopy(traced)
x, lengths, h0, c0 = \
torch.randn(7, 3, 10), torch.LongTensor([7, 5, 2]), torch.randn(2, 3, 20), torch.randn(2, 3, 20)
self.assertEqual(traced(x, lengths, (h0, c0)), imported(x, lengths, (h0, c0)))
def test_unique_state_dict(self):
class MyModule(torch.nn.Module):
def __init__(self):
super(MyModule, self).__init__()
shared_param = torch.nn.Parameter(torch.ones(1))
self.register_parameter('w1', shared_param)
self.register_parameter('w2', shared_param)
def forward(self, input):
return input + self.w1 + self.w2
model = MyModule()
unittest.TestCase.assertEqual(
self, len(torch.jit._unique_state_dict(model, keep_vars=False)), 1)
unittest.TestCase.assertEqual(
self, len(torch.jit._unique_state_dict(model, keep_vars=True)), 1)
def test_export_dropout(self):
test = torch.nn.Dropout()
test.eval()
traced = torch.jit.trace(test, (torch.rand(3, 4),), check_trace=False)
imported = self.getExportImportCopy(traced)
x = torch.randn(3, 4)
self.assertEqual(traced(x), imported(x))
def test_pretty_printer(self):
@torch.jit.script
def if_test(a, b):
# FIXME: use 0 instead of a.
# c = 0
c = a
if bool(a < b):
c = b
else:
c = a
return c
@torch.jit.script
def if_one(a, b):
c = b
if bool(a < b):
c = a
return c
@torch.jit.script
def while_test(a, i):
while bool(i < 3):
a *= a
i += 1
return a
@torch.jit.script
def while_if_test(a, b):
c = 0
while bool(a < 10):
a = a + 1
b = b + 1
if bool(a > b):
c = 2
else:
c = 3
return a + 1 + c
@torch.jit.script
def loop_use_test(y):
x = y + 1
z = x + 5
while bool(y < 8):
y += 1
z = x
return x, z
@torch.jit.ignore
def python_fn(x):
return x + 10
@torch.jit.script
def python_op_name_test(y):
return python_fn(y)
@torch.jit.script
def empty_int_list_test(y):
x = torch.jit.annotate(List[int], [])
return x[0]
@torch.jit.script
def empty_float_list_test(y):
return [1.0, 2.0, 3.0]
@torch.jit.script
def print_weird_test(y):
print("hi\016")
self.assertExpected(if_test.code, "if_test")
self.assertExpected(if_one.code, "if_one")
self.assertExpected(while_test.code, "while_test")
self.assertExpected(while_if_test.code, "while_if_test")
self.assertExpected(loop_use_test.code, "loop_use_test")
self.assertExpected(python_op_name_test.code, "python_op_name_test")
self.assertExpected(empty_int_list_test.code, "empty_int_list_test")
self.assertExpected(empty_float_list_test.code, "empty_float_list_test")
self.assertExpected(print_weird_test.code, "print_weird_test")
def test_cu_escaped_number(self):
cu = torch.jit.CompilationUnit('''
def foo(a):
print("hi\016")
''')
self.assertExpected(cu.foo.code)
def test_import_method(self):
with torch._jit_internal._disable_emit_hooks():
class Foo(torch.jit.ScriptModule):
def __init__(self):
super(Foo, self).__init__()
@torch.jit.script_method
def forward(self, x, y):
return 2 * x + y
foo = Foo()
buffer = io.BytesIO()
torch.jit.save(foo, buffer)
buffer.seek(0)
foo_loaded = torch.jit.load(buffer)
self.assertExpected(foo_loaded.forward.code)
@unittest.skip("temporarily disable the test for fwd compatibility")
def test_non_ascii_string(self):
class Foo(torch.jit.ScriptModule):
def __init__(self):
super(Foo, self).__init__()
self.a = "Over \u0e55\u0e57 57"
@torch.jit.script_method
def forward(self, x, y):
return self.a + "hi\xA1"
foo = Foo()
buffer = io.BytesIO()
torch.jit.save(foo, buffer)
buffer.seek(0)
foo_loaded = torch.jit.load(buffer)
self.assertExpected(foo_loaded.forward.code)
def test_function_default_values(self):
outer_var = torch.tensor(20)
outer_var2 = torch.tensor(30)
a = torch.tensor(0.5)
b = torch.tensor(10)
@torch.jit.script
def simple_fn(x, a=a, b=b, c=outer_var + outer_var2):
return x + a + b + c
self.assertEqual(
simple_fn(torch.ones(1)),
torch.ones(1) + 0.5 + 10 + (20 + 30))
self.assertEqual(
simple_fn(torch.ones(1), torch.tensor(1), torch.tensor(3), torch.tensor(4)),
torch.ones(1) + 1 + 3 + 4)
outer_c = torch.tensor(9)
outer_flag = torch.tensor(False)
@torch.jit.script
def bool_fn(x, a=outer_c, flag=outer_flag):
if bool(flag):
result = x
else:
result = x + a
return result
self.assertEqual(bool_fn(torch.ones(1)), torch.ones(1) + 9)
self.assertEqual(
bool_fn(torch.ones(1), torch.tensor(1), torch.tensor(True)),
torch.ones(1))
@torch.jit.script
def none_fn(x=None):
# type: (Optional[int]) -> Optional[int]
return x
self.assertEqual(none_fn(), None)
self.assertEqual(none_fn(1), 1)
@torch.jit.script
def hints(x, a=0.5, b=10):
# type: (Tensor, float, int) -> Tensor
return x + a + b
self.assertEqual(hints(torch.ones(1)), torch.ones(1) + 0.5 + 10)
with self.assertRaisesRegex(RuntimeError, "Expected a default value"):
@torch.jit.script
def hints_bad_types(x, a=10, b=0.5): # noqa: T484
# type: (Tensor, float, int) -> Tensor
return x + a + b
with self.assertRaisesRegex(RuntimeError, "Expected a default value"):
@torch.jit.script
def bad_no_optional(x=None):
# type: (Dict[str, int]) -> Dict[str, int]
return x
def test_module_default_values(self):
four = torch.tensor(4)
class Test(torch.jit.ScriptModule):
def __init__(self):
super(Test, self).__init__()
@torch.jit.script_method
def forward(self, input, other=four):
return input + other
t = Test()
self.assertEqual(t(torch.ones(1)), torch.ones(1) + 4)
def test_mutable_default_values(self):
with self.assertRaisesRegex(Exception, "Mutable default parameters"):
@torch.jit.script
def foo(x=(1, [])):
# type: (Tuple[int, List[Tensor]])
return x
class Test(torch.nn.Module):
def forward(self, input=[]): # noqa: B006
return input
with self.assertRaisesRegex(Exception, "Mutable default parameters"):
torch.jit.script(Test())
@skipIfTorchDynamo("TorchDynamo fails with unknown reason")
def test_warnings(self):
import warnings
def fn(x):
if bool(x < 2):
warnings.warn("x is less than 2")
return x
class M(torch.nn.Module):
def forward(self, x):
if bool(x < 2):
warnings.warn("x is less than 2")
return x
scripted_mod = torch.jit.script(M())
scripted_fn = torch.jit.script(fn)
with warnings.catch_warnings(record=True) as warns:
fn(torch.ones(1))
with warnings.catch_warnings(record=True) as script_warns:
scripted_fn(torch.ones(1))
with warnings.catch_warnings(record=True) as script_mod_warns:
scripted_mod(torch.ones(1))
self.assertEqual(str(warns[0]), str(script_warns[0]))
self.assertEqual(len(script_mod_warns), 1)
self.assertEqual(str(warns[0].message), str(script_mod_warns[0].message))
def test_no_erroneous_warnings(self):
import warnings
def fn(x):
if bool(x > 0):
warnings.warn('This should NOT be printed')
x += 1
return x
with warnings.catch_warnings(record=True) as warns:
fn_script = torch.jit.script(fn)
fn_script(torch.tensor(0))
warns = [str(w.message) for w in warns]
self.assertEqual(len(warns), 0)
@unittest.skipIf(True, "TODO: re-enable with https://github.com/pytorch/pytorch/pull/29339")
def test_torch_load_error(self):
class J(torch.jit.ScriptModule):
def __init__(self):
super(J, self).__init__()
@torch.jit.script_method
def forward(self, input):
return input + 100
j = J()
with TemporaryFileName() as fname:
j.save(fname)
with self.assertRaisesRegex(RuntimeError, "is a zip"):
torch.load(fname)
def test_torch_load_zipfile_check(self):
@torch.jit.script
def fn(x):
return x + 10
with TemporaryFileName() as fname:
fn.save(fname)
with io.open(fname, 'rb') as f:
self.assertTrue(torch.serialization._is_zipfile(f))
def test_python_bindings(self):
lstm_cell = torch.jit.script(LSTMCellS)
def lstm(x, hx, cx, w_ih, w_hh, b_ih, b_hh):
for i in range(x.size(0)):
hx, cx = lstm_cell(x[i], hx, cx, w_ih, w_hh, b_ih, b_hh)
return hx
slstm = torch.jit.script(lstm)
inputs = get_lstm_inputs('cpu', training=True, seq_length=10)
slstm(*inputs).sum().backward()
global fw_graph
fw_graph = slstm.graph_for(*inputs)
nodes = list(fw_graph.nodes())
tested_blocks = False
for node in nodes:
for output in node.outputs():
self.assertTrue(hasattr(output, 'type'))
self.assertTrue(output.type() is not None)
for input in node.inputs():
self.assertTrue(hasattr(input, 'type'))
self.assertTrue(input.type() is not None)
for block in node.blocks():
tested_blocks = True
self.assertTrue(hasattr(block, 'inputs'))
self.assertTrue(hasattr(block, 'outputs'))
for output in block.outputs():
self.assertTrue(hasattr(output, 'type'))
self.assertTrue(output.type() is not None)
for input in block.inputs():
self.assertTrue(hasattr(input, 'type'))
self.assertTrue(input.type() is not None)
self.assertTrue(hasattr(block, 'returnNode'))
self.assertTrue(type(block.returnNode()) == torch._C.Node)
self.assertTrue(hasattr(block, 'paramNode'))
self.assertTrue(type(block.paramNode()) == torch._C.Node)
self.assertTrue(tested_blocks)
def test_export_opnames(self):
class Foo(torch.jit.ScriptModule):
def __init__(self):
super(Foo, self).__init__()
def one(self, x, y):
# type: (Tensor, Tensor) -> Tensor
return x + y
def two(self, x):
# type: (Tensor) -> Tensor
return 2 * x
@torch.jit.script_method
def forward(self, x):
# type: (Tensor) -> Tensor
return self.one(self.two(x), x)
class Bar(torch.jit.ScriptModule):
def __init__(self):
super(Bar, self).__init__()
self.sub = Foo()
@torch.jit.script_method
def forward(self, x):
# type: (Tensor) -> Tensor
return self.sub.forward(x)
bar = Bar()
ops = torch.jit.export_opnames(bar)
expected = ['aten::add.Tensor', 'aten::mul.Scalar']
self.assertTrue(set(expected).issubset(set(ops)))
def test_pytorch_jit_env_off(self):
import subprocess
env = os.environ.copy()
env['PYTORCH_JIT'] = '0'
try:
subprocess.check_output([sys.executable, '-c', 'import torch'], env=env)
except subprocess.CalledProcessError as e:
raise RuntimeError("Could not 'import torch' with PYTORCH_JIT=0") from e
def test_print_op_module(self):
# Issue #19351: python2 and python3 go through different paths.
# python2 returns '<module 'torch.ops' (built-in)>'
# python3 uses __file__ and return
# '<module 'torch.ops' from '/scratch/ailzhang/pytorch/torch/_ops.py'>'
s = str(torch.ops)
self.assertRegex(s, r'ops')
def test_print_classes_module(self):
s = str(torch.classes)
self.assertRegex(s, r'classes')
def test_print_torch_ops_modules(self):
s = str(torch._ops.ops.quantized)
self.assertRegex(s, r'torch.ops')
s = str(torch._ops.ops.atan)
self.assertRegex(s, r'torch.ops')
@unittest.skipIf(IS_WINDOWS, 'TODO: fix occasional windows failure')
def test_profiler(self):
prev_opt = torch._C._get_graph_executor_optimize()
torch._C._set_graph_executor_optimize(False)
def other_fn(x):
return x * 2
x = torch.rand(3, 4)
traced_other_fn = torch.jit.trace(other_fn, x)
def fn(x):
y = traced_other_fn(x)
fut = torch.jit._fork(traced_other_fn, x)
y = torch.jit._wait(fut)
return y
traced_fn = torch.jit.trace(fn, x)
with torch.autograd.profiler.profile() as prof:
traced_fn(x)
# expecting to see other_fn TS function call
# with cpu time >= mul cpu time and
# a forked other_fn
mul_events = defaultdict(int)
other_fn_events = defaultdict(int)
for e in prof.function_events:
if e.name == "aten::mul":
self.assertTrue(e.thread not in mul_events)
mul_events[e.thread] = e.time_range.elapsed_us()
elif e.name == "other_fn":
self.assertTrue(e.thread not in other_fn_events)
other_fn_events[e.thread] = e.time_range.elapsed_us()
self.assertTrue(len(mul_events) == 2)
self.assertTrue(len(other_fn_events) == 2)
for thread, mul_time in mul_events.items():
self.assertTrue(thread in other_fn_events)
self.assertTrue(other_fn_events[thread] >= mul_time)
torch._C._set_graph_executor_optimize(prev_opt)
def test_hide_source_ranges_context_manager(self):
@torch.jit.script
def foo(x):
return torch.add(x, x)
graph = foo.graph
source_range_regex = "# .*\\.py"
self.assertRegex(graph.__repr__(), source_range_regex)
with torch.jit._hide_source_ranges():
self.assertNotRegex(graph.__repr__(), source_range_regex)
self.assertRegex(graph.str(print_source_ranges=True), source_range_regex)
self.assertRegex(graph.__repr__(), source_range_regex)
class TestFrontend(JitTestCase):
def test_instancing_error(self):
@torch.jit.ignore
class MyScriptClass(object):
def unscriptable(self):
return "a" + 200
class TestModule(torch.nn.Module):
def __init__(self):
super(TestModule, self).__init__()
def forward(self, x):
return MyScriptClass()
with self.assertRaises(torch.jit.frontend.FrontendError) as cm:
torch.jit.script(TestModule())
checker = FileCheck()
checker.check("Cannot instantiate class")
checker.check("def forward")
checker.run(str(cm.exception))
class TestScript(JitTestCase):
# Tests that calling torch.jit.script repeated on function is allowed.
def test_repeated_script_on_function(self):
@torch.jit.script
@torch.jit.script
def fn(x):
return x
torch.jit.script(torch.jit.script(fn))
def test_pretty_print_function(self):
@torch.jit.script
def foo(x):
return torch.nn.functional.interpolate(x)
FileCheck().check("interpolate").run(foo.code)
def test_inlined_graph(self):
"""
Check that the `inlined_graph` property correctly returns an inlined
graph, both through function calls and method calls.
"""
@torch.jit.script
def foo(x):
return torch.add(x, x)
class MyNestedMod(torch.nn.Module):
def __init__(self):
super(MyNestedMod, self).__init__()
def forward(self, x):
return torch.sub(x, x)
class MyMod(torch.nn.Module):
def __init__(self):
super(MyMod, self).__init__()
self.nested = MyNestedMod()
def forward(self, x):
x = self.nested(x) # sub
x = foo(x) # add
return torch.mul(x, x)
m = torch.jit.script(MyMod())
FileCheck().check("aten::sub") \
.check("aten::add") \
.check("aten::mul") \
.run(m.inlined_graph)
def test_static_method_on_module(self):
"""
Check that the `@staticmethod` annotation on a function on a module works.
"""
class MyCell(torch.nn.Module):
def __init__(self):
super(MyCell, self).__init__()
@staticmethod
def do_it(x, h):
new_h = torch.tanh(x + h)
return new_h, new_h
def forward(self, x, h):
return self.do_it(x, h)
my_cell = torch.jit.script(MyCell())
x = torch.rand(3, 4)
h = torch.rand(3, 4)
jitted_cell = my_cell(x, h)
non_jitted_cell = MyCell().do_it(x, h)
self.assertEqual(jitted_cell, non_jitted_cell)
def test_code_with_constants(self):
"""
Check that the `code_with_constants` property correctly returns graph CONSTANTS in the
CONSTANTS.cN format used in the output of the `code` property.
"""
@torch.jit.script
def foo(x=torch.ones(1)):
return x
class Moddy(torch.nn.Module):
def __init__(self):
super(Moddy, self).__init__()
def forward(self, x):
return foo()
m = torch.jit.script(Moddy())
src, CONSTANTS = m.code_with_constants
self.assertEqual(CONSTANTS.c0, torch.ones(1))
self.assertEqual(src, m.code)
def test_code_with_constants_restore(self):
"""
Check that the `code_with_constants` property correctly works on restoration after save() + load()
"""
@torch.jit.script
def foo(x=torch.ones(1)):
return x
class Moddy(torch.nn.Module):
def __init__(self):
super(Moddy, self).__init__()
def forward(self, x):
return foo()
m = torch.jit.script(Moddy())
src, CONSTANTS = m.code_with_constants
eic = self.getExportImportCopy(m)
src_eic, CONSTANTS_eic = eic.code_with_constants
self.assertEqual(src, src_eic)
self.assertEqual(CONSTANTS.c0, CONSTANTS_eic.c0)
def test_oneline_func(self):
def fn(x): return x # noqa: E704
self.checkScript(fn, (torch.ones(2, 2), ))
def test_request_bailout(self):
with enable_profiling_mode_for_profiling_tests():
def fct_loop(x):
for i in range(3):
x = torch.cat((x, x), 0)
return x
x = torch.ones(2, 3, 4, dtype=torch.float32)
expected = fct_loop(x)
jitted = torch.jit.script(fct_loop)
# profile
jitted(x)
# optimize
jitted(x)
dstate = jitted.get_debug_state()
eplan = get_execution_plan(dstate)
num_bailouts = eplan.code.num_bailouts()
for i in range(0, num_bailouts):
eplan.code.request_bailout(i)
self.assertEqual(jitted(x), expected)
@unittest.skip("bailouts are being deprecated")
def test_dominated_bailout(self):
with enable_profiling_mode_for_profiling_tests():
# functional dominated guard
@torch.jit.script
def foo(x):
dim = x.dim()
if dim == 0:
y = int(x)
else:
y = x.size()[dim - 1]
return y
x = torch.zeros(2)
self.assertEqual(foo(x), 2)
self.assertEqual(foo(x), 2)
g = torch.jit.last_executed_optimized_graph()
g_s = str(g)
g_s = g_s[0:g_s.find("return")]
FileCheck().check_count("prim::BailOut[", 1, exactly=True).run(g_s)
# dominated guard of non-functional value
@torch.jit.script
def foo(x):
dim = x.dim()
x.add_(3)
if dim == 0:
return 0
else:
return x.size()[dim - 1]
x = torch.zeros(2)
self.assertEqual(foo(x), 2)
self.assertEqual(foo(x), 2)
g = torch.jit.last_executed_optimized_graph()
FileCheck().check("prim::BailOut[").check("aten::add_").check_next("prim::BailOut[").check("return").run(g)
with torch.enable_grad():
@torch.jit.ignore
def disable_grad():
torch.set_grad_enabled(False)
@torch.jit.ignore
def enable_grad():
torch.set_grad_enabled(True)
@torch.jit.script
def foo(x):
x = x + 1
dim = x.dim()
disable_grad()
if dim == 0:
y = int(x)
else:
y = x.size()[dim - 1]
enable_grad()
return y
x = torch.zeros(2, requires_grad=True)
self.assertEqual(foo(x), 2)
self.assertEqual(foo(x), 2)
g = torch.jit.last_executed_optimized_graph()
# there should still be a Bailout after disable_grad call
FileCheck().check("disable_grad").check("BailOut[").check("BailoutTemplate").run(g)
@unittest.skipIf(GRAPH_EXECUTOR != ProfilingMode.PROFILING, "skip if profiling isn't enabled")
def test_profiling_merge(self):
@torch.jit.script
def test_not_const(x):
if x.size(0) == 1:
return 1
else:
return 2
with enable_profiling_mode_for_profiling_tests():
with num_profiled_runs(2):
test_not_const(torch.rand([1, 2]))
test_not_const(torch.rand([2, 2]))
graph_str = torch.jit.last_executed_optimized_graph()
FileCheck().check("profiled_type=Double(*, 2, strides=[2, 1], requires_grad=0, device=cpu").run(graph_str)
FileCheck().check_not("profiled_type=Double(1, 2, strides=[2, 1], requires_grad=0, device=cpu").run(graph_str)
def test_nested_bailouts(self):
@torch.jit.script
def fct_loop(x):
for i in range(3):
x = torch.cat((x, x), 0)
return x
x = torch.ones(2, 3, 4, dtype=torch.float32)
out = fct_loop(x)
jit_trace = torch.jit.trace(fct_loop, x)
out_trace = jit_trace(x)
def test_no_self_arg_ignore_function(self):
class MyModule(nn.Module):
@torch.jit.ignore # noqa: B902
def call_np(): # noqa: B902
# type: () -> int
return np.random.choice(2, p=[.95, .05])
def forward(self):
return self.call_np()
with self.assertRaisesRegex(Exception, "does not have a self argument"):
torch.jit.script(MyModule())
def test_loop_liveness(self):
with enable_profiling_mode_for_profiling_tests():
@torch.jit.script
def f(i):
# type: (int) -> Tensor
l = []
for n in [2, 1]:
l.append(torch.zeros(n, i))
return l[0]
f(2)
f(1)
def test_bailout_loop_carried_deps_name_clash(self):
with enable_profiling_mode_for_profiling_tests():
NUM_ITERATIONS = 10
@torch.jit.script
def fct_loop(z, size):
# type: (int, int) -> Tuple[Tensor, List[int]]
counters = torch.jit.annotate(List[int], [])
j = 0
y = torch.ones(2)
for i in range(size):
counters.append(i + j)
y = torch.cat((y, torch.ones(z)), 0)
j = j + 1
return y, counters
inputs = [1, 2, 3, 4]
expected = [x * 2 for x in range(NUM_ITERATIONS)]
for inp in inputs:
results = fct_loop(inp, NUM_ITERATIONS)
self.assertEqual(results[1], expected)
def test_bailout_loop_counter_transition(self):
with enable_profiling_mode_for_profiling_tests():
NUM_ITERATIONS = 10
@torch.jit.script
def fct_loop(z, size):
# type: (int, int) -> Tuple[Tensor, List[int]]
counters = torch.jit.annotate(List[int], [])
y = torch.ones(2)
for i in range(size):
counters.append(i)
y = torch.cat((y, torch.ones(z)), 0)
return y, counters
inputs = [1, 2, 3, 4]
expected = list(range(NUM_ITERATIONS))
for inp in inputs:
results = fct_loop(inp, NUM_ITERATIONS)
self.assertEqual(results[1], expected)
def test_ignored_method_binding(self):
class Bar(torch.nn.Module):
def __init__(self):
super(Bar, self).__init__()
self.x : int = 0
@torch.jit.export
def setx(self, x : int):
self.x = x
@torch.jit.export
def getx(self):
return self.x
@torch.jit.ignore
def ignored_getx(self):
return self.x
b = Bar()
b.setx(123)
sb = torch.jit.script(b)
self.assertEqual(sb.getx(), 123)
self.assertEqual(sb.ignored_getx(), 123)
sb.setx(456)
self.assertEqual(sb.getx(), 456)
self.assertEqual(sb.ignored_getx(), 456)
def test_set_attribute_through_optional(self):
class A(torch.nn.Module):
__annotations__ = {"x": Optional[torch.Tensor]}
def __init__(self):
super(A, self).__init__()
self.x = None
@torch.jit.ignore
def foo(self):
if self.x is None:
self.x = torch.tensor([3])
return self.x
def forward(self, x):
a = self.foo()
return x + 1
m = torch.jit.script(A())
self.assertEqual(m.x, None)
m(torch.rand(1))
self.assertEqual(m.x, torch.tensor([3]))
def test_mutate_constant(self):
class M(torch.jit.ScriptModule):
__constants__ = ["foo"]
def __init__(self, foo):
super(M, self).__init__()
self.foo = foo
m = M(5)
# m has a constant attribute, but we can't
# assign to it
with self.assertRaises(RuntimeError):
m.foo = 6
def test_class_attribute(self):
class M(torch.jit.ScriptModule):
FOO = 0
def __init__(self):
super(M, self).__init__()
self.foo = self.FOO
m = M()
self.assertEqual(m.foo, M.FOO)
def test_class_attribute_in_script(self):
class M(torch.jit.ScriptModule):
FOO = 0
def __init__(self):
super(M, self).__init__()
@torch.jit.script_method
def forward(self):
return self.FOO
with self.assertRaises(RuntimeError):
M()
def test_not_initialized_err(self):
class M(torch.jit.ScriptModule):
def __init__(self):
self.foo = torch.rand(2, 3)
with self.assertRaises(RuntimeError):
M()
def test_attribute_in_init(self):
class M(torch.jit.ScriptModule):
def __init__(self):
super(M, self).__init__()
self.foo = torch.jit.Attribute(0.1, float)
# we should be able to use self.foo as a float here
assert 0.0 < self.foo
M()
def test_scriptable_fn_as_attr(self):
class M(torch.nn.Module):
def __init__(self, fn):
super(M, self).__init__()
self.fn = fn
def forward(self, x):
return self.fn(x)
m = M(torch.sigmoid)
inp = torch.rand(2, 3)
self.checkModule(m, (inp, ))
def test_sequence_parsing(self):
tests = [
("return [x, x,]", True),
("return [x x]", "expected ]"),
("return x, x,", True),
("return bar(x, x,)", True),
("return bar()", "Argument x not provided"),
("for a, b, in x, x,:\n pass", "List of iterables"),
("a, b, = x, x,\n return a + b", True)
]
for exp, result in tests:
cu = torch.jit.CompilationUnit()
full = """
def bar(x, y):
return x + y
def foo(x):
{}
""".format(exp)
if isinstance(result, str):
with self.assertRaisesRegex(RuntimeError, result):
cu.define(full)
else:
cu.define(full)
def test_namedtuple_python(self):
global MyTuple, MyMod # see [local resolution in python]
MyTuple = namedtuple('MyTuple', ['a'])
@torch.jit.unused
def fn():
# type: () -> MyTuple
return MyTuple(1)
# Only check compilation
@torch.jit.script
def fn2():
# type: () -> MyTuple
return fn()
FileCheck().check("NamedTuple").run(fn2.graph)
class MyMod(torch.nn.Module):
def __init__(self):
super(MyMod, self).__init__()
@torch.jit.unused
def fn(self):
# type: () -> MyTuple
return MyTuple(1)
def forward(self, x):
if 1 == 1:
return MyTuple(torch.rand(2, 3))
else:
return self.fn()
# shouldn't throw a type error
torch.jit.script(MyMod())
def test_unused_decorator(self):
class MyMod(torch.nn.Module):
def __init__(self):
super(MyMod, self).__init__()
@torch.jit.unused
@torch.no_grad()
def fn(self, x):
# type: (Tensor) -> int
return next(x) # invalid, but should be ignored
def forward(self, x):
return self.fn(x)
torch.jit.script(MyMod())
@_inline_everything
def test_lazy_script(self):
def untraceable(x):
if x.ndim > 2:
print("hello")
else:
print("goodbye")
return x + 2
# Non-working example
def fn(x):
return untraceable(x)
with self.capture_stdout():
traced_bad = torch.jit.trace(fn, [torch.ones(2, 2)])
FileCheck().check_not("goodbye").check_not("hello").run(traced_bad.graph)
# Working example
untraceable = torch.jit.script_if_tracing(untraceable)
def fn2(x):
return untraceable(x)
with self.capture_stdout():
traced = torch.jit.trace(fn, [torch.ones(2, 2)])
FileCheck().check("goodbye").run(traced.graph)
def foo(x: int):
return x + 1
@torch.jit.script_if_tracing
def fee(x: int = 2):
return foo(1) + x
# test directly compiling function
fee_compiled = torch.jit.script(fee)
self.assertEqual(fee_compiled(), fee())
# test compiling it within another function
@torch.jit.script
def hum():
return fee(x=3)
self.assertEqual(hum(), 5)
def test_big_int_literals(self):
def ok():
# signed 64 bit max
a = 9223372036854775807
return a
def toobig():
a = 9223372036854775808
return a
def waytoobig():
a = 99999999999999999999
return a
self.checkScript(ok, [])
with self.assertRaisesRegex(RuntimeError, "out of range"):
torch.jit.script(toobig)
with self.assertRaisesRegex(RuntimeError, "out of range"):
torch.jit.script(waytoobig)
def test_hex_literals(self):
def test1():
return 0xaaaaaa
def test2():
return 0xaaaaaa
def test3():
return -0xaaaaaa
self.checkScript(test1, [])
self.checkScript(test2, [])
self.checkScript(test3, [])
def ok():
a = 0x7FFFFFFFFFFFFFFF
return a
def toobig():
a = 0xFFFFFFFFFFFFFFFF
return a
def waytoobig():
a = 0xFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFF
return a
self.checkScript(ok, [])
with self.assertRaisesRegex(RuntimeError, "out of range"):
torch.jit.script(toobig)
with self.assertRaisesRegex(RuntimeError, "out of range"):
torch.jit.script(waytoobig)
def test_big_float_literals(self):
def ok():
# Python interprets this as inf
a = 1.2E400
return a
def check(fn):
self.assertTrue(fn() == ok())
# checkScript doesn't work since assertEqual doesn't consider
# `inf` == `inf`
check(torch.jit.script(ok))
cu = torch.jit.CompilationUnit()
cu.define(dedent(inspect.getsource(ok)))
check(cu.ok)
def _test_device_type(self, dest):
def fn(x):
# type: (Device) -> Tuple[str, Optional[int]]
return x.type, x.index
device = torch.ones(2).to(dest).device
self.checkScript(fn, [device])
def test_device_type(self):
self._test_device_type('cpu')
@unittest.skipIf(not RUN_CUDA, "Requires CUDA")
def test_device_type_cuda(self):
self._test_device_type('cuda')
def test_string_device_implicit_conversion(self):
@torch.jit.script
def fn(x: torch.device):
return x
self.assertEqual(fn("cpu"), torch.device("cpu"))
with self.assertRaisesRegex(RuntimeError, "Expected one of"):
fn("invalid_device")
def test_eval_python(self):
def _test(m):
self.assertTrue(m(torch.ones(2, 2)))
self.assertTrue(m.training)
self.assertTrue(m._c.getattr('training'))
m.eval()
self.assertFalse(m.training)
self.assertFalse(m._c.getattr('training'))
self.assertFalse(m(torch.ones(2, 2)))
buffer = io.BytesIO()
torch.jit.save(m, buffer)
buffer.seek(0)
loaded = torch.jit.load(buffer)
self.assertFalse(loaded.training)
self.assertFalse(loaded._c.getattr('training'))
class M(nn.Module):
def __init__(self):
super(M, self).__init__()
def forward(self, x):
return self.training
class OldM(torch.jit.ScriptModule):
def __init__(self):
super(OldM, self).__init__()
@torch.jit.script_method
def forward(self, x):
return self.training
_test(torch.jit.script(M()))
_test(OldM())
def test_inherit_method(self):
class A(torch.jit.ScriptModule):
def __init__(self):
super(A, self).__init__()
@torch.jit.script_method
def forward(self, x):
return x + self.bar(x)
class B(A):
def __init__(self):
super(B, self).__init__()
@torch.jit.script_method
def bar(self, x):
return x * x
with self.assertRaisesRegex(RuntimeError, 'attribute'):
A() # cannot use because bar is not defined
v = torch.rand(3, 4)
b = B()
self.assertEqual(b(v), v + v * v)
class C(torch.jit.ScriptModule):
def __init__(self):
super(C, self).__init__()
@torch.jit.script_method
def bar(self, x):
return x
class D(C, B):
def __init__(self):
super(D, self).__init__()
self.assertEqual(D()(v), v + v)
def test_tensor_subclasses(self):
def check_subclass(x, tensor):
template = dedent("""
def func(input: {}) -> {}:
return torch.zeros((input.shape[0], 1), dtype=input.dtype)
""")
self._check_code(template.format(x, x), "func", [tensor])
check_subclass("torch.LongTensor", torch.LongTensor([[1, 2], [3, 4]]))
check_subclass("torch.DoubleTensor", torch.DoubleTensor([[1.2, 2.3], [3.4, 4.5]]))
check_subclass("torch.IntTensor", torch.IntTensor([[1, 2], [3, 4]]))
check_subclass("torch.BoolTensor", torch.BoolTensor([[False, True], [True, False]]))
def check_subclass_warn(input: torch.LongTensor) -> torch.LongTensor:
return torch.zeros((input.shape[0], 1), dtype=input.dtype)
with warnings.catch_warnings(record=True) as warns:
scripted = torch.jit.script(check_subclass_warn)
FileCheck().check("TorchScript will treat type annotations of Tensor").run(str(warns[0]))
def test_first_class_module(self):
class Foo(torch.jit.ScriptModule):
def __init__(self):
super(Foo, self).__init__()
self.foo = nn.Parameter(torch.rand(3, 4))
@torch.jit.script_method
def forward(self, input):
self.foo = input
return self.foo
foo = Foo()
input = torch.rand(3, 4)
foo.forward(input)
self.assertEqual(input, foo.foo)
@_tmp_donotuse_dont_inline_everything
def test_first_class_calls(self):
@torch.jit.script
class Foo(object):
def __init__(self, x):
self.bar = x
def stuff(self, x):
return self.bar + x
@torch.jit.script
def foo(x):
return x * x + Foo(x).stuff(2 * x)
@torch.jit.script
def bar(x):
return foo(x) * foo(x)
x = torch.rand(3, 4)
self.assertEqual(bar(x), (x * x + 3 * x) * (x * x + 3 * x))
def test_static_methods(self):
class M(nn.Module):
def __init__(self):
super(M, self).__init__()
@staticmethod
def my_method(x):
return x + 100
def forward(self, x):
return x + M.my_method(x)
class N(nn.Module):
def __init__(self):
super(N, self).__init__()
@staticmethod
def my_method(x):
return x * 100
def forward(self, x):
return x - M.my_method(x) + N.my_method(x)
self.checkModule(M(), (torch.ones(2, 2),))
self.checkModule(N(), (torch.ones(2, 2),))
def test_invalid_prefix_annotation(self):
with self.assertRaisesRegex(RuntimeError, "annotation prefix in line"):
with self.capture_stdout() as captured:
@torch.jit.script
def invalid_prefix_annotation1(a):
#type: (Int) -> Int # noqa: E265
return a + 2
with self.assertRaisesRegex(RuntimeError, "annotation prefix in line"):
with self.capture_stdout() as captured:
@torch.jit.script
def invalid_prefix_annotation2(a):
#type : (Int) -> Int # noqa: E265
return a + 2
with self.assertRaisesRegex(RuntimeError, "annotation prefix in line"):
with self.capture_stdout() as captured:
@torch.jit.script
def invalid_prefix_annotation3(a):
# type: (Int) -> Int
return a + 2
def test_builtin_function_attributes(self):
class Add(nn.Module):
def __init__(self):
super(Add, self).__init__()
self.add = torch.add
def forward(self, input):
return self.add(input, input)
self.checkModule(Add(), [torch.randn(2, 2)])
def test_pybind_type_comparisons(self):
@torch.jit.script
def f():
return None
node = list(f.graph.nodes())[0]
t = node.outputsAt(0).type()
self.assertIsNotNone(t)
@unittest.skipIf(IS_WINDOWS and sys.version_info >= (3, 8), 'TODO: need to fix the test case')
def test_unmatched_type_annotation(self):
message1 = re.escape("Number of type annotations (2) did not match the number of function parameters (1):")
message2 = 'def invalid2\\(a\\):\n\\s*~+\\.*\\s+<--- HERE\n\\s+# type: \\(Int, Int\\) -> Int\n\\s+return a \\+ 2'
message3 = 'def invalid4\\(a\\):\n\\s*~+\\.*\\s+<--- HERE\n\\s+# type: \\(Int, Int\\) -> Int\n\\s+return a \\+ 2'
with self.assertRaisesRegex(RuntimeError, message1):
@torch.jit.script
def invalid1(a):
# type: (Int, Int) -> Int
return a + 2
with self.assertRaisesRegex(RuntimeError, message2):
@torch.jit.script
def invalid2(a):
# type: (Int, Int) -> Int
return a + 2
with self.assertRaisesRegex(RuntimeError, message1):
def invalid3(a):
# type: (Int, Int) -> Int
return a + 2
torch.jit.script(invalid3)
with self.assertRaisesRegex(RuntimeError, message3):
def invalid4(a):
# type: (Int, Int) -> Int
return a + 2
torch.jit.script(invalid4)
def test_is_optional(self):
ann = Union[List[int], List[float]]
torch._jit_internal.is_optional(ann)
def test_interpreter_fuzz(self):
import builtins
# This test generates random tree-like programs to fuzz test
# that the interpreter does not have a bug in its stack manipulation
# code. An assert in that code ensures individual operators are
# not reordered.
templates = [
"torch.rand(3, 4)",
"({} + {})",
"-{}",
"({} * {})",
"torch.tanh({})",
"VAR {}",
]
def gen_code():
src_lines = ['def f():']
exprs = []
n_variables = 0
def get_expr(idx):
elem = exprs[idx]
exprs[idx] = exprs[-1]
exprs.pop()
return elem
def select_expr_or_var():
idx = random.randrange(0, len(exprs) + n_variables)
if idx < len(exprs):
return get_expr(idx)
else:
return 'v{}'.format(idx - len(exprs))
for i in range(50):
n = None
while n is None or n > len(exprs) + n_variables:
template = random.choice(templates)
n = template.count('{}')
if 'VAR' in template:
src_lines.append(' v{} = {}'.format(n_variables, select_expr_or_var()))
n_variables += 1
else:
exprs.append(template.format(*(select_expr_or_var() for _ in range(n))))
src_lines.append(' return ({})\n'.format(''.join('v{},'.format(i) for i in range(n_variables))))
return '\n'.join(src_lines)
for i in range(100):
g = {'torch': torch}
code = gen_code()
builtins.exec(code, g, None)
cu = torch.jit.CompilationUnit(code)
with freeze_rng_state():
o1 = g['f']()
with freeze_rng_state():
o2 = cu.f()
self.assertEqual(o1, o2)
@skipIfTorchDynamo("TorchDynamo fails with unknown reason")
def test_cpp_module_iterator(self):
a = nn.Module()
a.name = 'a'
a.p = nn.Parameter(torch.rand(3, 4))
a.foo = nn.Module()
a.foo.name = 'foo'
a.foo.register_buffer('b', torch.rand(1, 1))
a.foo.bar = nn.Module()
a.foo.bar.name = 'bar'
a.foo.bar.an_int = 4
a.another = nn.Module()
a.another.name = 'another'
sa = torch.jit.script(a)
result = torch._C._jit_debug_module_iterators(sa._c)
def replace(e):
if e is a.p:
return 'P'
elif e is a.foo.b:
return 'B'
elif isinstance(e, torch._C.ScriptModule):
return e.getattr('name')
return e
for k, v in result.items():
for i in range(len(v)):
if isinstance(v[i], tuple):
n, v2 = v[i]
v[i] = (n, replace(v2))
else:
v[i] = replace(v[i])
# module type creation is not deterministic, so we have to sort
# the result
v.sort()
expected = {'buffers': [],
'buffers_r': ['B'],
'children': ['another', 'foo'],
'modules': ['a', 'another', 'bar', 'foo'],
'named_attributes': [('_is_full_backward_hook', None),
('another', 'another'),
('foo', 'foo'),
('name', 'a'),
('p', 'P'),
('training', True)],
'named_attributes_r': [('_is_full_backward_hook', None),
('another', 'another'),
('another._is_full_backward_hook', None),
('another.name', 'another'),
('another.training', True),
('foo', 'foo'),
('foo._is_full_backward_hook', None),
('foo.b', 'B'),
('foo.bar', 'bar'),
('foo.bar._is_full_backward_hook', None),
('foo.bar.an_int', 4),
('foo.bar.name', 'bar'),
('foo.bar.training', True),
('foo.name', 'foo'),
('foo.training', True),
('name', 'a'),
('p', 'P'),
('training', True)],
'named_buffers': [],
'named_buffers_r': [('foo.b', 'B')],
'named_children': [('another', 'another'), ('foo', 'foo')],
'named_modules': [('', 'a'),
('another', 'another'),
('foo', 'foo'),
('foo.bar', 'bar')],
'named_parameters': [('p', 'P')],
'named_parameters_r': [('p', 'P')],
'parameters': ['P'],
'parameters_r': ['P']}
self.assertEqual(expected, result)
def test_parameter_order(self):
m = nn.Module()
for i, name in enumerate(string.ascii_letters):
setattr(m, name, nn.Parameter(torch.tensor([float(i)])))
ms = torch.jit.script(m)
print(torch.cat(list(m.parameters())))
print(torch.cat(list(ms.parameters())))
self.assertEqual(list(m.parameters()), list(ms.parameters()))
def test_python_op_builtins(self):
@torch.jit.unused
def fn(x):
# type: (List[int]) -> int
return sum(x)
@torch.jit.script
def script_fn(x):
# type: (List[int]) -> int
return fn(x)
def test_submodule_twice(self):
@torch.jit.script
def foo(x):
return x * x
class What(torch.jit.ScriptModule):
def __init__(self, x):
super(What, self).__init__()
self.foo = x
a = What(foo)
c = What(foo)
def test_training_param(self):
class What(torch.jit.ScriptModule):
def __init__(self):
super(What, self).__init__()
@torch.jit.script_method
def forward(self, x):
# type: (int) -> int
if self.training:
r = x
else:
r = x + 4
# check double use of training
if self.training:
r = r + 1
return r
w = What()
self.assertEqual(4, w(3))
w.train(False)
self.assertEqual(7, w(3))
self.assertFalse("training" in w.state_dict())
def test_class_as_attribute(self):
@torch.jit.script
class Foo321(object):
def __init__(self):
self.x = 3
class FooBar1234(torch.nn.Module):
def __init__(self):
super(FooBar1234, self).__init__()
self.f = Foo321()
def forward(self, x):
return x + self.f.x
scripted = torch.jit.script(FooBar1234())
eic = self.getExportImportCopy(scripted)
x = torch.rand(3, 4)
self.assertEqual(scripted(x), eic(x))
def test_module_str(self):
class Foo(torch.nn.Module):
def forward(self, x):
return torch.relu(x)
f = torch.jit.script(Foo())
self.assertEqual('ScriptObject', str(f._c))
def test_jitter_bug(self):
@torch.jit.script
def fn2(input, kernel_size):
# type: (Tensor, List[int]) -> Tensor
if kernel_size[0] > 1:
_stride = [2]
else:
_stride = kernel_size
print(_stride, kernel_size)
return input
@torch.jit.script
def fn(input):
# type: (Tensor) -> Tensor
return fn2(input, [1])
def test_parser_kwargonly(self):
cu = torch.jit.CompilationUnit('''
def foo(x, *, y) -> Tuple[Tensor, Tensor]:
return x, x
def bar(x):
return foo(x, y=x)
''')
self.assertTrue('*' in str(cu.foo.schema))
with self.assertRaisesRegex(RuntimeError, "not provided"):
torch.jit.CompilationUnit('''
def foo(x, *, y) -> Tuple[Tensor, Tensor]:
return x, x
def bar(x):
return foo(x, x)
''')
def test_annoying_doubles(self):
mod = types.ModuleType("temp")
mod.inf = float("inf")
mod.ninf = float("-inf")
mod.nan = float("nan")
with torch._jit_internal._disable_emit_hooks():
class Foo(torch.jit.ScriptModule):
def __init__(self):
super(Foo, self).__init__()
@torch.jit.script_method
def forward(self):
return math.pi, 0.1, mod.inf, mod.ninf, 2.225073858507201e-308, mod.nan
foo = Foo()
buffer = io.BytesIO()
torch.jit.save(foo, buffer)
buffer.seek(0)
foo_loaded = torch.jit.load(buffer)
r = foo()
r2 = foo_loaded()
# use precise assert, we are checking floating point details
self.assertTrue(r[:-1] == r2[:-1])
self.assertTrue(math.isnan(r[-1]) and math.isnan(r2[-1]))
def test_type_annotate(self):
def foo(a):
return torch.jit.annotate(torch.Tensor, a)
self.checkScript(foo, (torch.rand(3),))
def bar():
a = torch.jit.annotate(List[int], [])
for _ in range(10):
a.append(4)
return a
self.checkScript(bar, ())
def baz(a):
return torch.jit.annotate(float, a)
self.checkScript(baz, (torch.rand(()),))
# test annotate none types
def annotate_none():
return torch.jit.annotate(Optional[torch.Tensor], None)
self.checkScript(annotate_none, ())
def test_robust_op_resolution(self):
neg = torch.add # misleading name to make sure we resolve by function
def stuff(x):
return neg(x, x)
a = (torch.rand(3),)
self.checkScript(stuff, a)
def test_nested_aug_assign(self):
@torch.jit.script
class SomeClass(object):
def __init__(self):
self.num = 99
def __iadd__(self, x):
# type: (int)
self.num += x
return self
def __eq__(self, other):
# type: (SomeClass) -> bool
return self.num == other.num
@torch.jit.script
class SomeOutOfPlaceClass(object):
def __init__(self):
self.num = 99
def __add__(self, x):
# type: (int)
self.num = x
return self
def __eq__(self, other):
# type: (SomeClass) -> bool
return self.num == other.num
class Child(nn.Module):
def __init__(self):
super().__init__()
self.x = 2
self.o = SomeClass()
self.oop = SomeOutOfPlaceClass()
self.list = [1, 2, 3]
class A(nn.Module):
def __init__(self):
super().__init__()
self.child = Child()
def forward(self):
self.child.x += 1
self.child.o += 5
self.child.oop += 5
some_list = [1, 2]
self.child.list += some_list
self.child.list *= 2
return self.child.x, self.child.o, self.child.list, self.child.oop
a = A()
sa = torch.jit.script(A())
eager_result = a()
script_result = sa()
self.assertEqual(eager_result, script_result)
self.assertEqual(a.child.x, sa.child.x)
self.assertEqual(a.child.o, sa.child.o)
self.assertEqual(a.child.list, sa.child.list)
@torch.jit.script
class SomeNonAddableClass(object):
def __init__(self):
self.num = 99
def __eq__(self, other):
# type: (SomeClass) -> bool
return self.num == other.num
# with self.assertRaisesRegex(RuntimeError, "")
class A(nn.Module):
def __init__(self):
super().__init__()
self.x = SomeNonAddableClass()
def forward(self):
self.x += SomeNonAddableClass()
return self.x
with self.assertRaisesRegex(RuntimeError, "Cannot emit inplace op"):
torch.jit.script(A())
def test_var_aug_assign(self):
@torch.jit.script
class SomeNonAddableClass(object):
def __init__(self):
self.num = 99
def __eq__(self, other):
# type: (SomeNonAddableClass) -> bool
return self.num == other.num
with self.assertRaisesRegex(RuntimeError, "Cannot emit inplace op"):
@torch.jit.script
def fn():
a = SomeNonAddableClass()
a += SomeNonAddableClass()
return a
@torch.jit.script
class SomeClass(object):
def __init__(self):
self.num = 99
def __iadd__(self, x):
# type: (int)
self.num += x
return self
def __eq__(self, other):
# type: (SomeClass) -> bool
return self.num == other.num
@torch.jit.script
class SomeOutOfPlaceClass(object):
def __init__(self):
self.num = 99
def __add__(self, x):
# type: (int)
self.num = x
return self
def __eq__(self, other):
# type: (SomeClass) -> bool
return self.num == other.num
def fn2():
a = SomeClass()
a_copy = a
a += 20
assert a is a_copy
b = SomeOutOfPlaceClass()
b_copy = b
b += 99
assert b is b_copy
c = [1, 2, 3]
c_copy = c
c *= 2
assert c is c_copy
c += [4, 5, 6]
d = torch.ones(2, 2)
d_copy = d
d += torch.ones(2, 2)
assert d is d_copy
return a, b, c, d
self.checkScript(fn2, [])
def test_nested_list_construct(self):
def foo():
return [[4]] + [[4, 5]]
self.checkScript(foo, ())
def test_file_line_error(self):
def foobar(xyz):
return torch.blargh(xyz)
_, lineno = inspect.getsourcelines(foobar)
with self.assertRaisesRegex(RuntimeError, "test_jit.py\", line {}".format(lineno + 1)):
scripted = torch.jit.script(foobar)
def test_file_line_error_class_defn(self):
class FooBar(object):
def baz(self, xyz):
return torch.blargh(xyz)
_, lineno = inspect.getsourcelines(FooBar)
with self.assertRaisesRegex(RuntimeError, "test_jit.py\", line {}".format(lineno + 2)):
torch.jit.script(FooBar)
def test_file_line_graph(self):
def foobar(xyz):
return torch.neg(xyz)
scripted = torch.jit.script(foobar)
_, lineno = inspect.getsourcelines(foobar)
fc = FileCheck().check('test_jit.py:{}:19'.format(lineno + 1))
fc.run(scripted.graph)
fc.run(str(scripted.graph))
def test_file_line_save_load(self):
class Scripted(torch.jit.ScriptModule):
@torch.jit.script_method
def forward(self, xyz):
return torch.neg(xyz)
scripted = Scripted()
# NB: not using getExportImportCopy because that takes a different
# code path that calls CompilationUnit._import rather than
# going through the full save/load pathway
buffer = scripted.save_to_buffer()
bytesio = io.BytesIO(buffer)
scripted = torch.jit.load(bytesio)
_, lineno = inspect.getsourcelines(Scripted)
fc = FileCheck().check(':{}'.format(lineno + 3))
fc.run(scripted.graph)
fc.run(str(scripted.graph))
def test_file_line_string(self):
scripted = torch.jit.CompilationUnit('''
def foo(xyz):
return torch.neg(xyz)
''')
fc = FileCheck().check('<string>:3:11')
fc.run(scripted.foo.graph)
fc.run(str(scripted.foo.graph))
@skipIfCrossRef
def test_file_line_trace(self):
def foobar(xyz):
return torch.neg(xyz)
scripted = torch.jit.trace(foobar, (torch.rand(3, 4)))
_, lineno = inspect.getsourcelines(foobar)
fc = FileCheck().check('test_jit.py:{}:0'.format(lineno + 1))
fc.run(scripted.graph)
fc.run(str(scripted.graph))
def test_serialized_source_ranges(self):
class FooTest(torch.jit.ScriptModule):
@torch.jit.script_method
def forward(self, x, w):
return torch.mm(x, w.t())
ft = FooTest()
loaded = self.getExportImportCopy(ft)
_, lineno = inspect.getsourcelines(FooTest)
with self.assertRaisesRegex(RuntimeError, 'test_jit.py\", line {}'.format(lineno + 3)):
loaded(torch.rand(3, 4), torch.rand(30, 40))
def test_serialized_source_ranges_graph(self):
class FooTest3(torch.jit.ScriptModule):
@torch.jit.script_method
def forward(self, x, w):
return torch.mm(x, w.t())
ft = FooTest3()
loaded = self.getExportImportCopy(ft)
_, lineno = inspect.getsourcelines(FooTest3)
fc = FileCheck().check('test_jit.py:{}'.format(lineno + 3))
fc.run(loaded.graph)
def test_serialized_source_ranges2(self):
class FooTest2(torch.jit.ScriptModule):
@torch.jit.script_method
def forward(self):
raise RuntimeError('foo')
_, lineno = inspect.getsourcelines(FooTest2)
with self.assertRaisesRegex(torch.jit.Error, 'test_jit.py\", line {}'.format(lineno + 3)):
ft = FooTest2()
loaded = self.getExportImportCopy(ft)
loaded()
def test_serialized_source_ranges_dont_jitter(self):
class FooTest3(torch.jit.ScriptModule):
@torch.jit.script_method
def forward(self, lim):
first = 1
second = 1
i = 1
somenum = 5
dontmutateme = 3
third = 0
while bool(i < lim):
third = first + second
first = second
second = third
j = 0
while j < 10:
somenum = somenum * 2
j = j + 1
i = i + j
i = i + dontmutateme
st = second + third
fs = first + second
return third, st, fs
ft3 = FooTest3()
def debug_records_from_mod(self, mod):
buffer = io.BytesIO()
torch.jit.save(ft3, buffer)
buffer.seek(0)
archive = zipfile.ZipFile(buffer)
files = filter(lambda x: x.startswith('archive/code/'), archive.namelist())
debug_files = list(filter(lambda f: f.endswith('.debug_pkl'), files))
self.assertEqual(len(debug_files), 1)
debug_file = archive.open(debug_files[0])
return pickle.load(debug_file), buffer
records1, buffer = debug_records_from_mod(self, ft3)
buffer.seek(0)
loaded = torch.jit.load(buffer)
records2, buffer = debug_records_from_mod(self, loaded)
buffer.seek(0)
loaded2 = torch.jit.load(buffer)
records3, _ = debug_records_from_mod(self, loaded2)
self.assertEqual(records1, records2)
self.assertEqual(records2, records3)
def test_serialized_source_ranges_no_dups(self):
class FooTest3(torch.jit.ScriptModule):
@torch.jit.script_method
def forward(self, lim):
first = 1
second = 1
i = 1
somenum = 5
dontmutateme = 3
third = 0
while bool(i < lim):
third = first + second
first = second
second = third
j = 0
while j < 10:
somenum = somenum * 2
j = j + 1
i = i + j
i = i + dontmutateme
st = second + third
fs = first + second
return third, st, fs
ft3 = FooTest3()
def debug_records_from_mod(mod):
buffer = io.BytesIO()
torch.jit.save(ft3, buffer)
buffer.seek(0)
archive = zipfile.ZipFile(buffer)
files = list(filter(lambda x: x.startswith('archive/code/'), archive.namelist()))
debug_files = filter(lambda f: f.endswith('.debug_pkl'), files)
debug_files = (archive.open(f) for f in debug_files)
debug_files = (pickle.load(f) for f in debug_files)
debug_files = (f[2] for f in debug_files)
return list(debug_files)
debug_files = debug_records_from_mod(ft3)
for debug_file in debug_files:
for i in range(len(debug_file) - 1):
offset, source_range_tag, source_range = debug_file[i]
offset2, source_range_tag2, source_range2 = debug_file[i + 1]
self.assertNotEqual(source_range, source_range2)
def test_circular_dependency(self):
"""
https://github.com/pytorch/pytorch/issues/25871
"""
class A(torch.jit.ScriptModule):
def __init__(self):
super(A, self).__init__()
@torch.jit.script_method
def forward(self, x):
return x
class B(torch.jit.ScriptModule):
def __init__(self):
super(B, self).__init__()
self.foo = torch.nn.ModuleList([A()])
@torch.jit.script_method
def forward(self, x):
for f in self.foo:
x = f(x)
return x
class C(torch.jit.ScriptModule):
def __init__(self):
super(C, self).__init__()
self.foo = torch.nn.Sequential(B())
@torch.jit.script_method
def forward(self, x):
for f in self.foo:
x = f(x)
return x
self.getExportImportCopy(C())
def test_serialize_long_lines(self):
class OrderModuleLong(torch.nn.Module):
def forward(self, long_arg_name: List[torch.Tensor]):
return [(long_arg_name[1],), (long_arg_name[0].argmax(),)]
src = str(torch.jit.script(OrderModuleLong()).code)
# make long_arg_name[1] does not get reordered after the argmax
FileCheck().check("long_arg_name[1]").check("argmax").run(src)
def test_tensor_shape(self):
x = torch.empty(34, 56, 78)
def f(x):
return x.shape
self.checkScript(f, (x,))
def test_block_input_grad_in_loop(self):
x = torch.randn(3, 3, requires_grad=False)
y = torch.randn(3, 3, requires_grad=True)
def grad_in_loop(x, y):
for i in range(100):
x = y @ x
return x
scripted = torch.jit.script(grad_in_loop)
outer = scripted.graph_for(x, y)
loop = outer.findNode("prim::Loop")
loop_block = next(loop.blocks())
param_node = loop_block.paramNode()
x_value = list(param_node.outputs())[1]
self.assertTrue(x_value.requires_grad())
def test_tensor_grad(self):
x = torch.randn(3, 4, requires_grad=True)
y = torch.randn(3, 4, requires_grad=False)
def f_requires_grad(x):
return x.requires_grad
self.checkScript(f_requires_grad, (x,))
self.checkScript(f_requires_grad, (y,))
def f_grad(x):
return x.grad
x.sum().backward()
self.checkScript(f_grad, (x,))
self.checkScript(f_grad, (y,))
@unittest.skipIf(GRAPH_EXECUTOR != ProfilingMode.LEGACY, "shape analysis is only enabled in Legacy")
def test_prim_grad_undefined(self):
x = torch.ones(2)
def f_grad(x):
return x.grad
scripted = self.checkScript(f_grad, (x,))
g = scripted.graph_for(x)
prim_grad_node = g.findNode("prim::grad")
self.assertTrue(next(prim_grad_node.outputs()).type().undefined() is None)
def test_tensor_data(self):
x = torch.randn(3, 4, requires_grad=True)
y = torch.randn(4, 5)
def f_data(x):
return x.data
scripted_f_data = torch.jit.script(f_data)
scripted_x = scripted_f_data(x)
self.assertEqual(scripted_x, f_data(x))
self.assertEqual(scripted_x.requires_grad, False)
scripted_y = scripted_f_data(y)
self.assertEqual(scripted_y, f_data(y))
self.assertEqual(scripted_x.requires_grad, False)
def test_tensor_dtype(self):
x_byte = torch.empty(34, 56, 78, dtype=torch.uint8)
x_long = torch.empty(34, 56, 78, dtype=torch.long)
x_float32 = torch.empty(34, 56, 78, dtype=torch.float32)
@torch.jit.script
def byte(x):
return x.dtype == torch.uint8
@torch.jit.script
def long(x):
return x.dtype == torch.long
@torch.jit.script
def float32(x):
return x.dtype == torch.float32
self.assertTrue(byte(x_byte))
self.assertFalse(byte(x_long))
self.assertFalse(byte(x_float32))
self.assertFalse(long(x_byte))
self.assertTrue(long(x_long))
self.assertFalse(long(x_float32))
self.assertFalse(float32(x_byte))
self.assertFalse(float32(x_long))
self.assertTrue(float32(x_float32))
@unittest.skipIf(not RUN_CUDA, "device tests require CUDA")
def test_tensor_device(self):
cpu = torch.empty(34, 56, 78, device='cpu')
gpu = torch.empty(34, 56, 78, device='cuda')
@torch.jit.script
def same_device(x, y):
return x.device == y.device
self.assertTrue(same_device(cpu, cpu))
self.assertTrue(same_device(gpu, gpu))
self.assertFalse(same_device(cpu, gpu))
@unittest.skipIf(not RUN_CUDA, "device tests require CUDA")
def test_tensor_to_device(self):
def to_device(x):
return x.to(device="cuda").to(device=torch.device("cpu"))
self.checkScript(to_device, (torch.ones(3, 4),))
def test_tensor_to_cpu(self):
def to_cpu(x):
return x.cpu()
x = torch.ones(3, 4)
script_fn = torch.jit.script(to_cpu)
self.assertEqual(to_cpu(x).device, script_fn(x).device)
self.checkScript(to_cpu, (x,))
@unittest.skipIf(not RUN_CUDA, "device tests require CUDA")
def test_tensor_to_cuda(self):
def to_cuda(x):
return x.cuda()
x = torch.ones(3, 4)
script_fn = torch.jit.script(to_cuda)
self.assertEqual(to_cuda(x).device, script_fn(x).device)
self.checkScript(to_cuda, (x,))
def test_generic_list_errors(self):
with self.assertRaisesRegex(RuntimeError, "previously matched to type"):
@torch.jit.script
def foo(x):
return [[x]] + [[1]]
def test_script_cu(self):
cu = torch.jit.CompilationUnit('''
def foo(a):
b = a
return b
''')
a = Variable(torch.rand(1))
self.assertEqual(a, cu.foo(a))
# because the compilation unit ingests python strings
# to use an escape sequence escape the backslash (\\n = \n)
def test_string_cu(self):
cu = torch.jit.CompilationUnit('''
def foo(a):
print(a, """a\\n\tb\\n""", 2, "a\
a")
return a
''')
FileCheck().check("aa").check("a\\n\\tb\\n").run(str(cu.foo.graph))
def test_function_compilation_caching(self):
def fun():
return 1 + 2
fun_compiled = torch.jit.script(fun)
# python wrapper around the script function is a different pointer,
# but the underlying script function graph is the same
self.assertIs(fun_compiled.graph, torch.jit.script(fun).graph)
def fun():
return 3 + 4
num_ref_counts = sys.getrefcount(fun)
# caching doesn't get tripped up by same qualname
fun_compiled_2 = torch.jit.script(fun)
self.assertIsNot(fun_compiled, fun_compiled_2)
self.assertEqual(fun_compiled_2(), 7)
# caching doesnt increase refcounts to function (holds weak reference)
self.assertTrue(sys.getrefcount(fun), num_ref_counts)
def test_string_ops(self):
def foo():
a = "a" + "b"
return a + a, "ab" == "b", "ab" != "b", "ab" == "ab", "ab" != "ab"
self.checkScript(foo, ())
def test_string_sorted(self):
def foo(strs: List[str]):
return sorted(strs)
FileCheck() \
.check("graph") \
.check_next("str[] = aten::sorted") \
.check_next("return") \
.run(str(torch.jit.script(foo).graph))
inputs = ["str3", "str2", "str1"]
self.checkScript(foo, (inputs,))
def test_string_sort(self):
def foo(strs: List[str]):
strs.sort()
return strs
inputs = ["str3", "str2", "str1"]
self.checkScript(foo, (inputs,))
def test_tuple_sorted(self):
def foo(tups: List[Tuple[int, int]]):
return sorted(tups)
inputs = [(1, 2), (0, 2), (1, 3)]
self.checkScript(foo, (inputs,))
def test_tuple_sort(self):
def foo(tups: List[Tuple[int, int]]):
tups.sort()
return tups
inputs = [(1, 2), (0, 2), (1, 3)]
self.checkScript(foo, (inputs,))
def test_tuple_sort_reverse(self):
def foo(tups: List[Tuple[int, int]]):
tups.sort(reverse=True)
return tups
inputs = [(1, 2), (0, 2), (1, 3)]
self.checkScript(foo, (inputs,))
def test_tuple_unsortable_element_type(self):
@torch.jit.script
def foo():
tups = [({1: 2}, {2: 3})]
tups.sort()
return tups
with self.assertRaisesRegexWithHighlight(RuntimeError, "are not sortable", "tups.sort"):
foo()
def test_tuple_unsortable_diff_type(self):
@torch.jit.script
def foo(inputs: List[Any]):
inputs.sort()
return inputs
inputs = [(1, 2), ("foo", "bar")]
with self.assertRaisesRegexWithHighlight(RuntimeError, "Only values of same type can be compared", "inputs.sort"):
foo(inputs)
def test_tuple_nested_sort(self):
def foo(inputs: List[Tuple[int, Tuple[int, str]]]):
inputs.sort()
return inputs
inputs = [(1, (2, "foo")), (1, (2, "bar")), (1, (0, "bar"))]
self.checkScript(foo, (inputs,))
def test_tuple_unsortable_nested_diff_type(self):
@torch.jit.script
def foo(inputs: List[Any]):
inputs.sort()
return inputs
inputs = [(1, (2, 3)), (2, ("foo", "bar"))]
with self.assertRaisesRegexWithHighlight(RuntimeError, "Only values of same type can be compared", "inputs.sort"):
foo(inputs)
def test_string_new_line(self):
with self.assertRaisesRegex(RuntimeError, "expected a valid token*"):
torch.jit.CompilationUnit('''
def test_while(a):
print("
a")
return a
''')
def test_string_single_escape(self):
with self.assertRaisesRegex(RuntimeError, "expected a valid token*"):
torch.jit.CompilationUnit('''
def test_while(a):
print("\\")
return a
''')
def test_script_annotation(self):
@torch.jit.script
def foo(a):
return a + a + a
s = Variable(torch.rand(2))
self.assertEqual(s + s + s, foo(s))
def test_torch_pow(self):
def func(a, b):
return pow(a, b)
def func2(a, b, c, d):
return pow(pow(c + a, b), d)
def func3(a : int, b : float):
# type: (int, float) -> float
return pow(a, b)
def func4():
# type: () -> float
return pow(2, -2)
def func5(x, y):
return pow(x.item(), y.item())
def func6(a : int, b : int):
# type: (int, int) -> float
return pow(a, b)
a = torch.rand(1)
b = torch.rand(1)
c = torch.rand(1)
d = torch.rand(1)
self.checkScript(func, (a, b))
self.checkScript(func2, (a, b, c, d))
self.checkScript(func3, (4, -0.5))
self.checkScript(func4, ())
self.checkScript(func6, (2, 4))
inputs = [torch.tensor(2), torch.tensor(-2), torch.tensor(.5), torch.tensor(.2)]
for x in inputs:
for y in inputs:
if x < 0:
continue
else:
self.checkScript(func5, (x, y))
@unittest.skipIf(not RUN_CUDA, "device tests require CUDA")
def test_pow_scalar_backward_cuda(self):
# see that scalar exponent works with cuda base (#19253)
with enable_profiling_mode_for_profiling_tests():
for dtype in [torch.float, torch.double]:
@torch.jit.script
def func(a, b):
# type: (Tensor, float) -> Tensor
return (a * 2) ** b
a = torch.rand(1, requires_grad=True, device='cuda', dtype=dtype)
func(a, 1, profile_and_replay=True).backward()
@torch.jit.script
def func(a, b):
# type: (float, Tensor) -> Tensor
return a ** (b * 2 + 1)
a = torch.rand(1, requires_grad=True, device='cuda', dtype=dtype)
func(2, a, profile_and_replay=True).backward()
def _check_code(self, code_str, fn_name, inputs):
scope = {}
exec(code_str, globals(), scope)
cu = torch.jit.CompilationUnit(code_str)
self.assertEqual(cu.func(*inputs), scope[fn_name](*inputs))
@unittest.skipIf(not RUN_CUDA, 'no CUDA')
def test_scriptmodule_releases_tensors_cuda(self):
with enable_profiling_mode_for_profiling_tests():
@torch.jit.script
def fn(x, y):
return x.sigmoid() * y.tanh()
def test(backward=False):
x = torch.randn(3, 3, dtype=torch.double, device='cuda', requires_grad=True)
y = torch.randn(3, 3, dtype=torch.double, device='cuda', requires_grad=True)
out = fn(x, y, profile_and_replay=True)
if backward:
out.sum().backward()
with self.assertLeaksNoCudaTensors():
test()
test()
test()
if GRAPH_EXECUTOR != ProfilingMode.SIMPLE:
with self.assertLeaksNoCudaTensors():
test(backward=True)
test(backward=True)
test(backward=True)
def test_index(self):
def consec(size, start=0):
numel = torch.tensor(size).prod().item()
return torch.arange(numel).view(size)
def consec_list(size):
return list(range(size))
def random_string(size):
letters = string.ascii_lowercase
return "".join(random.choice(letters) for i in range(size))
def check_indexing(indexing, tensor):
template = dedent("""
def func(x):
return x{}
""")
self._check_code(template.format(indexing), "func", [tensor])
def check_dynamic_indexing(indexing, tensor, value1, value2):
value1 = torch.tensor(value1)
value2 = torch.tensor(value2)
template = dedent("""
def func(x, value1, value2):
i = int(value1)
j = int(value2)
return x{}
""")
self._check_code(template.format(indexing), "func", [tensor, value1, value2])
# Torchscript assumes type Tensor by default, so we need this explicit
# declaration.
def check_indexing_list_int(indexing, list):
template = dedent("""
def func(x):
# type: (List[int]) -> Any
return x{}
""")
self._check_code(template.format(indexing), "func", [list])
def check_indexing_str(indexing, str):
template = dedent("""
def func(x):
# type: (str) -> Any
return x{}
""")
self._check_code(template.format(indexing), "func", [str])
# basic slices
check_indexing('[0]', consec((3, 3)))
check_indexing('[1]', consec((3, 3), 10))
check_indexing('[2]', consec((3, 3), 19))
check_indexing('[2]', consec((3,)))
check_indexing('[-1]', consec((3, 3), 19))
check_indexing('[0:2]', consec((3, 3, 3)))
check_indexing('[1:-1]', consec((3, 3, 3)))
check_indexing('[-3:-1]', consec((6, 3)))
check_indexing('[1:]', consec((3, 3)))
check_indexing('[:1]', consec((3, 3)))
check_indexing('[:]', consec((3, 2)))
# multi-dim: indexes
check_indexing('[0, 1]', consec((3, 3)))
check_indexing('[0, 1]', consec((3, 3, 2)))
check_indexing('[1, 0, 2]', consec((3, 3, 3)))
check_indexing('[2, -1]', consec((3, 3)))
# multi-dim: mixed slicing and indexing
check_indexing('[0, 1:2]', consec((3, 3)))
check_indexing('[0, :1]', consec((3, 3, 2)))
check_indexing('[1, 2:]', consec((3, 3, 3)))
check_indexing('[-1, 1:, 0]', consec((3, 3, 3, 3)))
check_indexing('[1:, -1, 0]', consec((3, 3, 3, 3)))
check_indexing('[-1, 2:, 1:2]', consec((3, 3, 3, 3)))
check_indexing('[-1, 1:, 0]', consec((3, 3, 3, 3)))
check_indexing('[-1, :, 0, 2]', consec((3, 3, 3, 3)))
# zero-sized slices
check_indexing('[0:0]', consec((2, 2)))
check_indexing('[0:0, 1]', consec((3, 3)))
# trivial expression usage
check_indexing('[1+1]', consec((3, 3)))
check_indexing('[1:(0 + 2)]', consec((3, 3, 3)))
# None for new dimensions
check_indexing('[None, 0]', consec((3, 3)))
check_indexing('[1, None]', consec((3, 3), 10))
check_indexing('[None, None, 2]', consec((3, 3), 19))
check_indexing('[None, 2, None]', consec((3,)))
check_indexing('[0:2, None]', consec((3, 3, 3)))
check_indexing('[None, 1:-1]', consec((3, 3, 3)))
check_indexing('[None, -3:-1, None]', consec((6, 3)))
check_indexing('[-1, None, 2:, None, 1:2]', consec((3, 3, 3, 3)))
check_indexing('[None, -1, None, 2:, None, 1:2, None]', consec((3, 3, 3, 3)))
# dynamic expression usage
check_dynamic_indexing("[i + j]", consec((3, 3)), 0, 1)
check_dynamic_indexing("[i:j, i]", consec((3, 3, 2)), 0, 2)
# positive striding
check_indexing_list_int('[0]', consec_list(6))
check_indexing_list_int('[1]', consec_list(7))
check_indexing_list_int('[2]', consec_list(8))
check_indexing_list_int('[2]', consec_list(9))
check_indexing_list_int('[-1]', consec_list(10))
check_indexing_list_int('[0:2]', consec_list(11))
check_indexing_list_int('[1:-1]', consec_list(12))
check_indexing_list_int('[-3:-1]', consec_list(13))
check_indexing_list_int('[1:]', consec_list(15))
check_indexing_list_int('[:1]', consec_list(16))
check_indexing_list_int('[:]', consec_list(17))
check_indexing_list_int('[::]', consec_list(0))
check_indexing_list_int('[1000::]', consec_list(0))
check_indexing_list_int('[:1000:]', consec_list(0))
# negative striding
check_indexing_list_int('[::-1]', consec_list(7))
check_indexing_list_int('[:3:-1]', consec_list(7))
check_indexing_list_int('[3::-1]', consec_list(7))
check_indexing_list_int('[1000::-1]', consec_list(7))
check_indexing_list_int('[3:0:-1]', consec_list(7))
check_indexing_list_int('[3:-1000:-1]', consec_list(7))
check_indexing_list_int('[0:0:-1]', consec_list(7))
check_indexing_list_int('[0:-1000:-1]', consec_list(7))
# only step is specified
check_indexing_list_int('[::-1]', consec_list(0))
check_indexing_list_int('[::-1]', consec_list(7))
check_indexing_list_int('[::-2]', consec_list(7))
check_indexing_list_int('[::2]', consec_list(7))
check_indexing_list_int('[::42]', consec_list(7))
check_indexing_list_int('[::-42]', consec_list(7))
check_indexing_list_int('[::42]', consec_list(0))
check_indexing_list_int('[::-42]', consec_list(0))
check_indexing_list_int('[::9223372036854775807]', consec_list(42))
check_indexing_list_int('[::-9223372036854775807]', consec_list(42))
with self.assertRaisesRegex(RuntimeError, "out of bounds"):
check_indexing_list_int('[::-9223372036854775808]', consec_list(42))
with self.assertRaisesRegex(RuntimeError, "should have non-zero step"):
check_indexing_list_int('[::0]', consec_list(42))
# striding strings
check_indexing_str('[0]', random_string(6))
check_indexing_str('[1]', random_string(7))
check_indexing_str('[2]', random_string(8))
check_indexing_str('[2]', random_string(9))
check_indexing_str('[-1]', random_string(10))
check_indexing_str('[0:2]', random_string(11))
check_indexing_str('[1:-1]', random_string(12))
check_indexing_str('[-3:-1]', random_string(13))
check_indexing_str('[1:]', random_string(15))
check_indexing_str('[:1]', random_string(16))
check_indexing_str('[:]', random_string(17))
check_indexing_str('[::]', random_string(0))
check_indexing_str('[1000::]', random_string(0))
check_indexing_str('[:1000:]', random_string(0))
check_indexing_str('[::-1]', random_string(7))
check_indexing_str('[:3:-1]', random_string(7))
check_indexing_str('[3::-1]', random_string(7))
check_indexing_str('[1000::-1]', random_string(7))
check_indexing_str('[3:0:-1]', random_string(7))
check_indexing_str('[3:-1000:-1]', random_string(7))
check_indexing_str('[0:0:-1]', random_string(7))
check_indexing_str('[0:-1000:-1]', random_string(7))
check_indexing_str('[::-1]', random_string(0))
check_indexing_str('[::-1]', random_string(7))
check_indexing_str('[::-2]', random_string(7))
check_indexing_str('[::2]', random_string(7))
check_indexing_str('[::42]', random_string(7))
check_indexing_str('[::-42]', random_string(7))
check_indexing_str('[::42]', random_string(0))
check_indexing_str('[::-42]', random_string(0))
check_indexing_str('[::9223372036854775807]', random_string(42))
check_indexing_str('[::-9223372036854775807]', random_string(42))
with self.assertRaisesRegex(RuntimeError, "out of bounds"):
check_indexing_str('[::-9223372036854775808]', random_string(42))
with self.assertRaisesRegex(RuntimeError, "should have non-zero step"):
check_indexing_str('[::0]', random_string(42))
def test_module_copy_with_attributes(self):
class Vocabulary(torch.jit.ScriptModule):
def __init__(self, vocab_list):
super(Vocabulary, self).__init__()
self._vocab = torch.jit.Attribute(vocab_list, List[str])
self.some_idx = torch.jit.Attribute(2, int)
self.idx = torch.jit.Attribute(
{word: i for i, word in enumerate(vocab_list)}, Dict[str, int]
)
@torch.jit.script_method
def lookup_indices_1d(self, values):
# type: (List[str]) -> List[int]
result = torch.jit.annotate(List[int], [])
# Direct list iteration not supported
for i in range(len(values)):
value = values[i]
result.append(self.idx.get(value, self.some_idx))
return result
@torch.jit.script_method
def forward(self, values):
# type: (List[List[str]]) -> List[List[int]]
result = torch.jit.annotate(List[List[int]], [])
# Direct list iteration not supported
for i in range(len(values)):
result.append(self.lookup_indices_1d(values[i]))
return result
v = Vocabulary(list('uabcdefg'))
v.__copy__()
def test_tuple_to_opt_list(self):
@torch.jit.script
def foo(x):
# type: (Optional[List[int]]) -> int
return 1
@torch.jit.script
def tuple_call():
return foo((1, 2))
def test_keyword(self):
@torch.jit.script
def func(x):
return torch.sum(x, dim=0)
x = torch.rand(10, dtype=torch.float, requires_grad=True)
y = func(x)
y2 = torch.sum(x, dim=0)
self.assertEqual(y, y2)
def test_constant_pooling_none(self):
@torch.jit.script
def typed_nones(a=None, b=None, c=None):
# type: (Optional[int], Optional[bool], Optional[Tensor]) -> Tuple[Optional[int], Optional[bool], Optional[Tensor]]
return a, b, c
@torch.jit.script
def test(a):
# type: (bool) -> None
if a:
print(typed_nones())
else:
print(typed_nones())
graph_str = str(test.graph)
self.assertTrue(graph_str.count("NoneType = prim::Constant") == 1)
def test_constant_pooling_same_identity(self):
def foo():
a = torch.tensor([4])
b = (a,)
index = len(a) - 1
c = b[index]
d = b[index]
return c, d
foo_script = torch.jit.script(foo)
self.run_pass('constant_propagation', foo_script.graph)
self.run_pass('constant_pooling', foo_script.graph)
# even though the c & d escape scope, we are still able
# pool them into one constant because they are the same object
FileCheck().check_count("prim::Constant", 1, exactly=True).run(foo_script.graph)
self.assertEqual(foo(), foo_script())
def test_constant_pooling_introduce_aliasing(self):
@torch.jit.script
def foo():
a = torch.tensor(1)
b = torch.tensor(1)
return a, b
self.run_pass('constant_propagation', foo.graph)
self.run_pass('constant_pooling', foo.graph)
# dont pool constants bc it would introduce observable alias relationship changing
a, b = foo()
self.assertIsNot(a, b)
def test_literal(self):
def func1(a, b):
c = a, b
d, e = c
return d + e
def func2(a, b):
c = a, (a, b)
d, e = c
f, g = e
return d + f + g
def func3(a, b):
# type: (float, float) -> float
c = 0., (0., 0.)
x = True
while x:
x = False
c = a, (a, b)
d, e = c
f, g = e
return d + f + g
a = torch.rand(1, requires_grad=True)
b = torch.rand(1, requires_grad=True)
self.checkScript(func1, (a, b), optimize=True)
self.checkScript(func2, (a, b), optimize=True)
self.checkScript(func3, (a.item(), b.item()), optimize=True)
def test_expand(self):
@torch.jit.script
def func(x, y):
return x + y
x = torch.rand(2, 3, dtype=torch.float, requires_grad=True)
y = torch.rand(3, dtype=torch.float, requires_grad=True)
out = func(x, y)
self.assertEqual(func(x, y), x + y)
grad = torch.randn(2, 3, dtype=torch.float)
out.backward(grad)
self.assertEqual(x.grad, grad)
self.assertEqual(y.grad, grad.sum(dim=0))
def test_sum(self):
@torch.jit.script
def func(x):
return x.sum(dim=[4])
@torch.jit.script
def func2(x):
return x.sum(dim=4)
# test that shape analysis is written correctly for sum with OptionalIntArrayRef[1] dim argument
self.run_pass('constant_propagation', func.graph)
self.run_pass('constant_propagation', func2.graph)
g = _propagate_shapes(func.graph, (torch.zeros(1, 1, 1, 1, 4),), False)
g2 = _propagate_shapes(func2.graph, (torch.zeros(1, 1, 1, 1, 4),), False)
def test_cat(self):
with enable_profiling_mode_for_profiling_tests():
@torch.jit.script
def func(x):
return torch.cat((x, x), dim=0)
x = torch.rand(10, dtype=torch.float, requires_grad=True)
self.assertEqual(func(x, profile_and_replay=True), torch.cat((x, x), dim=0))
@torch.jit.script
def func2(x, y):
return torch.cat((x, x), y)
with disable_autodiff_subgraph_inlining():
for sizes in ((2, 2), (0, 2)):
x = torch.rand(sizes).requires_grad_()
y = torch.tensor(1)
output = func2(x, y, profile_and_replay=True)
output_ref = torch.cat((x, x), y)
self.assertEqual(output, output_ref)
if GRAPH_EXECUTOR != ProfilingMode.SIMPLE:
self.assertAutodiffNode(func2.graph_for(x, y), True, ['aten::cat'], [])
grad = torch.autograd.grad(output.sum(), x)
grad_ref = torch.autograd.grad(output_ref.sum(), x)
self.assertEqual(grad, grad_ref)
def test_cat_lifts(self):
@torch.jit.script
def foo(x):
return torch.cat([x, x], dim=1)
@torch.jit.script
def foo2(x):
return torch.cat([], dim=1)
@torch.jit.script
def foo3(x):
return torch.cat([x], dim=1)
for g in [foo.graph, foo2.graph, foo3.graph]:
FileCheck().check("int =").check("ListConstruct").check("aten::cat").run(str(g))
def test_stack(self):
with enable_profiling_mode_for_profiling_tests():
@torch.jit.script
def func(x):
return torch.stack((x, x), dim=1)
x = torch.rand(10, 10)
self.assertEqual(func(x, profile_and_replay=True), torch.stack((x, x), dim=1))
@torch.jit.script
def func2(x, y):
return torch.stack((x, y), dim=0)
with disable_autodiff_subgraph_inlining():
x = torch.randn([2, 2]).requires_grad_()
y = torch.randn([2, 2]).requires_grad_()
output = func2(x, y, profile_and_replay=True)
output_ref = torch.stack((x, y), 0)
self.assertEqual(output, output_ref)
if GRAPH_EXECUTOR != ProfilingMode.SIMPLE:
self.assertAutodiffNode(func2.graph_for(x, y), True, ['aten::stack'], [])
grads = torch.autograd.grad(output.sum(), (x, y))
grads_ref = torch.autograd.grad(output_ref.sum(), (x, y))
self.assertEqual(grads, grads_ref)
@unittest.skipIf(GRAPH_EXECUTOR != ProfilingMode.LEGACY,
"Profiling executor will be using different heuristics for constructing differentiable graphs")
def test_unbind(self):
with enable_profiling_mode_for_profiling_tests():
@torch.jit.script
def func(x, y):
# type: (Tensor, int) -> List[Tensor]
return torch.unbind(x, y)
with disable_autodiff_subgraph_inlining():
x = torch.rand([2, 2]).requires_grad_()
y = 0
outputs = func(x, y, profile_and_replay=True)
outputs_ref = torch.unbind(x, dim=y)
self.assertEqual(outputs, outputs_ref)
self.assertAutodiffNode(func.graph_for(x, y), True, [], [])
grad = torch.autograd.grad(_sum_of_list(outputs), x)
grad_ref = torch.autograd.grad(_sum_of_list(outputs_ref), x)
self.assertEqual(grad, grad_ref)
@unittest.skipIf(GRAPH_EXECUTOR == ProfilingMode.PROFILING,
"Profiling executor fails to recognize that tensors in a list require gradients")
def test_meshgrid(self):
with enable_profiling_mode_for_profiling_tests():
@torch.jit.script
def func(a):
# type: (List[Tensor]) -> List[Tensor]
return torch.meshgrid(a)
with disable_autodiff_subgraph_inlining():
a = torch.tensor([1.0, 2, 3]).requires_grad_()
b = torch.tensor([1.0, 2, 3, 4]).requires_grad_()
inputs = [a, b]
outputs_ref = torch.meshgrid(inputs)
outputs = func(inputs, profile_and_replay=True)
self.assertEqual(outputs, outputs_ref)
if GRAPH_EXECUTOR != ProfilingMode.SIMPLE:
self.assertAutodiffNode(func.graph_for(inputs), True, [], [])
grads = torch.autograd.grad(_sum_of_list(outputs), inputs)
grads_ref = torch.autograd.grad(_sum_of_list(outputs_ref), inputs)
self.assertEqual(grads, grads_ref)
def test_tensor_len(self):
def func(x):
return len(x)
self.checkScript(func, [torch.ones(4, 5, 6)])
def test_func_call(self):
def add(a, b):
return a + b
def mul(a, x):
return a * x
def func(alpha, beta, x, y):
return add(mul(alpha, x), mul(beta, y))
alpha = torch.rand(1, dtype=torch.float, requires_grad=True)
beta = torch.rand(1, dtype=torch.float, requires_grad=True)
x = torch.rand(3, dtype=torch.float, requires_grad=True)
y = torch.rand(3, dtype=torch.float, requires_grad=True)
# NOTE: cannot optimize yet because broadcasts are not inserted before the fuser runs
self.checkScript(func, [alpha, beta, x, y], optimize=False)
@unittest.skip("bailouts are being deprecated")
def test_profiling_graph_executor(self):
@torch.jit.script
def def_in_one_branch(x, z):
# type: (Tensor, bool) -> float
y = x
if z is False:
y = x + 1
return y.sum()
a = torch.rand(2, 3)
with enable_profiling_mode_for_profiling_tests():
# check prim::profile are inserted
profiled_graph_str = str(def_in_one_branch.graph_for(a, True))
FileCheck().check_count("prim::profile", 4).run(profiled_graph_str)
# this call is optimized for
# the given shape of (2, 3)
def_in_one_branch(a, False)
# change shape to (3)
# so we go down a bailout path
a = torch.ones(3)
# check prim::BailOuts are inserted
bailout_graph_str = str(def_in_one_branch.graph_for(a, True))
FileCheck().check_count("prim::BailOut", 3).run(bailout_graph_str)
# this triggers all 3 bailouts
self.assertEqual(def_in_one_branch(a, False), 6.0)
# this triggers 2 bailouts
self.assertEqual(def_in_one_branch(a, True), 3.0)
@unittest.skip("bailouts are being deprecated")
def test_maxpool_guard_elimination(self):
@torch.jit.script
def my_maxpool(x):
return F.max_pool1d(x, kernel_size=[1]) + torch.ones([32, 32, 32])
a = torch.rand(32, 32, 32)
with enable_profiling_mode_for_profiling_tests():
my_maxpool(a)
bailout_graph_str = str(my_maxpool.graph_for(a))
FileCheck().check_count("prim::BailOut", 1).run(bailout_graph_str)
@unittest.skip("bailouts are being deprecated")
def test_slice_guard_elimination(self):
@torch.jit.script
def my_slice(x):
return x[0:16:2] + x[0:16:2]
a = torch.rand(32, 4)
with enable_profiling_mode_for_profiling_tests():
my_slice(a)
bailout_graph_str = str(my_slice.graph_for(a))
FileCheck().check_count("prim::BailOut", 1).run(bailout_graph_str)
@unittest.skip("bailouts are being deprecated")
def test_unsqueeze_guard_elimination(self):
@torch.jit.script
def my_unsqueeze(x):
return torch.unsqueeze(x, 0) + torch.unsqueeze(x, 0)
a = torch.rand(32, 4)
with enable_profiling_mode_for_profiling_tests():
my_unsqueeze(a)
bailout_graph_str = str(my_unsqueeze.graph_for(a))
FileCheck().check_count("prim::BailOut", 2).run(bailout_graph_str)
def test_resize_input_ops(self):
# resize_ and resize_as resize the input tensor. because our shape analysis
# is flow invariant, we set any Tensor that can alias a resized Tensor
# to the base Tensor Type, without size information.
# testing that value which is an input of a graph gets handled
def out_op_graph_input():
@torch.jit.script
def test(x, y, z):
torch.mul(x, y, out=z)
return z
graph = _propagate_shapes(test.graph,
(torch.zeros(2, 1), torch.zeros(1, 2), torch.zeros(1, 1, 1)), False)
self.assertTrue(next(graph.outputs()).type() == TensorType.get())
out_op_graph_input()
def test_resize():
@torch.jit.script
def test(x):
after_resize_alias = torch.zeros([2])
for _i in range(5):
b = x + 1
f = [1]
before_resize_alias = b.sub_(1)
# for i in range(10):
f.append(1)
b.resize_(f)
after_resize_alias = b.add_(1)
return after_resize_alias
self.run_pass('constant_propagation', test.graph)
g = _propagate_shapes(test.graph, (torch.zeros(1, 1),), False)
resize_node = g.findNode("aten::resize_")
# first input and output of b.resize_ is b
self.assertTrue(next(resize_node.inputs()).type() == TensorType.get())
self.assertTrue(next(resize_node.outputs()).type() == TensorType.get())
# correctly propagates to b alias set
before_resize = g.findNode("aten::sub_")
self.assertTrue(next(before_resize.outputs()).type() == TensorType.get())
after_resize = g.findNode("aten::add_")
self.assertTrue(next(after_resize.outputs()).type() == TensorType.get())
test_resize()
def test_resize_as():
@torch.jit.script
def test(x):
b = torch.zeros([2, 2])
b.resize_as_(x)
return b
g = test.graph
self.run_pass('constant_propagation', g)
g = _propagate_shapes(test.graph, (torch.zeros(1, 1),), False)
# x doesn't alias a resized op so it shouldn't be set to base Tensor type
self.assertTrue(next(g.inputs()).type() != TensorType.get())
# return is resized
self.assertTrue(next(g.outputs()).type() == TensorType.get())
test_resize_as()
def test_uninitialized(self):
graph_str = """graph():
%1 : int = prim::Uninitialized()
%2 : int = prim::Constant[value=1]()
%3 : int = aten::add(%1, %2)
return (%3)
"""
g = parse_ir(graph_str)
m = self.createFunctionFromGraph(g)
self.getExportImportCopy(m)
with self.assertRaisesRegex(RuntimeError, "isInt"):
m()
@unittest.skipIf(GRAPH_EXECUTOR == ProfilingMode.SIMPLE, "Simple Executor doesn't use requires_grad information")
@unittest.skipIf(GRAPH_EXECUTOR == ProfilingMode.PROFILING, "Peeling is now disabled")
def test_requires_grad_loop(self):
@torch.jit.script
def test(x, y, z):
# type: (Tensor, Tensor, int) -> Tensor
for _ in range(z):
x = y
return x
# x requires grad, y does not
# testing that requires grad analysis correctly exits, with its input
# to the loop (x) requiring grad and its output to the loop not requiring grad
# and the output of the node conservatively setting grad to true
inps = (torch.tensor(1.0, requires_grad=True), torch.tensor(1), 10)
test(*inps, profile_and_replay=True)
graph = test.graph_for(*inps)
loop = graph.findNode("prim::Loop")
loop_body = next(loop.blocks())
loop_inputs = list(loop_body.inputs())
loop_outputs = list(loop_body.outputs())
if GRAPH_EXECUTOR == ProfilingMode.PROFILING:
# TODO: simplify this test as it's very sensitive
# the optimized graph will have 3 loops
# the original loop is peeled
# peeled loop also gets unrolled
index_of_x_in_peeled_unrolled_loop = -2
self.assertTrue(loop_inputs[index_of_x_in_peeled_unrolled_loop].requires_grad())
bailouts_in_outer_block = graph.findAllNodes("prim::BailOut", False)
last_bailout_index_on_loops_output = -1
self.assertFalse(bailouts_in_outer_block[last_bailout_index_on_loops_output].output().requires_grad())
else:
self.assertTrue(loop_inputs[1].requires_grad())
self.assertTrue(loop.output().requires_grad())
self.assertFalse(loop_outputs[1].requires_grad())
def test_view_shape_prop(self):
cu = torch.jit.CompilationUnit('''
def test_view_shape_prop(a):
return a.view(size=[-1])
''')
inputs = [torch.zeros(10, 10)]
outputs = torch.zeros(100)
real_outs = cu.test_view_shape_prop(*inputs)
self.assertEqual(real_outs, outputs)
def test_view_listconstruct_shape_prop(self):
def fn(x):
B = x.size(0)
C = x.size(1)
T = x.size(2)
return x.view(T, B, C)
x = torch.randn(3, 1, 5, requires_grad=True)
fn = torch.jit.script(fn)
graph = _propagate_shapes(fn.graph, (x,), False)
self.assertTrue(next(graph.outputs()).type().scalarType() == 'Double')
def test_shape_prop_promotion(self):
@torch.jit.script
def fn(x, y):
return x + y
x, y = torch.rand(3, 4, dtype=torch.float), torch.rand(3, 4, dtype=torch.double)
graph = _propagate_shapes(fn.graph, (x, y), False)
FileCheck().check('Double(*, *, device=cpu) = aten::add').run(graph)
def test_shape_prop_promote_scalar_arg(self):
@torch.jit.script
def fn(x):
return math.pi + x
x = torch.zeros(3, 4, dtype=torch.long)
graph = _propagate_shapes(fn.graph, (x,), False)
default = torch.get_default_dtype()
if(default == torch.float):
FileCheck().check('Float(*, *, requires_grad=0, device=cpu) = aten::add').run(graph)
else:
FileCheck().check('Double(*, *, requires_grad=0, device=cpu) = aten::add').run(graph)
def test_integral_shape_inference(self):
cu = torch.jit.CompilationUnit('''
def test_integral_shape_inference(a):
return a * a
''')
inputs = [torch.ones(10, 10, dtype=torch.long)]
outputs = torch.ones(10, 10, dtype=torch.long)
self.assertEqual(cu.test_integral_shape_inference(*inputs), outputs)
@unittest.skipIf(RUN_CUDA, 'This tests the CPU fuser')
@unittest.skipIf(IS_SANDCASTLE, "NYI: fuser support for Sandcastle")
@enable_cpu_fuser
def test_batchnorm_fuser_cpu(self):
code = '''
graph(%3 : Tensor,
%7 : Tensor,
%12 : Float(*, *),
%13 : Tensor,
%25 : Tensor):
%23 : int = prim::Constant[value=1]()
%22 : float = prim::Constant[value=1e-05]()
%26 : Tensor = aten::sqrt(%25)
%24 : Tensor = aten::add(%26, %22, %23)
%20 : Tensor = aten::reciprocal(%24)
%norm_invstd : Tensor = aten::mul(%20, %23)
%15 : Tensor = aten::sub(%12, %13, %23)
%11 : Tensor = aten::mul(%15, %norm_invstd)
%8 : Tensor = aten::mul(%11, %7)
%5 : Tensor = aten::add(%8, %3, %23)
%1 : Float(*, *) = aten::relu(%5)
return (%1)
'''
graph = parse_ir(code)
inputs = 5 * [torch.rand(26, 2048, dtype=torch.float)]
code = torch._C._jit_fuser_get_fused_kernel_code(graph, inputs)
FileCheck().check('sqrtf').run(code)
@slowTest
@unittest.skipIf(RUN_CUDA, 'This tests the CPU fuser')
@unittest.skipIf(IS_SANDCASTLE, "NYI: fuser support for Sandcastle")
@enable_cpu_fuser
def test_fuser_double_float_codegen(self):
fns = ['log', 'log10', 'log1p', 'log2', 'lgamma', 'exp', 'expm1', 'erf',
'erfc', 'cos', 'acos', 'cosh', 'sin', 'asin', 'sinh', 'tan',
'atan', 'tanh', 'sqrt', 'ceil', 'floor', 'round', 'trunc',
'frac']
def lookup_c_equivalent_fn(aten_fn):
return aten_fn
def test_dispatch(op, expects, dtype, binary=False):
if dtype == torch.double:
dtype_str = 'Double'
elif dtype == torch.float:
dtype_str = 'Float'
else:
raise RuntimeError('Unknown dtype')
if binary:
code = '''
graph(%3 : Tensor, %4 : Tensor):
%2 : {dtype}(*, *) = aten::{op}(%3, %4)
%1 : {dtype}(*, *) = aten::relu(%2)
return (%1)
'''.format(op=op, dtype=dtype_str)
else:
code = '''
graph(%3 : Tensor):
%2 : {dtype}(*, *) = aten::{op}(%3)
%1 : {dtype}(*, *) = aten::relu(%2)
return (%1)
'''.format(op=op, dtype=dtype_str)
graph = parse_ir(code)
inputs = (2 if binary else 1) * [torch.rand(26, 2048, dtype=dtype)]
code = torch._C._jit_fuser_get_fused_kernel_code(graph, inputs)
FileCheck().check(expects).run(code)
for fn in fns:
test_dispatch(fn, lookup_c_equivalent_fn(fn) + '(', torch.double)
test_dispatch(fn, lookup_c_equivalent_fn(fn) + 'f(', torch.float)
# 'min', 'max' were previously tested but are now replaced with ternary expressions
# instead of fmin() and fmax()
binary_fns = ['pow']
for fn in binary_fns:
test_dispatch(fn, lookup_c_equivalent_fn(fn) + '(', torch.double, binary=True)
test_dispatch(fn, lookup_c_equivalent_fn(fn) + 'f(', torch.float, binary=True)
@unittest.skipIf(RUN_CUDA, 'This tests the CPU fuser')
@unittest.skipIf(IS_SANDCASTLE, "NYI: fuser support for Sandcastle")
@enable_cpu_fuser
def test_fuser_double_literal_precision(self):
code = '''
graph(%2 : Float(*, *)):
%4 : int = prim::Constant[value=1]()
%3 : float = prim::Constant[value=1.282549830161864]()
%5 : Float(*, *) = aten::add(%2, %3, %4)
%1 : Float(*, *) = aten::relu(%5)
return (%1)
'''
graph = parse_ir(code)
code = torch._C._jit_fuser_get_fused_kernel_code(graph, [torch.rand(3, 4)])
FileCheck().check('1.282549830161864').run(code)
def test_fuser_multiple_blocks(self):
cu = torch.jit.CompilationUnit('''
def test_fuser_multiple_blocks(this, that, theother, meme):
i = 0
while i < 20:
this = torch.cat([this, meme], dim=0)
that = torch.cat([that, meme], dim=0)
theother = torch.cat([theother, meme], dim=0)
i = i + 1
return this, that, theother
''')
inputs = [torch.ones(0, 10, 10)] * 3
inputs += [torch.ones(1, 10, 10)]
outputs = [torch.ones(20, 10, 10)] * 3
self.assertEqual(cu.test_fuser_multiple_blocks(*inputs), outputs)
def test_dropout_script(self):
eg = torch.zeros(1, 2, 3, requires_grad=True)
@_trace(eg)
def foo(x):
x = torch.neg(x)
return F.dropout(x)
class MyDrop(nn.Module):
def forward(self, x):
return foo(x)
f = io.BytesIO()
with warnings.catch_warnings(record=True):
torch.onnx.export(MyDrop(), (eg,), f, verbose=False)
@unittest.skip("RuntimeError: VariableType::ID() not implemented")
def test_cast(self):
script = '''
def to_int(x):
return int(x)
'''
x = Variable(torch.FloatTensor([1.1, 2.3]), requires_grad=True)
out = Variable(torch.IntTensor([1, 2]), requires_grad=True)
self.checkScript(script, [x], optimize=True, outputs=[out], func='to_int')
def test_str_cast(self):
@torch.jit.script
def to_str(x):
# type: (int) -> str
return str((x, x))
self.assertEqual("(1, 1)", to_str(1))
def test_int_cast(self):
@torch.jit.script
def to_int(x):
# type: (str) -> int
return int(x)
self.assertEqual(5, to_int('5'))
self.assertEqual(-5, to_int('-5'))
self.assertEqual(2147483647, to_int('2147483647'))
self.assertEqual(-2147483648, to_int('-2147483648'))
with self.assertRaisesRegex(RuntimeError, "invalid literal for int()"):
to_int('0x20')
with self.assertRaisesRegex(RuntimeError, "invalid literal for int()"):
to_int('0b0001')
def test_python_frontend(self):
def fn(x, y, z):
q = None
q = x + y - z.sigmoid()
print(q)
w = -z
if not x and not y and z:
m = x if not z else y
while x < y > z:
q = x
assert 1 == 1, "hello"
return x
ast = torch.jit.frontend.get_jit_def(fn, fn.__name__)
self.assertExpected(str(ast))
def test_python_frontend_source_range(self):
def fn():
raise Exception("hello")
ast = torch.jit.frontend.get_jit_def(fn, fn.__name__)
FileCheck().check("SourceRange at:") \
.check("def fn():") \
.check("~~~~~~~~~") \
.check('raise Exception("hello")') \
.check('~~~~~~~~~~~~~~~~~ <--- HERE') \
.run(str(ast.range()))
def test_python_frontend_py3(self):
def fn():
raise Exception("hello")
ast = torch.jit.frontend.get_jit_def(fn, fn.__name__)
self.assertExpected(str(ast))
def _make_scalar_vars(self, arr, dtype):
return [torch.tensor(val, dtype=dtype) for val in arr]
def test_string_print(self):
def func(a):
print(a, "a" 'b' '''c''' """d""", 2, 1.5)
return a
inputs = self._make_scalar_vars([1], torch.int64)
self.checkScript(func, inputs, capture_output=True)
def test_while(self):
def func(a, b, max):
while bool(a < max):
a = a + 1
b = b + 1
c = a + b
return c
inputs = self._make_scalar_vars([1, 1, 10], torch.int64)
self.checkScript(func, inputs, optimize=True)
def test_fibb(self):
def func(lim):
first = 1
second = 1
i = 1
somenum = 5
dontmutateme = 3
third = 0
while bool(i < lim):
third = first + second
first = second
second = third
j = 0
while j < 10:
somenum = somenum * 2
j = j + 1
i = i + j
i = i + dontmutateme
st = second + third
fs = first + second
return third, st, fs
inputs = self._make_scalar_vars([10], torch.int64)
self.checkScript(func, inputs, optimize=True)
def test_fibb_totally_better(self):
def fib(x):
# type: (int) -> int
prev = 1
v = 1
for i in range(0, x):
save = v
v = v + prev
prev = save
return v
self.checkScript(fib, (10,))
def test_if(self):
def func(a, b):
# type: (int, int) -> int
d = 3
if bool(a > 10):
a = 3 + d
else:
b = 3 + d
d = 4
c = a + b
return c
inputs = self._make_scalar_vars([1, -1], torch.int64)
self.checkScript(func, inputs, optimize=True)
def test_if_for_in_range(self):
def func(a, b):
# type: (int, int) -> int
d = 3
for _ in range(20):
if bool(a > 10):
a = 3 + d
else:
b = 3 + d
d = 4
c = a + b
return d
inputs = self._make_scalar_vars([1, -1], torch.int64)
self.checkScript(func, inputs, optimize=True)
def test_if_noelse(self):
def func(a, b):
if bool(a > 10):
a = 3 + b
c = a + b
return c
inputs = self._make_scalar_vars([-1, 1], torch.int64)
self.checkScript(func, inputs, optimize=True)
def test_if_is_none_dispatch(self):
@torch.jit.script
def test_lhs_none_rhs_none():
# LHS, RHS both alwaysNone, dispatch always_none_branch
# only emit one prim::Constant
if None is None:
return 1
elif None is not None:
return 2
else:
return 3
self.assertTrue(str(test_lhs_none_rhs_none.graph).count(': int = prim::Constant') == 1)
@torch.jit.script
def test_lhs_opt_rhs_none(lhs=None):
# type: (Optional[Tensor]) -> int
# LHS maybeNone: emit normal if stmt that contains 3 constants
if lhs is not None:
return 2
elif lhs is None:
return 1
else:
return 3
self.assertTrue(str(test_lhs_opt_rhs_none.graph).count(': int = prim::Constant') == 3)
@torch.jit.script
def test_lhs_none_rhs_opt(rhs=None):
# type: (Optional[Tensor]) -> int
# RHS maybeNone, emit normal if stmt that contains 3 constants
if None is rhs:
return 1
elif None is not rhs:
return 2
else:
return 3
self.assertTrue(str(test_lhs_opt_rhs_none.graph).count(': int = prim::Constant') == 3)
@torch.jit.script
def test_lhs_never_rhs_none(lhs):
# LHS neverNone, RHS alwaysNone dispatch never_none_branch
# only emit one prim::Constant
if lhs is None:
return 1
elif lhs is not None:
return 2
else:
return 3
self.assertTrue(str(test_lhs_never_rhs_none.graph).count(': int = prim::Constant') == 1)
@torch.jit.script
def test_lhs_none_rhs_never(rhs):
# LHS alwaysNone, RHS neverNone dispatch never_none_branch
# only emit one prim::Constant
if None is rhs:
return 1
elif None is not rhs:
return 2
else:
return 3
self.assertTrue(str(test_lhs_none_rhs_never.graph).count(': int = prim::Constant') == 1)
@torch.jit.script
def test_bool_arith_and(lhs):
if lhs is None and lhs is not None:
return 1
else:
return 2
self.assertEqual(test_bool_arith_and(torch.zeros(3)), 2)
self.assertTrue(str(test_bool_arith_and.graph).count('if') == 0)
@torch.jit.script
def test_bool_arith_or(lhs):
if lhs is None or lhs is not None:
return 1
else:
return 2
self.assertEqual(test_bool_arith_or(torch.zeros(3)), 1)
self.assertTrue(str(test_bool_arith_or.graph).count('if') == 0)
@torch.jit.script
def test_bool_arith_not(lhs):
if not (lhs is None):
return 1
else:
return 2
self.assertEqual(test_bool_arith_not(torch.zeros(3)), 1)
self.assertTrue(str(test_bool_arith_not.graph).count('if') == 0)
def test_conditional_casting(self):
def test_bool_cast_tensor(x):
if x:
return 1
else:
return 0
for make_one_dim in [True, False]:
for inp_val in [0.1, 0.0, -0.0, -0.1, -1, 0, 1]:
inp_val = [inp_val] if make_one_dim else inp_val
self.checkScript(test_bool_cast_tensor, (torch.tensor(inp_val),))
self.checkScriptRaisesRegex(test_bool_cast_tensor, (torch.tensor([1, 1]),), Exception,
"Boolean value of Tensor with more than one value")
def test_not_cast(x):
if not x:
return 1
else:
return 0
self.checkScript(test_not_cast, (torch.tensor(1),))
self.checkScript(test_not_cast, (torch.tensor(0),))
with self.assertRaisesRegex(RuntimeError, r"Could not cast value of type Tuple\[Tensor, Tensor\]"): # noqa: W605
@torch.jit.script
def test_mult(x, y):
return not(x, y)
def test_cast_int(x):
# type: (int) -> int
if x:
return 1
else:
return 0
self.checkScript(test_cast_int, (1,))
self.checkScript(test_cast_int, (0,))
self.checkScript(test_cast_int, (-1,))
def test_cast_float(x):
# type: (float) -> int
if x:
return 1
else:
return 0
self.checkScript(test_cast_float, (1.,))
self.checkScript(test_cast_float, (0.,))
self.checkScript(test_cast_float, (-1.,))
with self.assertRaisesRegex(RuntimeError, r"Could not cast value of type Tuple\[int, int\] to bool"): # noqa: W605
@torch.jit.script
def test_bad_conditional(x):
if (1, 2): # noqa: F634
return
else:
return 0
def test_while_nonexistent_value(self):
with self.assertRaisesRegex(RuntimeError, "undefined value x"):
torch.jit.CompilationUnit('''
def test_while(a, b):
while bool(a < 10):
a = a + x
b = b + 1
return a + b
''')
def test_while_nonexistent_cond_value(self):
with self.assertRaisesRegex(RuntimeError, "undefined value x"):
torch.jit.CompilationUnit('''
def test_while(a, b):
while a < x:
a = a + 1
b = b + 1
return a + b
''')
@torch.jit.script
def test_ternary(x):
# type: (Optional[int]) -> int
x = x if x is not None else 2
return x
@torch.jit.script
def test_not_none(x):
# type: (Optional[int]) -> None
if x is not None:
print(x + 1)
@torch.jit.script
def test_and(x, y):
# type: (Optional[int], Optional[int]) -> None
if x is not None and y is not None:
print(x + y)
@torch.jit.script
def test_not(x, y):
# type: (Optional[int], Optional[int]) -> None
if not (x is not None and y is not None):
pass
else:
print(x + y)
@torch.jit.script
def test_bool_expression(x):
# type: (Optional[int]) -> None
if x is not None and x < 2:
print(x + 1)
@torch.jit.script
def test_nested_bool_expression(x, y):
# type: (Optional[int], Optional[int]) -> int
if x is not None and x < 2 and y is not None:
x = x + y
else:
x = 5
return x + 2
@torch.jit.script
def test_or(x, y):
# type: (Optional[int], Optional[int]) -> None
if y is None or x is None:
pass
else:
print(x + y)
# backwards compatibility
@torch.jit.script
def test_manual_unwrap_opt(x):
# type: (Optional[int]) -> int
if x is None:
x = 1
else:
x = torch.jit._unwrap_optional(x)
return x # noqa: T484
with self.assertRaisesRegex(RuntimeError, "Arguments for call are not valid"):
@torch.jit.script
def or_error(x, y):
# type: (Optional[int], Optional[int]) -> None
if x is None or y is None:
print(x + y) # noqa: T484
with self.assertRaisesRegex(RuntimeError, "Arguments for call are not valid"):
@torch.jit.script
def and_error(x, y):
# type: (Optional[int], Optional[int]) -> None
if x is None and y is None:
pass
else:
print(x + y) # noqa: T484
with self.assertRaisesRegex(RuntimeError, "Arguments for call are not valid"):
@torch.jit.script
def named_var(x):
# type: (Optional[int]) -> None
x_none = x is not None
if x_none:
print(x + 1) # noqa: T484
with self.assertRaisesRegex(RuntimeError, "Arguments for call are not valid"):
@torch.jit.script
def named_var_and(x, y):
# type: (Optional[int], Optional[int]) -> None
x_none = x is not None
if y is not None and x_none:
print(x + y) # noqa: T484
def test_assertion_optional_refinement(self):
@torch.jit.script
def test(x, y):
# type: (Optional[int], Optional[int]) -> int
assert x is not None and y is not None
return x + y
self.assertEqual(test(2, 2), 4)
with self.assertRaisesRegex(Exception, ""):
test(1, None)
@unittest.skipIf(GRAPH_EXECUTOR != ProfilingMode.LEGACY, "the current version of Profiler doesn't profile/specialize Optionals")
def test_optional_tensor(self):
@torch.jit.script
def fn(x, y):
# type: (Optional[Tensor], int) -> int
if x is None:
return y
else:
return 0
res = fn(None, 1)
self.assertEqual(res, 1)
g = torch.jit.last_executed_optimized_graph()
first_input = next(g.inputs())
# check if input is disconnected
self.assertEqual(first_input.type().kind(), 'OptionalType')
self.assertEqual(first_input.uses(), [])
t = torch.ones(1)
res = fn(t, 1)
self.assertEqual(res, 0)
g = torch.jit.last_executed_optimized_graph()
self.assertEqual(next(g.inputs()).type().kind(), 'TensorType')
@torch.jit.script
def fn(x, y, b):
# type: (Optional[Tensor], Tensor, bool) -> Tensor
if b:
res = y
else:
res = torch.jit._unwrap_optional(x)
return res
t2 = torch.zeros(1)
res = fn(t, t2, True)
self.assertEqual(res, t2)
with self.assertRaisesRegex(RuntimeError, "Unwrapping null optional"):
res = fn(None, t2, False)
res = fn(None, t2, True)
g = torch.jit.last_executed_optimized_graph()
self.assertIn(next(g.outputs()).type().str(), ("Tensor", "Tensor(requires_grad=1)"))
@unittest.skipIf(GRAPH_EXECUTOR != ProfilingMode.LEGACY, "the current version of Profiler doesn't profile/specialize Optionals")
def test_optional_list(self):
@torch.jit.script
def fn(x, y):
# type: (Optional[List[int]], int) -> int
if x is None:
return y
else:
res = 0
for d in x:
res += d
return res
res = fn(None, 1)
self.assertEqual(res, 1)
g = torch.jit.last_executed_optimized_graph()
first_input = next(g.inputs())
# check if input is disconnected
self.assertEqual(first_input.type().kind(), 'OptionalType')
self.assertEqual(first_input.uses(), [])
l = [2, 3]
res = fn(l, 1)
self.assertEqual(res, 5)
g = torch.jit.last_executed_optimized_graph()
self.assertEqual(next(g.inputs()).type().kind(), 'ListType')
@torch.jit.script
def fn(x, y, b):
# type: (Optional[List[int]], List[int], bool) -> List[int]
if b:
l = torch.jit._unwrap_optional(x)
else:
l = y
return l
l2 = [0, 1]
res = fn(l, l2, True)
self.assertEqual(res, l)
with self.assertRaisesRegex(RuntimeError, "Unwrapping null optional"):
res = fn(None, l2, True)
res = fn(None, l2, False)
g = torch.jit.last_executed_optimized_graph()
self.assertEqual(next(g.outputs()).type().str(), "int[]")
def test_alias_covariant_type_containers(self):
@torch.jit.script
def foo(x):
# type: (bool)
if x:
a = (None,)
else:
a = ([],)
return a
@torch.jit.script
def foo2(x, li):
# type: (bool, Tuple[Optional[List[Tensor]]])
if x:
li = (None,)
return li
def test_while_write_outer_then_read(self):
def func(a, b):
while bool(a < 10):
a = a + 1
b = a + 1
return a + b
inputs = self._make_scalar_vars([42, 1337], torch.int64)
self.checkScript(func, inputs, optimize=True)
@skipIfTorchDynamo("TorchDynamo fails with unknown reason")
def test_while_nest_if(self):
def func(a, b):
# type: (int, int) -> int
c = 0
while a < 10:
a = a + 1
b = b + 1
if a > b:
c = -a
else:
c = -b
return c + 1
inputs = self._make_scalar_vars([-1234, 4321], torch.int64)
self.checkScript(func, inputs, optimize=True)
def test_divmod(self):
def func_int(a, b):
# type: (int, int) -> Tuple[int, int]
return divmod(a, b)
def func_float(a, b):
# type: (float, float) -> Tuple[float, float]
return divmod(a, b)
def func_int_float(a, b):
# type: (int, float) -> Tuple[float, float]
return divmod(a, b)
def func_float_int(a, b):
# type: (float, int) -> Tuple[float, float]
return divmod(a, b)
def divmod_test_iterator(func, num, den):
for i in num:
for j in den:
self.checkScript(func, (i, j), frames_up=2)
num_int = [1024, -1024]
den_int = [10, -10]
num_float = [5.3, -5.3]
den_float = [2.0, -2.0]
divmod_test_iterator(func_int, num_int, den_int)
divmod_test_iterator(func_float, num_float, den_float)
divmod_test_iterator(func_int_float, num_int, den_float)
divmod_test_iterator(func_float_int, num_float, den_int)
with self.assertRaisesRegex(RuntimeError, "ZeroDivisionError: integer division or modulo by zero"):
cu = torch.jit.CompilationUnit(dedent(inspect.getsource(func_int)))
cu.func_int(1024, 0)
with self.assertRaisesRegex(RuntimeError, "ZeroDivisionError: float divmod()"):
cu = torch.jit.CompilationUnit(dedent(inspect.getsource(func_float)))
cu.func_float(5.3, 0.0)
with self.assertRaisesRegex(RuntimeError, "ZeroDivisionError: float divmod()"):
cu = torch.jit.CompilationUnit(dedent(inspect.getsource(func_int_float)))
cu.func_int_float(1024, 0.0)
with self.assertRaisesRegex(RuntimeError, "ZeroDivisionError: float divmod()"):
cu = torch.jit.CompilationUnit(dedent(inspect.getsource(func_float_int)))
cu.func_float_int(5.3, 0)
def test_math_ops(self):
def checkMathWrap(func_name, num_args=1, is_float=True, **args):
if is_float:
checkMath(func_name, num_args, True, **args)
checkMath(func_name, num_args, False, **args)
else:
checkMath(func_name, num_args, is_float, **args)
inf = float("inf")
NaN = float("nan")
mx_int = 2**31 - 1
mn_int = -2**31
float_vals = ([inf, NaN, 0.0, 1.0, 2.2, -1.0, -0.0, -2.2, -inf, 1, 0, 2] +
[10.0 ** i for i in range(5)] + [-(10.0 ** i) for i in range(5)])
int_vals = list(range(-5, 5, 1)) + [mx_int + 5, mx_int * 2, mn_int - 5, mn_int * 2]
def checkMath(func_name, num_args, is_float=True, ret_type="float", debug=False, vals=None, args_type=None):
funcs_template = dedent('''
def func(a, b):
# type: {args_type} -> {ret_type}
return math.{func}({args})
''')
if num_args == 1:
args = "a"
elif num_args == 2:
args = "a, b"
else:
raise RuntimeError("Test doesn't support more than 2 arguments")
if args_type is None:
args_type = "(float, float)" if is_float else "(int, int)"
funcs_str = funcs_template.format(func=func_name, args=args, args_type=args_type, ret_type=ret_type)
scope = {}
execWrapper(funcs_str, globals(), scope)
cu = torch.jit.CompilationUnit(funcs_str)
f_script = cu.func
f = scope['func']
if vals is None:
vals = float_vals if is_float else int_vals
vals = [(i, j) for i in vals for j in vals]
for a, b in vals:
res_python = None
res_script = None
try:
res_python = f(a, b)
except Exception as e:
res_python = e
try:
res_script = f_script(a, b)
except Exception as e:
res_script = e
if debug:
print("in: ", a, b)
print("out: ", res_python, res_script)
# We can't use assertEqual because of a couple of differences:
# 1. nan == nan should return true
# 2. When python functions throw an exception, we usually want to silently ignore them.
# (ie: We want to return `nan` for math.sqrt(-5))
if res_python != res_script:
if isinstance(res_python, Exception):
continue
if type(res_python) == type(res_script):
if isinstance(res_python, tuple) and (math.isnan(res_python[0]) == math.isnan(res_script[0])):
continue
if isinstance(res_python, float) and math.isnan(res_python) and math.isnan(res_script):
continue
msg = ("Failed on {func_name} with inputs {a} {b}. Python: {res_python}, Script: {res_script}"
.format(func_name=func_name, a=a, b=b, res_python=res_python, res_script=res_script))
self.assertEqual(res_python, res_script, msg=msg, atol=(1e-4) * max(abs(res_python), res_script), rtol=0)
unary_float_ops = ["log", "log1p", "log10", "exp", "sqrt", "gamma", "lgamma", "erf",
"erfc", "expm1", "fabs", "acos", "asin", "atan", "cos", "sin", "tan",
"asinh", "atanh", "acosh", "sinh", "cosh", "tanh", "degrees", "radians"]
binary_float_ops = ["atan2", "fmod", "copysign"]
for op in unary_float_ops:
checkMathWrap(op, 1)
for op in binary_float_ops:
checkMathWrap(op, 2)
checkMath("modf", 1, ret_type="Tuple[float, float]")
checkMath("frexp", 1, ret_type="Tuple[float, int]")
checkMath("isnan", 1, ret_type="bool")
checkMath("isinf", 1, ret_type="bool")
checkMath("ldexp", 2, is_float=False, ret_type="float", args_type="(float, int)",
vals=[(i, j) for i in float_vals for j in range(-10, 10)])
checkMath("pow", 2, is_float=False, ret_type="float")
checkMath("pow", 2, is_float=True, ret_type="float")
checkMathWrap("floor", ret_type="int")
checkMathWrap("ceil", ret_type="int")
checkMathWrap("gcd", 2, is_float=False, ret_type="int")
checkMath("isfinite", 1, ret_type="bool")
checkMathWrap("remainder", 2)
checkMathWrap("factorial", 1, is_float=False, ret_type="int", vals=[(i, 0) for i in range(-2, 10)])
@skipIfTorchDynamo("TorchDynamo fails with unknown reason")
def test_if_nest_while(self):
def func(a, b):
# type: (int, int) -> int
c = 0
if a > b:
while a > b:
b = b + 1
c = -b
return c
inputs = self._make_scalar_vars([4321, 1234], torch.int64)
self.checkScript(func, inputs)
def test_script_optional_none(self):
def none_stmt(x):
output = None
output = x
return output
def none_args(x):
# type: (Optional[Tensor]) -> Optional[Tensor]
return None
self.checkScript(none_stmt, [torch.arange(0, 2)], optimize=True)
self.checkScript(none_args, [None], optimize=True)
# test undefined tensor None as default param
def test_script_optional_tensor_none(x=None):
# type: (Optional[Tensor]) -> Tensor
res = torch.zeros(1, dtype=torch.int8)
if x is None:
res = res + 1
else:
res = x
return res
fn = test_script_optional_tensor_none
scripted_fn = torch.jit.script(fn)
self.assertEqual(fn(), scripted_fn())
self.assertEqual(fn(torch.zeros(1)), scripted_fn(torch.zeros(1)))
# test typical None as default param
def test_script_optional_other_none(x=None):
# type: (Optional[float]) -> float
res = 2.0
if x is None:
res = res + 1.0
else:
res = x
return res
fn = test_script_optional_other_none
scripted_fn = torch.jit.script(fn)
self.assertEqual(fn(), scripted_fn())
self.assertEqual(fn(1.0), scripted_fn(1.0))
def test_script_clamp_none(self):
def test_script_clamp_max_none(x):
return torch.clamp(x, min=2, max=None)
def test_script_clamp_max(x):
return torch.clamp(x, max=2)
def test_script_clamp_min_none(x):
return torch.clamp(x, min=None, max=2)
def test_script_clamp_min(x):
return torch.clamp(x, min=2)
input = [torch.arange(0, 3)]
self.checkScript(test_script_clamp_max_none, input, optimize=True)
self.checkScript(test_script_clamp_max, input, optimize=True)
self.checkScript(test_script_clamp_min_none, input, optimize=True)
self.checkScript(test_script_clamp_min, input, optimize=True)
def test_script_bool_constant(self):
def test_script_bool_constant():
a = True
return a
self.checkScript(test_script_bool_constant, [])
def test_ternary(self):
def func(a, b):
c = 3
c = a + b if bool(a > 3) else b
return c
inputs_true = self._make_scalar_vars([5, 2], torch.int64)
inputs_false = self._make_scalar_vars([1, 0], torch.int64)
self.checkScript(func, inputs_true, optimize=True)
self.checkScript(func, inputs_false, optimize=True)
def test_ternary_module_type_hint(self):
class M1(torch.nn.Module):
def forward(self) -> Any:
return 'out' if self.training else {}
class M2(torch.nn.Module):
def forward(self) -> Any:
out: Any = 'out' if self.training else {}
return out
class M3(torch.nn.Module):
def forward(self) -> Optional[int]:
return None if self.training else 1
for module in [M1, M2, M3]:
self.checkModule(module().train(), ())
self.checkModule(module().eval(), ())
def test_ternary_static_if(self):
# Test for True branch when condition variable
# is annotated as Final
class M1(torch.nn.Module):
flag: torch.jit.Final[bool]
def __init__(self):
super().__init__()
self.flag = True
def forward(self) -> torch.Tensor:
return torch.ones(3) if self.flag else {}
# Test for True branch when condition variable
# is annotated as Final
class M2(torch.nn.Module):
flag: torch.jit.Final[bool]
def __init__(self):
super().__init__()
self.flag = False
def forward(self) -> torch.Tensor:
return {} if self.flag else torch.ones(3)
model1 = M1()
model2 = M2()
script_model_1 = torch.jit.script(model1)
script_model_2 = torch.jit.script(model2)
self.assertEqual(model1.forward(), script_model_1.forward())
self.assertEqual(model2.forward(), script_model_2.forward())
def test_ternary_right_associative(self):
def plus_123(x: int):
return x + 1 if x == 1 else x + 2 if x == 2 else x + 3
self.checkScript(plus_123, (1,))
self.checkScript(plus_123, (2,))
self.checkScript(plus_123, (3,))
def test_print(self):
def func(x, y):
q = (x + y).sigmoid()
print(q, 1, 2, [1, 2], [1.0, 2.0])
w = -q
return w * w
x = torch.arange(4., requires_grad=True)
y = torch.arange(0., 8, 2, requires_grad=True)
self.checkScript(func, [x, y], optimize=True, capture_output=True)
def test_format(self):
def func(x):
print("{}, I'm a {}".format("Hello", "test"))
print("format blank".format())
print("stuff before {}".format("hi"))
print("{} stuff after".format("hi"))
return x + 1
x = torch.arange(4., requires_grad=True)
self.checkScript(func, [x], optimize=True, capture_output=True)
def test_logical_short_circuit(self):
@torch.jit.script
def testNoThrows(t):
c1 = 1
if (False and bool(t[1])) or (True or bool(t[1])):
c1 = 0
return c1
FileCheck().check_not("prim::If").run(testNoThrows.graph)
self.assertEqual(0, testNoThrows(torch.randn(0)))
self.assertEqual(0, testNoThrows(torch.randn([2, 3])))
@torch.jit.script
def throwsOr(t):
c0 = False or bool(t[1])
print(c0)
@torch.jit.script
def throwsAnd(t):
c0 = True and bool(t[1])
print(c0)
t = torch.randn(0)
with self.assertRaisesRegex(RuntimeError, "index 1 out of range for tensor of size"):
throwsOr(t)
with self.assertRaisesRegex(RuntimeError, "index 1 out of range for tensor of size"):
throwsAnd(t)
def test_type_cast(self):
template = dedent('''
def func(v):
# type: ({from_type}) -> {to_type}
return {to_type}(v)
''')
def check_cast(from_type, to_type, value, raises=False):
code = template.format(from_type=from_type, to_type=to_type)
self.checkScript(code, (value,))
check_cast('int', 'float', 1)
check_cast('int', 'bool', 1)
check_cast('int', 'bool', 0)
check_cast('float', 'int', 1.)
check_cast('float', 'bool', 1.)
check_cast('float', 'bool', 0.)
check_cast('bool', 'int', True)
check_cast('bool', 'float', True)
def test_multiple_assignment(self):
def outer_func(x):
return x * 2, x + 2
@torch.jit.script
def func(x):
y, z = outer_func(x)
return y + z
x = torch.arange(4)
self.assertEqual(func(x), x * 2 + x + 2)
def test_literals(self):
def func(a):
return a.view(size=[1, 2, 3])
a = torch.randn(6)
self.checkScript(func, [a], optimize=True)
def test_return(self):
def no_return(a):
a + 1
def void_return(a):
return
def one_return(a):
return a + 1.
def multiple_returns(a):
return a * 1., a * 2., a * 3.
a = torch.randn(1, dtype=torch.float)
self.checkScript(no_return, [a], optimize=True)
self.checkScript(void_return, [a], optimize=True)
self.checkScript(one_return, [a], optimize=True)
self.checkScript(multiple_returns, [a], optimize=True)
with self.assertRaisesRegex(RuntimeError, "does not return along all paths"):
torch.jit.CompilationUnit('''
def no_return_bad_annotation(a):
# type: (Tensor) -> Tensor
a + 1
''')
def test_error(self):
@torch.jit.script
def foo(a):
return a.t()
s = Variable(torch.rand(5, 5, 5))
# XXX: this should stay quiet in stay propagation and only fail in the interpreter
with self.assertRaisesRegex(RuntimeError, "failed in the TorchScript interpreter"):
foo(s)
@torch.jit.script
def bar(c, b):
return c + b
with self.assertRaisesRegex(RuntimeError, "failed in the TorchScript interpreter"):
bar(Variable(torch.rand(10), requires_grad=True), Variable(torch.rand(9), requires_grad=True))
def test_error_stacktrace(self):
@torch.jit.script
def baz(c, b):
return c + b
@torch.jit.script
def foo(c, b):
return baz(c, b)
@torch.jit.script
def bar(c, b):
return foo(c, b)
with self.assertRaises(RuntimeError) as cm:
bar(torch.rand(10), torch.rand(9))
FileCheck().check("The following operation failed in the TorchScript interpreter") \
.check("Traceback") \
.check("in foo").check("in baz").run(str(cm.exception))
def test_error_stacktrace_interface(self):
@torch.jit.script
def baz(c, b):
return c + b
@torch.jit.script
def foo(c, b):
return baz(c, b)
@torch.jit.script
def bar(c, b):
return foo(c, b)
@torch.jit.script
class Bar(object):
def one(self, x, y):
return bar(x, y)
@torch.jit.interface
class IFace(object):
def one(self, x, y):
# type: (Tensor, Tensor) -> Tensor
pass
make_global(IFace)
@torch.jit.script
def as_interface(x):
# type: (IFace) -> IFace
return x
f = as_interface(Bar())
with self.assertRaises(RuntimeError) as cm:
x = f.one(torch.rand(10), torch.rand(9))
bar(torch.rand(10), torch.rand(9))
FileCheck().check("The following operation failed in the TorchScript interpreter") \
.check("Traceback") \
.check("in foo").check("in baz").run(str(cm.exception))
def test_operator_precedence(self):
def double(x):
# type: (int) -> int
return 2 * x
def complicated_arithmetic_operation():
# TODO we need to test exponent operator '**' and bitwise not
# operator '~' once they are properly supported.
list = [0, 1, 2, 3]
result = list[1:3][0] + double(4) + (-3 + 8) * 6 // 2 % 4 << 2 + 1 >> 1 | 23 & 16 + 3 ^ 4
return result
self.checkScript(complicated_arithmetic_operation, ())
def test_in_operator_with_two_strings(self):
def fn() -> bool:
return "a" in "abcd"
self.checkScript(fn, ())
def test_bitwise_ops(self):
def int_test():
return 2 & 3, 2 ^ 3, 2 | 3, 2 << 3, 2 >> 3
self.checkScript(int_test, ())
def bool_test(x, y):
# type: (bool, bool) -> Tuple[bool, bool, bool]
return x & y, x ^ y, x | y
self.checkScript(bool_test, (True, False))
self.checkScript(bool_test, (True, True))
def tensor_test(x, y):
return x & y, x ^ y, x | y
def tensor_with_int_test(x, y):
# type: (Tensor, int) -> Tuple[Tensor, Tensor]
return x << y, x >> y
x = torch.tensor(2)
y = torch.tensor(3)
self.checkScript(tensor_test, (x, y))
self.checkScript(tensor_with_int_test, (x, 2))
def not_test(x):
return ~x
self.checkScript(not_test, (torch.tensor([2, 4]), ))
def test_all(self):
@torch.jit.script
def test_all_tensor(x):
return all(x)
self.assertFalse(test_all_tensor(torch.tensor([1, 0, 3], dtype=torch.uint8)))
self.assertTrue(test_all_tensor(torch.tensor([3.14, 3, 99], dtype=torch.uint8)))
self.assertTrue(test_all_tensor(torch.tensor([True, True], dtype=torch.uint8)))
self.assertFalse(test_all_tensor(torch.tensor([True, False], dtype=torch.uint8)))
@torch.jit.script
def test_all_bool_list(x):
# type: (List[bool]) -> bool
return all(x)
self.assertTrue(test_all_bool_list([True, True]))
self.assertTrue(test_all_bool_list([True, 1]))
self.assertFalse(test_all_bool_list([True, False]))
self.assertFalse(test_all_bool_list([True, 0]))
self.assertFalse(test_all_bool_list([False, 0]))
self.assertTrue(test_all_bool_list([]))
@torch.jit.script
def test_all_int_list(x):
# type: (List[int]) -> bool
return all(x)
self.assertTrue(test_all_int_list([3, 6]))
self.assertFalse(test_all_int_list([2, 0]))
@torch.jit.script
def test_all_float_list(x):
# type: (List[float]) -> bool
return all(x)
self.assertTrue(test_all_float_list([3.14, 8.1]))
self.assertFalse(test_all_float_list([3.14, 0, 8.9]))
def test_number_math(self):
ops_template = dedent('''
def func():
return {scalar1} {op} {scalar2}
''')
ops = ['+', '-', '*', '%', '<', '<=', '>', '>=', '==', '!=', '//']
funcs_template = dedent('''
def func():
return {func}({scalar1}, {scalar2})
''')
funcs = ['min', 'max']
scalars = ['7', '2', '3', '-3', '3.14', '0.125', '-0.5', '2.0', '-2.0']
scalar_pairs = [(scalar1, scalar2) for scalar1 in scalars for scalar2 in scalars]
def run_test(code):
scope = {}
execWrapper(code, globals(), scope)
cu = torch.jit.CompilationUnit(code)
self.assertEqual(cu.func(), scope['func']())
for scalar1, scalar2 in scalar_pairs:
for op in ops:
code = ops_template.format(op=op, scalar1=scalar1, scalar2=scalar2)
run_test(code)
for func in funcs:
code = funcs_template.format(func=func, scalar1=scalar1, scalar2=scalar2)
run_test(code)
# test Scalar overloads
for scalar1, scalar2 in scalar_pairs:
item1 = 'torch.tensor(' + scalar1 + ').item()'
item2 = 'torch.tensor(' + scalar2 + ').item()'
for op in ops:
code = ops_template.format(op=op, scalar1=item1, scalar2=scalar2)
run_test(code)
code = ops_template.format(op=op, scalar1=scalar1, scalar2=item2)
run_test(code)
code = ops_template.format(op=op, scalar1=item1, scalar2=item2)
run_test(code)
for func in funcs:
code = funcs_template.format(func=func, scalar1=item1, scalar2=scalar2)
run_test(code)
code = funcs_template.format(func=func, scalar1=scalar1, scalar2=item2)
run_test(code)
code = funcs_template.format(func=func, scalar1=item1, scalar2=item2)
run_test(code)
def test_number_abs(self):
def func1(x):
# type: (float) -> float
return abs(x)
def func2(x):
# type: (int) -> int
return abs(x)
def func3(x):
return abs(x)
self.checkScript(func1, (-3.14,))
self.checkScript(func1, (3.14,))
self.checkScript(func2, (-10,))
self.checkScript(func2, (10,))
self.checkScript(func3, (torch.tensor([-5, -10, -20]),))
self.checkScript(func3, (torch.tensor([5, 10, 20]),))
self.checkScript(func3, (torch.tensor([-5, 10, -20]),))
def test_number_div(self):
self.assertEqual(div_int_future(), torch.jit.script(div_int_future)())
self.checkScript(div_float_future, ())
self.checkScript(div_int_nofuture, ())
self.checkScript(div_float_nofuture, ())
# Testing bitwise shorthand aug assignment
def test_bool_augassign_bitwise_or(self):
def func(a: bool, b: bool) -> bool:
a |= b
return a
self.checkScript(func, (True, False), optimize=True)
self.checkScript(func, (True, True), optimize=True)
self.checkScript(func, (False, False), optimize=True)
self.checkScript(func, (False, True), optimize=True)
def test_bool_augassign_bitwise_and(self):
def func(a: bool, b: bool) -> bool:
a &= b
return a
self.checkScript(func, (True, False), optimize=True)
self.checkScript(func, (True, True), optimize=True)
self.checkScript(func, (False, False), optimize=True)
self.checkScript(func, (False, True), optimize=True)
def test_bool_augassign_bitwise_xor(self):
def func(a: bool, b: bool) -> bool:
a ^= b
return a
self.checkScript(func, (True, False), optimize=True)
self.checkScript(func, (True, True), optimize=True)
self.checkScript(func, (False, False), optimize=True)
self.checkScript(func, (False, True), optimize=True)
def test_number_augassign_bitwise_lshift(self):
def func() -> int:
z = 8
z <<= 2
return z
self.checkScript(func, (), optimize=True)
def test_number_augassign_bitwise_rshift(self):
def func() -> int:
z = 8
z >>= 2
return z
self.checkScript(func, (), optimize=True)
def test_number_augassign_bitwise_pow(self):
def func() -> float:
z = 8
z **= 2
return z
self.checkScript(func, (), optimize=True)
def test_number_augassign(self):
def func():
z = 1
z += 2
return z
self.checkScript(func, (), optimize=True)
def test_nested_select_assign(self):
class SubSubModule(torch.nn.Module):
def __init__(self):
super(SubSubModule, self).__init__()
self.abc = 11
def forward(self, x):
return self.abc
class SubModule(torch.nn.Module):
def __init__(self):
super(SubModule, self).__init__()
self.a = 11
self.nested = SubSubModule()
def forward(self, x):
return self.a
class TestModule(torch.nn.Module):
def __init__(self):
super(TestModule, self).__init__()
self.sub = SubModule()
self.hi = 1
def forward(self):
self.hi = 5
self.sub.a = 1
self.sub.nested.abc = 5
return self.sub.a * 20 + self.sub.nested.abc * 3 + self.hi
self.checkModule(TestModule(), ())
def test_number_neg(self):
# int -> int
def func1():
return -8
# float -> float
def func2():
return -3.14
self.checkScript(func1, (), optimize=True)
self.checkScript(func2, (), optimize=True)
def test_compare_two_bool_inputs(self):
def compare_eq(a: bool, b: bool):
return a == b
def compare_ne(a: bool, b: bool):
return a != b
scripted_fn_eq = torch.jit.script(compare_eq)
scripted_fn_ne = torch.jit.script(compare_ne)
self.assertEqual(scripted_fn_eq(True, False), compare_eq(True, False))
self.assertEqual(scripted_fn_eq(False, True), compare_eq(False, True))
self.assertEqual(scripted_fn_eq(True, True), compare_eq(True, True))
self.assertEqual(scripted_fn_eq(False, False), compare_eq(False, False))
self.assertEqual(scripted_fn_ne(True, False), compare_ne(True, False))
self.assertEqual(scripted_fn_ne(False, True), compare_ne(False, True))
self.assertEqual(scripted_fn_ne(True, True), compare_ne(True, True))
self.assertEqual(scripted_fn_ne(False, False), compare_ne(False, False))
def _test_tensor_number_math(self, device='cpu'):
template = dedent('''
def func(t):
return {lhs} {op} {rhs}
''')
def test(op, tensor, const, swap_args, template=template):
args = ('t', const)
if swap_args:
args = (const, 't')
code = template.format(lhs=args[0], rhs=args[1], op=op)
scope = {}
execWrapper(code, globals(), scope)
cu = torch.jit.CompilationUnit(code)
message = 'with code `{} {} {}` and t={}'.format(args[0], op, args[1], tensor)
res1 = cu.func(tensor)
res2 = scope['func'](tensor)
self.assertEqual(res1, res2, msg=message + "\nres1=" + str(res1) + "\nres2=" + str(res2))
self.assertEqual(res1.dtype, res2.dtype, msg=message + "\nres1=" + str(res1) + "\nres2=" + str(res2))
var_int = [2, -2]
var_float = [1.4321, -1.2]
ops = ['+', '-', '*', '%', '<', '<=', '>', '>=', '==', '!=', '/']
float_tensor = torch.randn(5, 5, device=device)
double_tensor = torch.randn(5, 5, dtype=torch.double, device=device)
long_tensor = torch.randint(-5, 5, (5, 5), dtype=torch.long, device=device)
long_tensor[long_tensor == 0] = 2
tensors = [float_tensor, double_tensor, long_tensor]
consts = var_int + var_float
for op, tensor, const, swap_args in product(ops, tensors, consts, [True, False]):
# FIXME: things like 2 / long_tensor are not implemented correctly
# Look in torch/_tensor.py to see how pytorch implements it.
if op == '/' and tensor.data_ptr() == long_tensor.data_ptr():
continue
# % operator does not take: const % tensor
if op == '%' and swap_args is True:
continue
test(op, tensor, const, swap_args)
def test_tensor_number_math(self):
self._test_tensor_number_math()
def test_torch_tensor_bad_input(self):
with self.assertRaisesRegex(RuntimeError, "must be of ints, floats, "
"or bools, got None"):
@torch.jit.script
def test():
return torch.tensor([None])
test()
with self.assertRaisesRegex(RuntimeError, r"Empty lists default to List\[Tensor\]"):
@torch.jit.script
def tmp():
return torch.tensor([])
tmp()
@torch.jit.script
def foo():
return torch.tensor([[2, 2], [1]])
with self.assertRaisesRegex(RuntimeError, "Expected sequence of length"):
foo()
@suppress_warnings
def test_torch_tensor_as_tensor_empty_list(self):
tensor_template = dedent('''
def func():
empty_list = torch.jit.annotate(List[int], [])
ten1 = torch.{tensor_op}({input})
return ten1
''')
ops = ['tensor', 'as_tensor']
inputs = ['empty_list', '[empty_list, empty_list]', '[[[empty_list]]]']
for op in ops:
for inp in inputs:
code = tensor_template.format(tensor_op=op, input=inp)
scope = {}
exec(code, globals(), scope)
cu = torch.jit.CompilationUnit(code)
t1 = cu.func()
t2 = scope['func']()
if inp == 'empty_list':
# torchscript returns int tensor, python returns float tensor
self.assertNotEqual(t1.dtype, t2.dtype)
self.assertEqual(t1, t2, exact_dtype=False)
self.assertEqual(t1.device, t2.device)
@unittest.skipIf(GRAPH_EXECUTOR != ProfilingMode.LEGACY, "Simple Executor doesn't have any shapes to propagate")
def test_tensor_as_tensor_shape_prop(self):
tensor_template = dedent('''
def func():
return torch.{tensor_op}({input})
''')
ops = ['tensor', 'as_tensor']
inputs = ['[1]', '[False]', '[2.5]', '0.5', '1', 'False', '[[1]]', 'torch.jit.annotate(List[List[int]], [])']
expected_shape = ["Long(*, device=cpu)", "Bool(*, device=cpu)",
"Double(*, device=cpu)", "Double(device=cpu)",
"Long(device=cpu)", "Bool(device=cpu)", "Long(*, *, device=cpu)"]
for op in ops:
for inp, expect in zip(inputs, expected_shape):
code = tensor_template.format(tensor_op=op, input=inp)
scope = {}
exec(code, globals(), scope)
cu = torch.jit.CompilationUnit(code)
torch._C._jit_pass_complete_shape_analysis(cu.func.graph, (), False)
FileCheck().check(expect).check("aten::{tensor_op}".format(tensor_op=op)).run(cu.func.graph)
@torch.jit.script
def test_dtype(inp_dtype: torch.dtype):
a = torch.tensor(1.0, dtype=torch.float, requires_grad=True)
return a, torch.tensor(1.0, dtype=inp_dtype)
if GRAPH_EXECUTOR == ProfilingMode.PROFILING:
g = test_dtype.graph_for(5, profile_and_replay=True)
# both should have completed shapes
FileCheck().check("Tensor = aten::tensor").check("Float(device=cpu) = prim::BailOut") \
.check("Tensor = aten::tensor").check("Half(device=cpu) = prim::BailOut").run(g)
else:
g = test_dtype.graph_for(5)
# first should have type set second should not
FileCheck().check("Float(requires_grad=1, device=cpu) = aten::tensor") \
.check("Tensor(requires_grad=0) = aten::tensor").run(g)
@torch.jit.script
def test_as_tensor_tensor_input(input):
a = torch.as_tensor(input, dtype=input.dtype)
return a, torch.as_tensor(input, dtype=torch.float)
if GRAPH_EXECUTOR == ProfilingMode.PROFILING:
g = test_as_tensor_tensor_input.graph_for(torch.ones(3, 4), profile_and_replay=True)
FileCheck().check("Tensor = aten::as_tensor").check("Float(3, 4) = prim::BailOut") \
.check("Tensor = aten::as_tensor").check("Float(3, 4) = prim::BailOut").run(g)
else:
g = test_as_tensor_tensor_input.graph_for(torch.ones(3, 4))
FileCheck().check("Tensor = aten::as_tensor").check("Float(*, *, requires_grad=0, device=cpu) = aten::as_tensor").run(g)
@unittest.skipIf(GRAPH_EXECUTOR != ProfilingMode.LEGACY, "testing legacy behavior")
def test_tensor_requires_grad(self):
@torch.jit.script
def test(b):
# type: (bool) -> Tuple[Tensor, Tensor, Tensor]
a = torch.tensor(1., requires_grad=b)
b = torch.tensor(1., requires_grad=True)
c = torch.tensor(1., requires_grad=False)
return a, b, c
g = test.graph_for(True)
out = next(g.outputs())
out_inp = list(out.node().inputs())
self.assertTrue(out_inp[0].requires_grad())
self.assertTrue(out_inp[1].requires_grad())
self.assertFalse(out_inp[2].requires_grad())
def test_grad_from_script(self):
def test():
a = torch.tensor(2.5, requires_grad=True)
b = a * 2
return a, b
a, b = test()
b.backward()
a_script, b_script = torch.jit.script(test)()
b_script.backward()
self.assertEqual(a.grad, a_script.grad)
def test_torch_tensor_as_tensor(self):
tensor_template = dedent('''
def func():
li = {list_create}
ten1 = torch.{tensor_op}(li {options})
return ten1
''')
lists = ["2.5", "4", "True", "False", "[2]", "[-.5]", "[False, True, False]", "[2, 2]", "(1, 1)",
"torch.jit.annotate(List[List[int]], [])",
"torch.jit.annotate(List[int], [])", "[2.5, 2.5]", "[[2], [2]]", "[[-.5], [2.2]]", "[[False], [True]]"]
dtypes = ["", ", dtype=torch.float", ", dtype=torch.double", ", dtype=torch.half",
", dtype=torch.uint8", ", dtype=torch.int8", ", dtype=torch.short",
", dtype=torch.int", ", dtype=torch.long", ", dtype=torch.cfloat",
", dtype=torch.cdouble"]
ops = ['tensor', 'as_tensor']
devices = ['', ", device='cpu'"]
if RUN_CUDA:
devices.append(", device='cuda'")
option_pairs = [dtype + device for dtype in dtypes for device in devices]
for op in ops:
for li in lists:
for option in option_pairs:
# tensor from empty list is type float in python and annotated type in torchscript
if "annotate" in li and "dtype" not in option:
continue
# Skip unsigned tensor initializaton for signed values on 3.10
if sys.version_info[:2] >= (3, 10) and "torch.uint8" in option and "-" in li:
continue
code = tensor_template.format(list_create=li, tensor_op=op, options=option)
scope = {}
exec(code, globals(), scope)
cu = torch.jit.CompilationUnit(code)
t1 = cu.func()
t2 = scope['func']()
if t1.dtype == torch.float16: # equality NYI for half tensor
self.assertTrue(str(t1) == str(t2))
else:
self.assertEqual(t1, t2)
self.assertEqual(t1.dtype, t2.dtype)
self.assertEqual(t1.device, t2.device)
def test_as_tensor_tensor_input(input):
# type: (Tensor) -> Tuple[Tensor, Tensor, Tensor]
return torch.as_tensor(input, dtype=torch.cfloat), torch.as_tensor(input, dtype=torch.float), \
torch.as_tensor(input, dtype=torch.int32)
inp = torch.randn(3, 4, dtype=torch.cfloat)
self.checkScript(test_as_tensor_tensor_input, (inp,))
def test_torch_tensor_dtype(self):
def foo(s: float):
return torch.tensor(s), torch.tensor([s, s])
# need to clear function cache so we re run shape analysis
with set_default_dtype(torch.double):
self.assertEqual(torch.jit.script(foo)(1.), foo(1.), exact_dtype=True)
if GRAPH_EXECUTOR == ProfilingMode.LEGACY:
FileCheck().check("Double").check_same("aten::tensor").run(torch.jit.last_executed_optimized_graph())
with set_default_dtype(torch.float):
del torch.jit._state._jit_caching_layer[foo]
self.assertEqual(torch.jit.script(foo)(1.), foo(1.), exact_dtype=True)
if GRAPH_EXECUTOR == ProfilingMode.LEGACY:
FileCheck().check("Float").check_same("aten::tensor").run(torch.jit.last_executed_optimized_graph())
with set_default_dtype(torch.half):
del torch.jit._state._jit_caching_layer[foo]
self.assertEqual(torch.jit.script(foo)(1.), foo(1.), exact_dtype=True)
if GRAPH_EXECUTOR == ProfilingMode.LEGACY:
FileCheck().check("Half").check_same("aten::tensor").run(torch.jit.last_executed_optimized_graph())
def test_shape_analysis_grad_property(self):
@torch.jit.script
def foo(x):
return torch.sub(x, torch.tanh(x))
torch._C._jit_pass_complete_shape_analysis(foo.graph, (torch.tensor([0.39]),), False)
# requires_grad property shouldn't be accidentally set by shape analysis
self.assertTrue(foo.graph.findNode("aten::sub").output().requiresGrad() is None)
def test_empty_like_memory_format_bc(self):
def f(x):
# type: (Tensor) -> Tensor
return torch.zeros_like(x, memory_format=None)
scripted_f = torch.jit.script(f)
x = torch.rand(3, 4)
self.assertEqual(scripted_f(x), f(x))
def test_multiline_string_dedents(self):
def foo() -> None:
multiline_string_dedent_1 = """
This is a string dedent """
multiline_string_dedent_2 = """ This is a
string dedent """
multiline_string_dedent_3 = """
This is a string
dedent """
multiline_string_dedent_4 = """ This is a string dedent """
scripted_foo = torch.jit.script(foo)
self.assertEqual(scripted_foo(), foo())
def test_class_with_comment_at_lower_indentation(self):
class Foo(torch.nn.Module):
def forward(self, x):
x = torch.neg(x)
# This comment is at the wrong indent
return x
torch.jit.script(Foo())
# adapted from test in test_torch
def test_tensor_to(self):
template = dedent('''
def func(t):
cuda = "{cuda}"
device = "{device}"
non_blocking = {non_blocking}
return {to_str}
''')
def s(t, to_str, non_blocking=None, device=None, cuda=None):
device = device if device is not None else str(t.device)
non_blocking = non_blocking if non_blocking is not None else False
cuda = "cuda" if cuda is None else cuda
code = template.format(to_str=to_str, device=device, non_blocking=non_blocking, cuda=cuda)
scope = {}
cu = torch.jit.CompilationUnit(code)
return cu.func(t, profile_and_replay=True)
def test_copy_behavior(t, non_blocking=False):
self.assertIs(t, s(t, 't.to(t, non_blocking=non_blocking)', non_blocking))
self.assertIs(t, s(t, 't.to(t.dtype, non_blocking=non_blocking)', non_blocking))
self.assertIs(t, s(t, 't.to(torch.empty_like(t), non_blocking=non_blocking)', non_blocking))
self.assertIsNot(t, s(t, 't.to(t, non_blocking=non_blocking, copy=True)', non_blocking))
self.assertIsNot(t, s(t, 't.to(t.dtype, non_blocking=non_blocking, copy=True)', non_blocking))
self.assertIsNot(t, s(t, 't.to(torch.empty_like(t), non_blocking=non_blocking, copy=True)', non_blocking))
devices = [t.device]
if t.device.type == 'cuda':
if t.device.index == -1:
devices.append('cuda:{}'.format(torch.cuda.current_device()))
elif t.device.index == torch.cuda.current_device():
devices.append('cuda')
for device in devices:
self.assertIs(t, s(t, 't.to(device, non_blocking=non_blocking)', non_blocking, device))
self.assertIs(t, s(t, 't.to(device, t.dtype, non_blocking=non_blocking)', non_blocking, device))
self.assertIsNot(t, s(t, 't.to(device, non_blocking=non_blocking, copy=True)', non_blocking, device))
self.assertIsNot(t, s(t, 't.to(device, t.dtype, non_blocking=non_blocking, copy=True)',
non_blocking, device))
t = torch.tensor(5)
test_copy_behavior(t)
self.assertEqual(t.device, s(t, "t.to('cpu')").device)
self.assertEqual(t.device, s(t, "t.to('cpu', dtype=torch.float32)").device)
self.assertIs(torch.float32, s(t, "t.to('cpu', dtype=torch.float32)").dtype)
self.assertEqual(t.device, s(t, "t.to(torch.float32)").device)
self.assertIs(torch.float32, s(t, "t.to(dtype=torch.float32)").dtype)
self.assertEqual(t.data_ptr(), s(t, "t.to('cpu')").data_ptr())
self.assertEqual(t.data_ptr(), s(t, "t.to(dtype=t.dtype, device=t.device, copy=False)").data_ptr())
self.assertEqual(t.data_ptr(), s(t, "t.to('cpu', copy=False)").data_ptr())
self.assertNotEqual(t.data_ptr(), s(t, "t.to('cpu', copy=True)").data_ptr())
a = torch.tensor(5)
if torch.cuda.is_available():
for non_blocking in [True, False]:
for cuda in ['cuda', 'cuda:0' if torch.cuda.device_count() == 1 else 'cuda:1']:
b = torch.tensor(5., device=cuda)
test_copy_behavior(b, non_blocking)
self.assertEqual(b.device, s(b, "t.to(cuda, non_blocking=non_blocking).device", cuda=cuda))
self.assertEqual(a.device, s(b, "t.to('cpu', non_blocking=non_blocking).device"))
self.assertEqual(b.device, s(b, "t.to(cuda, non_blocking=non_blocking).device", cuda=cuda))
self.assertIs(torch.int32, s(b, "t.to('cpu', dtype=torch.int32, non_blocking=non_blocking)").dtype)
self.assertEqual(a.device, s(b, "t.to('cpu', dtype=torch.int32, non_blocking=non_blocking)").device)
self.assertIs(torch.int32, s(b, "t.to(dtype=torch.int32)").dtype)
self.assertEqual(b.device, s(b, "t.to(dtype=torch.int32)").device)
# Test AD: aten::to(Tensor self, int dtype, bool non_blocking, bool copy) -> Tensor
t = torch.tensor(5).float().requires_grad_()
out_ref = t.to(torch.float32)
out = s(t, "t.to(torch.float32)")
self.assertEqual(out_ref, out)
grad_ref = torch.autograd.grad(out_ref.sum(), t)
grad = torch.autograd.grad(out.sum(), t)
self.assertEqual(grad_ref, grad)
# Test AD: aten::to(Tensor self, Device? device, int? dtype, bool non_blocking, bool copy) -> Tensor
out_ref = t.to('cpu')
out = s(t, "t.to('cpu')")
self.assertEqual(out_ref, out)
grad_ref = torch.autograd.grad(out_ref.sum(), t)
grad = torch.autograd.grad(out.sum(), t)
self.assertEqual(grad_ref, grad)
# Test AD: aten::to(Tensor self, Tensor other, bool non_blocking, bool copy) -> Tensor
@torch.jit.script
def func2(t, t_ref):
return t.to(t_ref)
with disable_autodiff_subgraph_inlining():
t_ref = torch.tensor(4).double()
out_ref = t.to(t_ref)
out = func2(t, t_ref)
grad_ref = torch.autograd.grad(out_ref.sum(), t)
grad = torch.autograd.grad(out.sum(), t)
self.assertEqual(grad_ref, grad)
@unittest.skipIf(not RUN_CUDA, "No CUDA")
def test_tensor_number_math_cuda(self):
self._test_tensor_number_math(device='cuda')
def test_not(self):
# test not operator in python
# TODO: add more tests when bool conversions ready
def test_not_op(a):
return not bool(a > 1)
self.checkScript(test_not_op, (torch.tensor(2), ), optimize=True)
def test_is_isnot(self):
# test is and is not operator in python
template = dedent('''
def func():
# type: () -> bool
return {lhs} {op} {rhs}
''')
def test(op, args):
code = template.format(lhs=args[0], rhs=args[1], op=op)
scope = {}
execWrapper(code, globals(), scope)
cu = torch.jit.CompilationUnit(code)
self.assertEqual(
cu.func(),
scope['func'](),
msg="Failed with op: {}, lhs: {}, rhs: {}"
.format(op, args[0], args[1])
)
ops = ['is', 'is not']
type_literals = [True, False, None, [1, 1], 1, 2, .5, 1.5]
# do literals product to try any types combinations
for op, lhs, rhs in product(ops, type_literals, type_literals):
test(op, [lhs, rhs])
def test_isinstance_refinement(self):
@torch.jit.script
def foo(a):
# type: (Optional[int]) -> int
if isinstance(a, int):
return a + 3
else:
return 4
self.assertEqual(foo(4), 7)
self.assertEqual(foo(None), 4)
@torch.jit.script
def foo2(a, b):
# type: (Optional[int], Optional[int]) -> int
if not isinstance(a, int) or not isinstance(b, int):
return 0
else:
return a + b
self.assertEqual(foo2(3, 4), 7)
self.assertEqual(foo2(None, 4), 0)
self.assertEqual(foo2(4, None), 0)
@torch.jit.script
def any_refinement(a, b):
# type: (Any, Any) -> int
if isinstance(a, int) and isinstance(b, int):
return a + b
return 0
self.assertEqual(any_refinement(3, 4), 7)
self.assertEqual(any_refinement(3, "hi"), 0)
@torch.jit.script
def any_refinement2(a):
# type: (Any) -> Tensor
if isinstance(a, Tensor):
return a
return torch.tensor(3)
self.assertEqual(any_refinement2(3), torch.tensor(3))
self.assertEqual(any_refinement2(torch.tensor(5)), torch.tensor(5))
@unittest.skipIf(GRAPH_EXECUTOR == ProfilingMode.LEGACY, "bug persists in deprecated executor")
def test_unspecialized_any_binding(self):
# any binding will infer the type, if it infers
# a specialized tensor type `x` Dict type will fail isinstance check
@torch.jit.script
def foo(x: Any):
assert isinstance(x, Dict[str, torch.Tensor])
foo({"1": torch.tensor(3)})
with self.assertRaises(Exception):
foo(2)
def test_isinstance(self):
# test isinstance operator for static type checking
template = dedent('''
def func(x):
# type: ({type_hint}) -> bool
return isinstance(x, {typ})
''')
def test(inp, typ, type_hint):
code = template.format(typ=typ, type_hint=type_hint)
scope = {}
execWrapper(code, globals(), scope)
cu = torch.jit.CompilationUnit(code)
self.assertEqual(
cu.func(inp),
scope['func'](inp),
msg="Failed with typ: {}"
.format(typ)
)
inputs = [True, 1, 1.0, torch.tensor(1), [1, 2], (1.0,), [1, 2], 1]
type_literals = ['bool', 'int', 'float', 'torch.Tensor', 'list', 'tuple',
'(list, tuple)', '(int, float, bool)']
type_annotations = ['bool', 'int', 'float', 'Tensor', 'List[int]', 'Tuple[float]',
'List[int]', 'int']
# do zipping to try different types
for inp, typ, type_hint in zip(inputs, type_literals, type_annotations):
test(inp, typ, type_hint)
# test optional isinstance check
@torch.jit.script
def opt_func(x):
# type: (Optional[int]) -> bool
return isinstance(x, int)
self.assertTrue(opt_func(3))
self.assertFalse(opt_func(None))
def test_dropout_eval(self):
class ScriptedConv2d(torch.jit.ScriptModule):
def __init__(self, in_channels, out_channels, **kwargs):
super(ScriptedConv2d, self).__init__()
self.conv = nn.Conv2d(in_channels, out_channels, bias=False, **kwargs)
self.bn = nn.BatchNorm2d(out_channels, eps=0.001)
@torch.jit.script_method
def forward(self, x):
x = self.conv(x)
x = self.bn(x)
return F.relu(x, inplace=True)
class ScriptMod(torch.jit.ScriptModule):
def __init__(self):
super(ScriptMod, self).__init__()
self.Conv2d_1a_3x3 = ScriptedConv2d(3, 32, kernel_size=3, stride=2)
@torch.jit.script_method
def forward(self, x):
x = self.Conv2d_1a_3x3(x)
return F.dropout(x, training=self.training)
class EagerConv2d(torch.nn.Module):
def __init__(self, in_channels, out_channels, **kwargs):
super(EagerConv2d, self).__init__()
self.conv = nn.Conv2d(in_channels, out_channels, bias=False, **kwargs)
self.bn = nn.BatchNorm2d(out_channels, eps=0.001)
def forward(self, x):
x = self.conv(x)
x = self.bn(x)
return F.relu(x, inplace=True)
class EagerMod(torch.nn.Module):
def __init__(self):
super(EagerMod, self).__init__()
self.Conv2d_1a_3x3 = EagerConv2d(3, 32, kernel_size=3, stride=2)
def forward(self, x):
x = self.Conv2d_1a_3x3(x)
return F.dropout(x, training=self.training)
script_input = torch.rand(4, 3, 299, 299)
eager_input = script_input.clone()
with freeze_rng_state():
script_mod = ScriptMod()
script_mod.eval()
script_output = script_mod(script_input)
with freeze_rng_state():
eager_mod = EagerMod()
eager_mod.eval()
eager_output = eager_mod(eager_input)
self.assertEqual(script_output, eager_output)
with freeze_rng_state():
script_mod = ScriptMod()
script_mod.train()
script_output = script_mod(script_input)
with freeze_rng_state():
eager_mod = EagerMod()
eager_mod.train()
eager_output = eager_mod(eager_input)
self.assertEqual(script_output, eager_output)
def test_nested_breaks(self):
def no_bool_loop_outputs(g):
# testing that the "did exit" transform values are not loop block
# outputs (and thus not affecting one loop from another)
loops = g.findAllNodes("prim::Loop")
for loop in loops:
for out in loop.outputs():
self.assertTrue(out.type() != BoolType.get())
def test(y):
# type: (int)
ret = 0
tensor = torch.tensor(0)
while int(tensor.add_(1)) < 4:
if y == 1:
continue
for i in range(y):
continue
ret += 1
ret += 1
return ret, int(tensor)
self.assertEqual(torch.jit.script(test)(1), test(1))
self.assertEqual(torch.jit.script(test)(2), test(2))
no_bool_loop_outputs(torch.jit.script(test).graph)
def foo():
y = torch.tensor(0)
z = 0
while int(y.add_(1)) < 20:
if int(y) < 10:
for i in range(6):
if i == 3:
continue
else:
if i > 3:
break
z += 2
if int(y) == 18:
break
if int(y) == 15:
continue
z += 1
return int(y), z
no_bool_loop_outputs(torch.jit.script(foo).graph)
self.checkScript(foo, ())
def test_nested_two():
i = 0
k = 0
while i < 5:
for j in range(5):
k += 1
if j == 3:
continue
i += 1
k += 1
if i == 4:
break
return i, k
self.checkScript(test_nested_two, ())
no_bool_loop_outputs(torch.jit.script(test_nested_two).graph)
def test_breaks_continues(self):
def foo_continue(cond):
# type: (int)
j = 1
for i in range(5):
if i == cond:
continue
j += 1
return j
def foo_break(cond):
# type: (int)
j = 1
for i in range(5):
if i == cond:
break
j += 1
return j
for i in range(1, 4):
self.checkScript(foo_continue, (i,))
self.checkScript(foo_break, (i,))
def test_refine_outside_loop():
if 1 == 1:
x = None
else:
x = 1
i = 0
j = 0
while (x is None or torch.jit._unwrap_optional(x) > 3):
if i < 3:
if i < 3:
x = torch.jit.annotate(Optional[int], None)
i += 1
continue
x = 1
else:
x = 1 if x is None else x
x = x + 1
j = x + x
return x, j
self.checkScript(test_refine_outside_loop, ())
def assign_after_break(y):
# type: (int)
x = 0
for i in range(y):
x = y * 2 + i
break
x = 4
return x
self.checkScript(assign_after_break, (1,))
self.checkScript(assign_after_break, (2,))
self.checkScript(assign_after_break, (3,))
def assign_after_break_nested(y):
# type: (int)
x = 0
for i in range(y):
if y == 1:
x = 5
break
assert 1 == 2
else:
x = x + 1
break
assert 1 == 2
x = -30
assert 1 == 2
return x
self.checkScript(assign_after_break_nested, (1,))
self.checkScript(assign_after_break_nested, (2,))
self.checkScript(assign_after_break_nested, (3,))
def may_break(y):
# type: (int)
x = 0
for i in range(y):
if y == 1:
x = 5
else:
x = x + 1
break
x = -30
return x
self.checkScript(may_break, (1,))
self.checkScript(may_break, (2,))
self.checkScript(may_break, (3,))
def test(x, y):
# type: (int, int)
a = 1
while (x > 0):
if y == 3:
for i in range(y):
a += (1 % (i + 1))
x -= 1
if x == 3:
a = x * 3
break
if x < 3:
if x == 1:
a -= 2
x -= 1
break
a -= 1
x -= 3
return a, x
self.checkScript(test, (10, 3))
self.checkScript(test, (10, 2))
self.checkScript(test, (3, 2))
self.checkScript(test, (5, 3))
self.checkScript(test, (2, 3))
def test_delete_after_break(x):
# type: (int)
a = 1
b = 1
for i in range(x):
a = i * 3
break
b = i * 5
return a, b
self.checkScript(test_delete_after_break, (0,))
self.checkScript(test_delete_after_break, (1,))
def test_will_break_after_guard(x):
# type: (int)
a = 1
for i in range(x):
if i == 4:
a = 3
break
a -= 1
break
assert 1 == 2
a -= -100
return a
self.checkScript(test_will_break_after_guard, (0,))
self.checkScript(test_will_break_after_guard, (2,))
self.checkScript(test_will_break_after_guard, (4,))
def test_varexit(cond):
# type: (int)
m = 0
for i in range(3):
if cond == 2:
if cond == 2:
m = 2
break
k = 1
else:
k = 2
m += k
return m
# use of k tests the pathway where we have to insert unitialized
self.checkScript(test_varexit, (3,))
self.checkScript(test_varexit, (2,))
def test_break_true():
i = 0
while True:
i += 1
if i == 3:
break
while False:
i += 1
return i
self.checkScript(test_break_true, ())
def test_break_continue_error(self):
with self.assertRaisesRegex(RuntimeError, "Syntax"):
cu = torch.jit.CompilationUnit('''
def other_func(a):
break
''')
with self.assertRaisesRegex(RuntimeError, "Syntax"):
cu = torch.jit.CompilationUnit('''
def other_func(a):
for i in range(5):
def foo():
break
''')
with self.assertRaisesRegex(RuntimeError, "do not support break or continue inside"):
@torch.jit.script
def foo(x):
i = 0
for a in (1, "2", 1.5):
b = a
if x:
break
return b
def test_python_call(self):
def pyfunc(a):
return a * 3.0
cu = torch.jit.CompilationUnit('''
def other_func(a):
return a + a
def test_call_python(a):
b = pyfunc(a)
b = other_func(b)
i = 0
step = 1
while i < 10:
b = pyfunc(b)
if bool(b > 3.0):
b = pyfunc(b)
i = 11
return b
''')
inputs = self._make_scalar_vars([1], torch.float)
outputs = self._make_scalar_vars([54], torch.float)
self.assertEqual(cu.test_call_python(*inputs), outputs[0])
def test_python_call_failure(self):
with self.assertRaisesRegex(RuntimeError, "undefined value pyfunc2"):
def pyfunc(a):
return a * 3.0
cu = torch.jit.CompilationUnit('''
def other_func(a):
return a + a
def test_call_python(a):
b = pyfunc(a)
b = other_func(b)
i = 0
step = 1
while i < 10:
b = pyfunc2(b)
if b > 3.0:
b = pyfunc(b)
i = 11
return b
''')
inputs = self._make_scalar_vars([1], torch.float)
outputs = self._make_scalar_vars([54], torch.float)
self.assertEqual(cu.test_call_python(*inputs), outputs)
def test_type_call_in_script(self):
@torch.jit.script
def fn(x):
return type(x)
with self.assertRaisesRegex(RuntimeError, "value of type _TensorMeta"):
fn(torch.tensor(.5))
def test_python_call_annotation(self):
def pyfunc(a):
return a * 3.0
@torch.jit.script
def foo(a):
return pyfunc(a) + pyfunc(a)
inputs = self._make_scalar_vars([1], torch.float)
outputs = self._make_scalar_vars([6], torch.float)
self.assertEqual(foo(*inputs), outputs[0])
def test_python_call_annoytation_failure(self):
with self.assertRaisesRegex(RuntimeError, "undefined value pyfunc2"):
def pyfunc(a):
return a * 3.0
@torch.jit.script
def foo(a):
return pyfunc2(a) + pyfunc(a)
inputs = self._make_scalar_vars([1], torch.float)
outputs = self._make_scalar_vars([6], torch.float)
self.assertEqual(foo(*inputs), outputs[0])
def test_desugar_module(self):
import torch.nn.functional as F
def fn(x, slope):
a = torch.abs(x)
b = torch.nn.functional.prelu(x, slope)
c = F.prelu(x, slope)
return a, b, c
x = torch.arange(-3., 4)
slope = torch.tensor([0.5])
self.checkScript(fn, [x, slope], optimize=True)
def test_script_docstring(self):
@torch.jit.script
def with_docstring(x):
"""test str"""
y = x
"""y is the same as x"""
return y
self.assertEqual(with_docstring.__doc__, 'test str')
def test_script_method_docstring(self):
class A(torch.jit.ScriptModule):
@torch.jit.script_method
def with_docstring(self, x):
"""test str"""
y = x
"""y is the same as x"""
return y
a = A()
self.assertEqual(a.with_docstring.__doc__, 'test str')
def test_script_module(self):
class M1(torch.jit.ScriptModule):
def __init__(self):
super(M1, self).__init__()
self.weight = nn.Parameter(torch.randn(2))
@torch.jit.script_method
def forward(self, thing):
return self.weight + thing
class PModule(nn.Module):
def __init__(self):
super(PModule, self).__init__()
self.a = nn.Parameter(torch.randn(2, 3))
def forward(self, a):
return self.a.mm(a)
class M2(torch.jit.ScriptModule):
def __init__(self):
super(M2, self).__init__()
# test submodule
self.sub = M1()
self.sub2 = PModule()
# test parameters
self.weight = nn.Parameter(torch.randn(2, 3))
self.bias = nn.Parameter(torch.randn(2))
# test defining a method from a string
self.define("""
def hi(self, a):
return self.weight.mm(a)
""")
# test script methods
@torch.jit.script_method
def doit(self, input):
# test use of parameter
return self.weight.mm(input)
@torch.jit.script_method
def doit2(self, input):
return self.weight.mm(input)
@torch.jit.script_method
def forward(self, input):
a = self.doit(input)
b = self.doit2(input)
c = self.hi(input)
d = self.sub2(input)
return a + b + self.bias + self.sub(a) + c + d
with torch.jit.optimized_execution(False):
m2 = M2()
input = torch.randn(3, 2)
a = m2.weight.mm(input)
b = m2.weight.mm(input)
c = m2.weight.mm(input)
d = m2.sub2.a.mm(input)
ref = a + b + m2.bias + m2.sub.weight + a + c + d
self.assertEqual(ref, m2.forward(input))
m2.weight = nn.Parameter(torch.zeros_like(m2.weight))
m2.bias = nn.Parameter(torch.zeros_like(m2.bias))
m2.sub.weight = nn.Parameter(torch.zeros_like(m2.sub.weight))
m2.sub2.a.data.zero_()
self.assertEqual(torch.zeros(2, 2), m2.forward(torch.randn(3, 2)))
def test_irparser(self):
graph_str = """graph(%0 : Double(5, 5)):
# CHECK: aten::relu
%1 : Double(5, 5) = aten::relu(%0)
return (%1)
"""
FileCheck().run(graph_str, parse_ir(graph_str))
def test_parse_tensor_constants(self):
def foo():
return torch.zeros([4, 4])
foo_s = torch.jit.script(foo)
torch._C._jit_pass_constant_propagation(foo_s.graph)
g = str(foo_s.graph)
g_parsed = parse_ir(g, parse_tensor_constants=True)
self.assertEqual(str(canonical(g_parsed)), str(canonical(foo_s.graph)))
func = torch._C._create_function_from_graph("forward", g_parsed)
out_parsed = func()
out_func = foo()
# not checking data, just dtype, size etc
out_parsed[:] = 0
out_func[:] = 0
self.assertEqual(out_func, out_parsed)
with self.assertRaises(RuntimeError):
parse_ir(g, parse_tensor_constants=False)
def test_parse_nested_names(self):
g_str = """
graph(%x.1 : Tensor):
%3 : int = prim::Constant[value=1]()
%2 : int = prim::Constant[value=2]()
%hi.submod.value.5 : Tensor = aten::add(%x.1, %2, %3)
return (%hi.submod.value.5)
"""
g = parse_ir(g_str)
round_trip_g = parse_ir(str(g))
self.assertEqual(canonical(g), canonical(round_trip_g))
func1 = torch._C._create_function_from_graph("forward", g)
func2 = torch._C._create_function_from_graph("forward", round_trip_g)
self.assertEqual(func1(torch.ones([2])), func2(torch.ones([2])))
def test_is_after_use(self):
def sorted_input_use(g):
uses = list(next(g.inputs()).uses())
return sorted(uses, key=functools.cmp_to_key(type(uses[0]).isAfter))
@torch.jit.script
def foo(x):
a = x + 1
return (x, x, a)
uses_sorted = sorted_input_use(foo.graph)
# sorts last use to the end
self.assertFalse(uses_sorted[0].isAfter(uses_sorted[1]))
self.assertTrue(uses_sorted[0].user.kind() == "aten::add")
self.assertEqual(uses_sorted[1].offset, 0)
@torch.jit.script
def foo(x, cond: bool):
if cond:
return x + 3
else:
return x - 3
uses_sorted = sorted_input_use(foo.graph)
self.assertTrue(uses_sorted[0].user.kind() == "aten::add")
self.assertTrue(uses_sorted[1].user.kind() == "aten::sub")
@torch.jit.script
def foo(x, cond: bool, cond2: bool):
if cond:
return x + 3
elif cond2 :
return x - 3
return x / 3
graph1 = foo.graph
@torch.jit.script
def foo(x, cond: bool, cond2: bool):
if cond:
return x + 3
else:
if cond2 :
return x - 3
return x / 3
graph2 = foo.graph
for graph in [graph1, graph2]:
uses_sorted = sorted_input_use(graph)
self.assertTrue(uses_sorted[0].user.kind() == "aten::add")
self.assertTrue(uses_sorted[1].user.kind() == "aten::sub")
self.assertTrue(uses_sorted[2].user.kind() == "aten::div")
def test_canonicalize_control_outputs(self):
def test_all_outputs(g):
ifs = g.findAllNodes("prim::If")
loops = g.findAllNodes("prim::Loop")
def contained_blocks(node):
return len(node.findAllNodes("prim::If")) * 2 + len(node.findAllNodes("prim::Loop"))
for node in ifs + loops:
outs = list(node.outputs())
out_name = [x.debugName() for x in outs]
if len(out_name) == 0:
continue
fc = FileCheck()
# find the last output, then all subsequent uses
fc.check(out_name[-1] + " : ")
# skip past node body
for i in range(contained_blocks(node)):
fc.check("->")
if (node.kind() == "prim::If"):
fc.check("->").check("->").check("\n")
else:
fc.check("->").check("\n")
# the canonical order is the same order as the first use
# appears in text
for name in out_name:
fc.check(name)
fc.run(g)
@torch.jit.script
def test(x):
# type: (bool) -> Tuple[int, int]
b = 2
a = 1
if x:
a = 1
b = 2
x = False
if x:
b = a
else:
a = b
return a, b
test_all_outputs(test.graph)
@torch.jit.script
def test2(x):
# type: (bool) -> Tuple[int, int]
b = 2
a = 1
if x:
a = 1
b = 2
x = False
if x:
print(a)
else:
if x:
print(b)
return a, b
test_all_outputs(test2.graph)
@torch.jit.script
def test_loop(x, iter):
# type: (bool, int) -> (None)
a = 1
b = 2
c = 3
for i in range(iter):
a = 4
b = 5
c = 6
x = True
print(c)
if x:
print(a, b)
test_all_outputs(test_loop.graph)
@torch.jit.script
def loop_unused(iter):
# type: (int) -> (None)
a = 1
b = 2
c = 3
for i in range(iter):
c = c + 1
b = b + 1
a = a + 1
print(a, b)
print(c)
# c is used, then unused should be ordered by alphabetical
FileCheck().check(r"%c : int, %a : int, %b : int").run(loop_unused.graph)
def test_filecheck(self):
def test_check():
file = "232"
FileCheck().check("2").check("3").check("2").run(file)
FileCheck().check("232").run(file)
with self.assertRaisesRegex(RuntimeError, 'Expected to find "22"'):
FileCheck().check("22").run(file)
with self.assertRaisesRegex(RuntimeError, "CHECK: 3"):
FileCheck().check("3").check("3").run(file)
test_check()
def test_check_count():
file = "22222"
FileCheck().check_count("2", 5).run(file)
FileCheck().check_count("22", 2).run(file)
FileCheck().check_count("222", 1).run(file)
with self.assertRaisesRegex(RuntimeError, 'Expected to not find'):
FileCheck().check_count("2", 4, exactly=True).run(file)
with self.assertRaisesRegex(RuntimeError, 'Expected to find "22"'):
FileCheck().check_count("22", 3).run(file)
with self.assertRaisesRegex(RuntimeError, "CHECK-COUNT-6: 2"):
FileCheck().check_count("2", 6).run(file)
test_check_count()
def test_check_same():
file = "22\n33"
FileCheck().check_same("22").run(file)
with self.assertRaisesRegex(RuntimeError, "Expected to not find"):
FileCheck().check_same("33").run(file)
file = "22 1 3"
FileCheck().check("2").check_same("3").run(file)
FileCheck().check_count("2", 2).check_same("3").run(file)
test_check_same()
def test_check_next():
file = "\n1\n2\n3"
FileCheck().check("1").check_next("2").check_next("3").run(file)
FileCheck().check_next("1").check_next("2").check_next("3").run(file)
with self.assertRaisesRegex(RuntimeError, "Expected to find"):
FileCheck().check("1").check_next("2").run("12")
with self.assertRaisesRegex(RuntimeError, "Expected to not find"):
FileCheck().check("1").check_next("2").run("1\n\n2")
test_check_next()
def test_check_dag():
fc = FileCheck().check_dag("1").check_dag("2").check_not("2")
fc.run("12")
fc.run("21")
fc = FileCheck()
fc.check_not("3").check_dag("1").check_dag("2").check_not("3")
fc.run("1 3 2")
fc.run("2 3 1")
fc = FileCheck().check_dag("1").check_dag("2").check("3")
with self.assertRaisesRegex(RuntimeError, 'Expected to find "3" but did not find it'):
fc.run("1 3 2")
test_check_dag()
def test_check_not():
FileCheck().check_not("2").check("1").run("12")
FileCheck().check("2").check_not("2").run("12")
with self.assertRaisesRegex(RuntimeError, 'Expected to not find "2"'):
FileCheck().check_not("2").check("1").run("21")
with self.assertRaisesRegex(RuntimeError, 'Expected to not find "1"'):
FileCheck().check("2").check_not("1").run("21")
# checks with distinct range matchings
fb = FileCheck().check_count("2", 2).check_count("2", 2).check_not("2")
with self.assertRaisesRegex(RuntimeError, 'Expected to not find "2"'):
fb.run("22 2 22")
fb = FileCheck().check_count("2", 2).check_not("1").check_count("2", 2)
with self.assertRaisesRegex(RuntimeError, 'Expected to not find "1"'):
fb.run("22 1 22")
def _dtype_to_jit_name(self, dtype):
if(dtype == torch.float32):
return "Float"
if(dtype == torch.float64):
return "Double"
if(dtype == torch.int64):
return "Long"
if(dtype == torch.int32):
return "Int"
if(dtype == torch.bool):
return "Bool"
raise RuntimeError('dtype not handled')
def _dtype_to_expect(self, dtype, dim=0):
param = ', '.join(['*'] * dim + ['device=cpu'])
param = '(' + param + ')'
jit_type = self._dtype_to_jit_name(dtype)
if dim >= 0:
return jit_type + param
# special case representing wrapped number
else:
return jit_type.lower()
def _test_dtype_op_shape(self, ops, args, input_dims=1):
if input_dims < 1:
raise RuntimeError("input dims must be at least 1")
dtypes = [torch.float32, torch.float64, torch.int64, torch.int32]
str_args = ', '.join([str(arg) for arg in args]) + (', ' if len(args) else '')
tensor_data = ('[' * input_dims) + '1, 2, 3' + (input_dims * ']')
template = dedent('''
def func():
return {return_line}
''')
for op in ops:
for dtype in (dtypes + [None]):
for tensor_type in dtypes:
# a couple of ops aren't implemented for non-floating types
if(not tensor_type.is_floating_point or (dtype is not None and not dtype.is_floating_point)):
if op in ['mean', 'softmax', 'log_softmax']:
continue
return_line = "torch.tensor({}, dtype={}).{}({}dtype={})".format(tensor_data, tensor_type, op, str_args, dtype)
# uncomment for debugging a failed test:
# print("testing {}".format(return_line))
code = template.format(return_line=return_line)
scope = {}
exec(code, globals(), scope)
cu = torch.jit.CompilationUnit(code)
graph = cu.func.graph
torch._C._jit_pass_complete_shape_analysis(graph, (), False)
input_array = [1, 2, 3]
for _ in range(1, input_dims):
input_array = [input_array]
t = torch.tensor(input_array, dtype=tensor_type)
attr = getattr(t, op)
kwargs = {'dtype': dtype}
result = attr(*args, **kwargs)
expect = self._dtype_to_expect(result.dtype, result.dim())
FileCheck().check("aten::tensor").check(expect).run(graph)
def test_dtype_op_shape(self):
ops = ['prod']
self._test_dtype_op_shape(ops, args=[])
self._test_dtype_op_shape(ops, args=[0, False])
self._test_dtype_op_shape(ops, args=[0, False])
self._test_dtype_op_shape(ops, args=[0, True])
def test_dtype_op_shape2(self):
ops = ['cumprod', 'cumsum', 'softmax', 'log_softmax']
self._test_dtype_op_shape(ops, args=[0])
self._test_dtype_op_shape(ops, args=[1], input_dims=4)
def _test_binary_op_shape(self, ops, input_dims=1):
dtypes = [torch.float32, torch.float64, torch.int64, torch.int32, torch.bool]
if input_dims == 0:
shape = '1'
else:
shape = '[' + ('1,' * 4) + ']'
for _ in range(1, input_dims):
shape = '[' + ",".join([shape] * 4) + ']'
template = dedent('''
def func():
arg1 = {}
arg2 = {}
return torch.{}(arg1, arg2)
''')
args = []
for dtype in dtypes:
args = args + ["torch.tensor({}, dtype={})".format(shape, dtype)]
args = args + [1, 1.5]
def isBool(arg):
return type(arg) == bool or (type(arg) == str and "torch.bool" in arg)
for op in ops:
for first_arg in args:
for second_arg in args:
# subtract not supported for bool
if (op == 'sub' or op == 'div') and (isBool(first_arg) or isBool(second_arg)):
continue
# div is not implemented correctly for mixed-type or int params
if (op == 'div' and (type(first_arg) != type(second_arg) or
isinstance(first_arg, int) or
(isinstance(first_arg, str) and 'int' in first_arg))):
continue
return_line = "torch.{}({}, {})".format(op, first_arg, second_arg)
# uncomment for debugging a failed test:
# print("testing {}".format(return_line))
code = template.format(first_arg, second_arg, op)
scope = {}
exec(code, globals(), scope)
non_jit_result = scope['func']()
cu = torch.jit.CompilationUnit(code)
graph = cu.func.graph
torch._C._jit_pass_complete_shape_analysis(graph, (), False)
# use dim=-1 to represent a python/jit scalar.
dim = -1 if type(first_arg) != str and type(second_arg) != str else non_jit_result.dim()
dtype = non_jit_result.dtype
# jit only supports int/float scalars.
if dim < 0:
if dtype == torch.int64:
dtype = torch.int32
if dtype == torch.float64:
dtype = torch.float32
expect = self._dtype_to_expect(dtype, dim)
jit_output = next(graph.outputs())
check = FileCheck()
check.check(expect).run(str(jit_output))
def test_binary_op_shape(self):
self._test_binary_op_shape(['mul', 'div', 'add', 'sub'], 0)
self._test_binary_op_shape(['mul', 'div', 'add', 'sub'], 3)
def test_no_dtype_shape(self):
@torch.jit.script
def foo(x):
scalar_number = x.item()
return x.add(scalar_number)
@torch.jit.script
def foo2(x):
scalar_number = x.item()
return torch.tensor(1).add(scalar_number)
t = torch.tensor(5)
g = foo.graph_for(t)
type = next(g.outputs())
self.assertTrue(type.type() == torch._C.TensorType.get())
g2 = foo2.graph_for(t)
type = next(g.outputs())
self.assertTrue(type.type() == torch._C.TensorType.get())
def test_filecheck_parse(self):
def test_check():
file = """
# CHECK: 2
# CHECK: 3
# CHECK: 2
232
"""
FileCheck().run(checks_file=file, test_file=file)
file = """
# CHECK: 232
232
"""
FileCheck().run(file, "232")
with self.assertRaisesRegex(RuntimeError, 'Expected to find "232"'):
FileCheck().run(file, "22")
with self.assertRaisesRegex(RuntimeError, 'Expected to find "22"'):
FileCheck().run("# CHECK: 22", "23")
test_check()
def test_check_count():
file = "22222"
FileCheck().run("# CHECK-COUNT-5: 2", file)
FileCheck().run("# CHECK-COUNT-EXACTLY-5: 2", file)
FileCheck().run("# CHECK-COUNT-2: 22", file)
FileCheck().run("# CHECK-COUNT-1: 222", file)
with self.assertRaisesRegex(RuntimeError, 'Expected to not find'):
FileCheck().run("# CHECK-COUNT-EXACTLY-2: 2", file)
test_check_count()
def test_check_same():
file = "22\n33"
FileCheck().run("# CHECK-SAME: 22", file)
with self.assertRaisesRegex(RuntimeError, "Expected to not find"):
FileCheck().run("# CHECK-SAME: 33", file)
file = "22 1 3"
FileCheck().run("# CHECK: 2\n # CHECK-SAME: 3", file)
FileCheck().run("# CHECK-COUNT-2: 2\n # CHECK-SAME: 3", file)
test_check_same()
def test_bad_input():
with self.assertRaisesRegex(RuntimeError, "Check for bad input"):
FileCheck().run("", "1")
with self.assertRaisesRegex(RuntimeError, "Could not parse check"):
FileCheck().run("# CHECK1", "")
test_bad_input()
def test_script_module_call_noscript(self):
class M(torch.jit.ScriptModule):
def __init__(self):
super(M, self).__init__()
self.value = 1
@torch.jit.ignore
def foo(self):
return torch.ones(2, 2) + self.value
@torch.jit.script_method
def forward(self, input):
return input + self.foo()
with torch.jit.optimized_execution(False):
m = M()
input = torch.randn(2, 2)
o = m(input)
self.assertEqual(o, input + torch.ones(2, 2) + 1)
# check that we can change python attributes
# and that those changes are picked up in script methods
m.value = 2
o = m(input)
self.assertEqual(o, input + torch.ones(2, 2) + 2)
def test_script_module_nochange_submodule(self):
class M(torch.jit.ScriptModule):
def __init__(self):
super(M, self).__init__()
self.sub = nn.Linear(5, 5)
@torch.jit.script_method
def forward(self, input):
return self.sub(input)
with torch.jit.optimized_execution(False):
m = M()
input = torch.randn(1, 5, 5)
o = m(input)
self.assertEqual(o, m.sub(input))
with self.assertRaisesRegex(RuntimeError, "Cannot re-assign"):
m.sub = nn.Linear(5, 5)
def test_module_apis(self):
class Sub(torch.nn.Module):
def __init__(self):
super(Sub, self).__init__()
def forward(self, thing):
return thing - 2
class Double(torch.nn.Module):
def __init__(self):
super(Double, self).__init__()
def forward(self, thing):
return thing * 2
class MyMod(torch.nn.Module):
def __init__(self):
super(MyMod, self).__init__()
self.mod = (Sub())
self.mod2 = (Sub())
self.mod3 = nn.Sequential(nn.Sequential(Sub()))
self.mod4 = nn.Sequential(Sub(), Double())
@torch.jit.export
def method(self, x, x1, y, y1):
mod_names = ""
for name, mod in self.named_modules():
mod_names = mod_names + " " + name
x = mod(x)
children_names = ""
for name, mod in self.named_children():
children_names = children_names + " " + name
x1 = mod(x1)
for mod in self.modules():
y = mod(y)
for mod in self.children():
y1 = mod(y1)
return mod_names, children_names, x, x1, y, y1
def forward(self, x):
return x + 2
mod = torch.jit.script(MyMod())
inps = tuple([torch.tensor(i) for i in range(1, 5)])
self.assertEqual(mod.method(*inps), MyMod().method(*inps))
def test_script_module_const(self):
class M(torch.jit.ScriptModule):
__constants__ = ['b', 'i', 'c', 's']
def __init__(self):
super(M, self).__init__()
self.b = False
self.i = 1
self.c = 3.5
self.s = ["hello"]
@torch.jit.script_method
def forward(self):
return self.b, self.i, self.c
with torch.jit.optimized_execution(False):
m = M()
o0, o1, o2 = m()
self.assertEqual(o0, 0)
self.assertEqual(o1, 1)
self.assertEqual(o2, 3.5)
def test_script_module_fail_exist(self):
class M(torch.jit.ScriptModule):
def __init__(self):
super(M, self).__init__()
@torch.jit.script_method
def forward(self, x):
return x + self.whatisgoingon
with self.assertRaisesRegex(RuntimeError, "Module 'M' has no attribute"):
M()
@unittest.skip("[module dedupe] currently NoneType refinement on optional attributes doesn't work.")
def test_script_module_none_exist_fail(self):
class M(torch.jit.ScriptModule):
def __init__(self, my_optional):
super(M, self).__init__()
self.my_optional = my_optional
@torch.jit.script_method
def forward(self, x):
if self.my_optional is not None:
return torch.neg(x) + self.my_optional
return torch.neg(x)
with self.assertRaisesRegex(RuntimeError, "has no attribute 'my_optional'"):
x = torch.rand(3, 4)
fb = M(None)
fb(x)
def test_script_module_invalid_consts(self):
class Foo(torch.jit.ScriptModule):
__constants__ = ['invalid']
def __init__(self):
super(Foo, self).__init__()
self.invalid = [nn.Linear(3, 4)]
with self.assertRaisesRegex(
TypeError,
"Linear' object in attribute 'Foo.invalid' is not a valid constant"):
Foo()
class Foo2(torch.jit.ScriptModule):
__constants__ = ['invalid']
def __init__(self):
super(Foo2, self).__init__()
self.invalid = type(1)
with self.assertRaisesRegex(TypeError, "not a valid constant"):
Foo2()
class Foo3(torch.jit.ScriptModule):
__constants__ = ['invalid']
def __init__(self):
super(Foo3, self).__init__()
self.invalid = (3, 4, {})
with self.assertRaisesRegex(TypeError, "not a valid constant"):
Foo3()
class Foo4(torch.jit.ScriptModule):
__constants__ = ['invalid']
def __init__(self):
super(Foo4, self).__init__()
self.invalid = np.int64(5)
# verify that we capture human understandable class name
with self.assertRaisesRegex(TypeError, "numpy.int64"):
Foo4()
def test_script_module_param_buffer_mutation(self):
# TODO: add param mutation test case after JIT support it
class ModuleBufferMutate(torch.jit.ScriptModule):
def __init__(self):
super(ModuleBufferMutate, self).__init__()
self.register_buffer('running_var', torch.tensor(0, dtype=torch.long))
@torch.jit.script_method
def forward(self):
if self.training:
self.running_var += 1
return self.running_var
with torch.jit.optimized_execution(False):
m = ModuleBufferMutate()
self.assertEqual(m(), 1)
m.eval()
self.assertEqual(m(), 1)
def test_script_module_for(self):
class M(torch.jit.ScriptModule):
__constants__ = ['b']
def __init__(self):
super(M, self).__init__()
self.b = [1, 2, 3, 4]
@torch.jit.script_method
def forward(self):
sum = 0
for i in self.b:
sum += i
return sum
with torch.jit.optimized_execution(False):
m = M()
self.assertEqual(m(), 10)
def test_override_magic(self):
class OverrideMagic(nn.Module):
def __init__(self):
super(OverrideMagic, self).__init__()
@torch.jit.export
def __len__(self):
return 10
mod = OverrideMagic()
self.assertEqual(len(mod), len(torch.jit.script(mod)))
class OverrideMagicSeq(nn.Sequential):
def __init__(self):
super(OverrideMagicSeq, self).__init__()
@torch.jit.export
def __len__(self):
return 10
mod = OverrideMagicSeq()
self.assertEqual(len(mod), len(torch.jit.script(mod)))
self.assertTrue(torch.jit.script(mod))
def test_script_module_for2(self):
class Sub(torch.jit.ScriptModule):
def __init__(self):
super(Sub, self).__init__()
self.weight = nn.Parameter(torch.randn(2))
@torch.jit.script_method
def forward(self, thing):
return self.weight + thing
class M(torch.jit.ScriptModule):
def __init__(self):
super(M, self).__init__()
self.mods = nn.ModuleList([Sub() for i in range(10)])
@torch.jit.script_method
def forward(self, v):
for m in self.mods:
v = m(v)
return v
with torch.jit.optimized_execution(False):
i = torch.empty(2)
m = M()
o = m(i)
v = i
for sub in m.mods:
v = sub(v)
self.assertEqual(o, v)
with self.assertRaisesRegex(Exception, "object is not iterable"):
print(list(m))
def test_attr_qscheme_script(self):
class Foo(torch.nn.Module):
def __init__(self):
super(Foo, self).__init__()
self.qscheme = torch.per_tensor_affine
def forward(self):
if self.qscheme == torch.per_tensor_symmetric:
return 3
else:
return 4
f = Foo()
scripted = torch.jit.script(f)
self.assertEqual(f(), scripted())
def test_script_module_const_submodule_fail(self):
class Sub(torch.jit.ScriptModule):
def __init__(self):
super(Sub, self).__init__()
self.weight = nn.Parameter(torch.randn(2))
@torch.jit.script_method
def forward(self, thing):
return self.weight + thing
class M(torch.jit.ScriptModule):
def __init__(self):
super(M, self).__init__()
self.mods = [Sub() for _ in range(10)]
@torch.jit.script_method
def forward(self):
for _ in self.mods:
print(1)
return 4
with self.assertRaisesRegex(RuntimeError, "has no attribute 'mods'"):
M()
class DerivedStateModule(torch.jit.ScriptModule):
def __init__(self):
super(TestScript.DerivedStateModule, self).__init__()
self.param = torch.nn.Parameter(torch.ones(3, 4, dtype=torch.float))
self.register_buffer('derived', torch.neg(self.param).detach().clone())
# This is a flag so we can test that the pack method was called
self.register_buffer('pack_called', torch.zeros(1, dtype=torch.long))
# This is a flag so we can test that the unpack method was called
self.register_buffer('unpack_called', torch.zeros(1, dtype=torch.long))
@torch.jit.script_method
def _pack(self):
self.pack_called.set_(torch.ones(1, dtype=torch.long))
self.derived.set_(torch.rand(1, dtype=torch.float).detach())
@torch.jit.script_method
def _unpack(self):
self.unpack_called.set_(torch.ones(1, dtype=torch.long))
self.derived.set_(torch.neg(self.param).detach())
@torch.jit.script_method
def forward(self, x):
return x + self.derived
def test_pack_unpack_state(self):
sm = TestScript.DerivedStateModule()
x = torch.rand(3, 4, dtype=torch.float)
torch.testing.assert_close(sm(x), x + torch.neg(torch.ones(3, 4, dtype=torch.float)))
# Test save path
self.assertFalse(sm.pack_called.item())
self.assertFalse(sm.unpack_called.item())
imported = self.getExportImportCopyWithPacking(sm)
# ensure pack was called before serialization
self.assertTrue(sm.pack_called.item())
# ensure unpack was called after serialization so as to leave the module in an initialized state
self.assertTrue(sm.unpack_called.item())
torch.testing.assert_close(sm.derived, torch.neg(sm.param))
# Test load paths
self.assertTrue(imported.unpack_called.item())
torch.testing.assert_close(imported(x), x + torch.neg(torch.ones(3, 4, dtype=torch.float)))
@unittest.skipIf(not TEST_MKL, "PyTorch is built without MKL support")
@unittest.skipIf(True, "Skipping while landing PR stack")
def test_torch_functional(self):
def stft(input, n_fft):
# type: (Tensor, int) -> Tensor
return torch.stft(input, n_fft, return_complex=True)
inps = (torch.randn(10), 7)
self.assertEqual(stft(*inps), torch.jit.script(stft)(*inps))
def istft(input, n_fft):
# type: (Tensor, int) -> Tensor
return torch.istft(input, n_fft)
inps2 = (stft(*inps), inps[1])
self.assertEqual(istft(*inps2), torch.jit.script(istft)(*inps2))
def lu_unpack(x):
A_LU, pivots = torch.linalg.lu_factor(x)
return torch.lu_unpack(A_LU, pivots)
for shape in ((3, 3), (5, 3, 3), (7, 3, 5, 5), (7, 5, 3, 3, 3)):
a = torch.randn(*shape)
self.checkScript(lu_unpack, (a,))
def cdist_fn():
a = torch.tensor([[0.9041, 0.0196], [-0.3108, -2.4423], [-0.4821, 1.059]])
b = torch.tensor([[-2.1763, -0.4713], [-0.6986, 1.3702]])
return torch.cdist(a, b, compute_mode="use_mm_for_euclid_dist")
self.checkScript(cdist_fn, ())
def norm():
c = torch.tensor([[1, 2, 3], [-1, 1, 4]], dtype=torch.float)
return torch.norm(c, p="fro"), torch.norm(c, p="nuc"), torch.norm(c), torch.norm(c, p=.5)
self.checkScript(norm, ())
def torch_unique(dim: Optional[int]):
ten = torch.unique(torch.tensor([[1, 3], [2, 3]], dtype=torch.long))
a = torch.unique(ten, dim=dim)
b = torch.unique(ten, return_counts=True, dim=dim)
c = torch.unique(ten, return_inverse=True, dim=dim)
d = torch.unique(ten, return_counts=True, return_inverse=True, dim=dim)
return a, b, c, d
self.checkScript(torch_unique, (None,))
self.checkScript(torch_unique, (0,))
def torch_unique_consecutive(dim: Optional[int]):
ten = torch.unique(torch.tensor([[1, 3], [3, 2], [3, 2], [2, 3]], dtype=torch.long))
a = torch.unique_consecutive(ten, dim=dim)
b = torch.unique_consecutive(ten, return_counts=True, dim=dim)
c = torch.unique_consecutive(ten, return_inverse=True, dim=dim)
d = torch.unique_consecutive(ten, return_counts=True, return_inverse=True, dim=dim)
return a, b, c, d
self.checkScript(torch_unique_consecutive, (None,))
self.checkScript(torch_unique_consecutive, (0,))
def test_torch_functional_tensordot_int(self):
def tensordot_dims_int(a: torch.Tensor, b: torch.Tensor, dims: int):
return torch.tensordot(a, b, dims=dims)
a = torch.arange(120.).reshape(2, 3, 4, 5)
b = torch.arange(840.).reshape(4, 5, 6, 7)
dims = 2
self.checkScript(tensordot_dims_int, (a, b, dims))
def test_torch_functional_tensordot_tensor(self):
def tensordot_dims_tensor(a: torch.Tensor, b: torch.Tensor, dims: torch.Tensor):
return torch.tensordot(a, b, dims=dims)
a = torch.arange(120.).reshape(2, 3, 4, 5)
b = torch.arange(840.).reshape(4, 5, 6, 7)
dims = torch.tensor([2])
self.checkScript(tensordot_dims_tensor, (a, b, dims))
a = torch.arange(60.).reshape(3, 4, 5)
b = torch.arange(24.).reshape(4, 3, 2)
dims = torch.tensor([[1, 0], [0, 1]], dtype=torch.long)
self.checkScript(tensordot_dims_tensor, (a, b, dims))
def test_torch_functional_tensordot_list(self):
def tensordot_dims_list(a: torch.Tensor, b: torch.Tensor, dims: List[List[int]]):
return torch.tensordot(a, b, dims=dims)
a = torch.arange(60.).reshape(3, 4, 5)
b = torch.arange(24.).reshape(4, 3, 2)
dims = [[1, 0], [0, 1]]
self.checkScript(tensordot_dims_list, (a, b, dims))
def test_torch_functional_tensordot_tuple(self):
def tensordot_dims_tuple(a: torch.Tensor, b: torch.Tensor, dims: Tuple[List[int], List[int]]):
return torch.tensordot(a, b, dims=dims)
a = torch.arange(60.).reshape(3, 4, 5)
b = torch.arange(24.).reshape(4, 3, 2)
dims = ([1, 0], [0, 1])
self.checkScript(tensordot_dims_tuple, (a, b, dims))
def test_missing_getstate(self):
class Foo(torch.nn.Module):
def __init__(self):
super(Foo, self).__init__()
self.x = 1
def forward(self, x):
return x * self.x
@torch.jit.export
def __setstate__(self, state):
self.x = state[0]
self.training = state[1]
with self.assertRaisesRegex(RuntimeError, "getstate"):
scripted = torch.jit.script(Foo())
def test_inlining_cleanup(self):
def foo(x):
return F.linear(x, x)
@torch.jit.script
def fee(x):
return foo(x)
# inlining optimizations should have cleaned up linear if statement
self.run_pass("inline", fee.graph)
FileCheck().check_not("prim::If").run(fee.graph)
def test_pack_unpack_nested(self):
class SubSubMod(torch.jit.ScriptModule):
def __init__(self):
super(SubSubMod, self).__init__()
self.register_buffer('buf', torch.ones(3, 4) * 3)
@torch.jit.script_method
def _pack(self):
self.buf.set_(torch.zeros(1, dtype=torch.double))
@torch.jit.script_method
def _unpack(self):
self.buf.set_(torch.ones(3, 4, dtype=torch.double) * 3)
@torch.jit.script_method
def forward(self, x):
return x + self.buf
class SubMod(torch.jit.ScriptModule):
def __init__(self):
super(SubMod, self).__init__()
self.register_buffer('buf', torch.ones(3, 4) * 2)
self.ssm = SubSubMod()
@torch.jit.script_method
def _pack(self):
self.buf.set_(torch.zeros(1, dtype=torch.double))
@torch.jit.script_method
def _unpack(self):
self.buf.set_(torch.ones(3, 4, dtype=torch.double) * 2)
@torch.jit.script_method
def forward(self, x):
return self.ssm(x + self.buf)
class Mod(torch.jit.ScriptModule):
def __init__(self):
super(Mod, self).__init__()
self.submod = SubMod()
self.register_buffer('buf', torch.ones(3, 4) * 1)
@torch.jit.script_method
def _pack(self):
self.buf.set_(torch.zeros(1, dtype=torch.double))
@torch.jit.script_method
def _unpack(self):
self.buf.set_(torch.ones(3, 4, dtype=torch.double))
@torch.jit.script_method
def forward(self, x):
return self.submod(x + self.buf)
m = Mod()
torch.testing.assert_close(m(torch.zeros(3, 4)), torch.ones(3, 4) * 6)
m.apply(lambda s: s._pack())
torch.testing.assert_close(m(torch.zeros(3, 4)), torch.zeros(3, 4))
m.apply(lambda s: s._unpack())
torch.testing.assert_close(m(torch.zeros(3, 4)), torch.ones(3, 4) * 6)
def test_torch_any(self):
def fn(x):
return torch.any(x)
def fn1(x, dim: int):
return torch.any(x, dim)
self.checkScript(fn, (torch.randn(3, 4), ))
self.checkScript(fn, (torch.empty(3), ))
self.checkScript(fn, (torch.empty(1), ))
self.checkScript(fn, (torch.ones(3, 4),))
self.checkScript(fn, (torch.zeros(5, 7, 1),))
self.checkScript(fn1, (torch.empty(3, 4), -2))
self.checkScript(fn1, (torch.randn(3, 8), 1))
self.checkScript(fn1, (torch.zeros(3, 6, 9), -3))
self.checkScript(fn1, (torch.empty(5), 0))
def test_any(self):
def fn(x: List[int]):
return any(x)
def fn1(x: List[float]):
return any(x)
def fn2(x: List[bool]):
return any(x)
def fn3(x: List[str]):
return any(x)
self.checkScript(fn, ([0, 0, 0, 0], ))
self.checkScript(fn, ([0, 3, 0], ))
self.checkScript(fn, ([], ))
self.checkScript(fn1, ([1.0, 2.0, 3.0], ))
self.checkScript(fn1, ([0.0, 0.0, 0.0], ))
self.checkScript(fn1, ([0, 0, 0], ))
self.checkScript(fn1, ([], ))
self.checkScript(fn2, ([True, False, False], ))
self.checkScript(fn2, ([False, False, False], ))
self.checkScript(fn2, ([True, True, True, True], ))
self.checkScript(fn2, ([], ))
self.checkScript(fn3, (["", "", ""], ))
self.checkScript(fn3, (["", "", "", "-1"], ))
self.checkScript(fn3, ([], ))
def test_script_module_not_tuple(self):
class M(torch.jit.ScriptModule):
__constants__ = ['mods']
def __init__(self):
super(M, self).__init__()
self.mods = 1
@torch.jit.script_method
def forward(self, v):
for m in self.mods:
print(m)
return v
with self.assertRaisesRegex(RuntimeError, "'int' object is not iterable"):
M()
def test_attr_module_constants(self):
class M2(torch.jit.ScriptModule):
def __init__(self, mod_list):
super(M2, self).__init__()
self.mods = mod_list
@torch.jit.script_method
def forward(self, x):
return self.mods.forward(x)
with torch.jit.optimized_execution(False):
m = M2(nn.Sequential(nn.ReLU()))
self.assertExportImportModule(m, (torch.randn(2, 2),))
def test_script_sequential_for(self):
class Sub(torch.jit.ScriptModule):
def __init__(self):
super(Sub, self).__init__()
self.weight = nn.Parameter(torch.randn(2))
@torch.jit.script_method
def forward(self, thing):
return self.weight + thing
class M(torch.jit.ScriptModule):
def __init__(self):
super(M, self).__init__()
self.mods = nn.Sequential(Sub(), Sub(), Sub())
@torch.jit.script_method
def forward(self, v):
for m in self.mods:
v = m(v)
return v
@torch.jit.script_method
def forward2(self, v):
return self.mods(v)
with torch.jit.optimized_execution(False):
i = torch.empty(2)
m = M()
o = m(i)
v = i
for sub in m.mods._modules.values():
v = sub(v)
self.assertEqual(o, v)
o2 = m.forward2(i)
self.assertEqual(o2, v)
def test_script_sequential_sliced_iteration(self):
class seq_mod(nn.Module):
def __init__(self):
super(seq_mod, self).__init__()
self.layers = [nn.ReLU(), nn.ReLU(), nn.ReLU()]
self.layers = nn.Sequential(*self.layers)
def forward(self, input):
x = self.layers[0].forward(input)
for layer in self.layers[1:3]:
x = layer.forward(x)
for layer in self.layers[2:]:
x = layer.forward(x)
return x
seq = seq_mod()
self.checkModule(seq, [torch.tensor([-2, 1, -1, 2])])
def test_script_sequential_orderdict(self):
class M(torch.jit.ScriptModule):
def __init__(self):
super(M, self).__init__()
self.mods = nn.Sequential(OrderedDict([
("conv", nn.Conv2d(1, 20, 5)),
("relu", nn.ReLU())
]))
@torch.jit.script_method
def forward(self, input):
return self.mods(input)
m = M()
self.assertTrue('mods.conv.weight' in m.state_dict().keys())
def test_script_sequential_multi_output_fail(self):
class Sub(torch.jit.ScriptModule):
def __init__(self):
super(Sub, self).__init__()
self.weight = nn.Parameter(torch.randn(2))
@torch.jit.script_method
def forward(self, thing):
return self.weight + thing
class ReturnMulti(torch.jit.ScriptModule):
def __init__(self):
super(ReturnMulti, self).__init__()
@torch.jit.script_method
def forward(self, x):
return x, x, x
class HaveSequential(torch.jit.ScriptModule):
def __init__(self):
super(HaveSequential, self).__init__()
self.someseq = nn.Sequential(
Sub(),
ReturnMulti(),
Sub()
)
@torch.jit.script_method
def forward(self, x):
return self.someseq(x)
with self.assertRaisesRegex(RuntimeError, "(Tensor, Tensor, Tensor)"):
with torch.jit.optimized_execution(False):
hs = HaveSequential()
i = torch.empty(2)
hs(i)
@_tmp_donotuse_dont_inline_everything
def test_script_sequential_in_mod_list(self):
class Sub(torch.jit.ScriptModule):
def __init__(self):
super(Sub, self).__init__()
self.weight = nn.Parameter(torch.randn(2))
@torch.jit.script_method
def forward(self, thing):
return self.weight + thing
class M(torch.jit.ScriptModule):
def __init__(self):
super(M, self).__init__()
self.mods = nn.ModuleList([Sub(), nn.Sequential(Sub(), nn.Sequential(Sub(), Sub()), Sub())])
@torch.jit.script_method
def forward(self, v):
for mod in self.mods:
v = mod(v)
return v
m = M()
graph = str(m.graph)
self.assertTrue(graph.count("prim::CallMethod") == 2)
self.assertTrue("python" not in graph)
@_tmp_donotuse_dont_inline_everything
def test_script_nested_mod_list(self):
class Sub(torch.jit.ScriptModule):
def __init__(self):
super(Sub, self).__init__()
self.weight = nn.Parameter(torch.randn(2))
@torch.jit.script_method
def forward(self, thing):
return self.weight + thing
class M(torch.jit.ScriptModule):
def __init__(self):
super(M, self).__init__()
self.mods = nn.ModuleList([nn.ModuleList([Sub()]), nn.Sequential(Sub()), nn.ModuleList([Sub(), Sub()])])
@torch.jit.script_method
def forward(self, v):
for mod in self.mods:
for m in mod:
v = m(v)
return v
m = M()
graph = str(m.graph)
self.assertTrue(graph.count("prim::CallMethod") == 4)
self.assertTrue("python" not in graph)
def test_constant_as_attr(self):
class M(torch.jit.ScriptModule):
__constants__ = ['dim']
def __init__(self):
super(M, self).__init__()
self.dim = 1
@torch.jit.script_method
def forward(self, v):
return torch.cat([v, v, v], dim=self.dim)
v = torch.zeros(1, 1)
with torch.jit.optimized_execution(False):
self.assertEqual(torch.cat([v, v, v], dim=1), M()(v))
class StarTestSumStarred(torch.nn.Module):
def __init__(self):
super(TestScript.StarTestSumStarred, self).__init__()
def forward(self, *inputs):
output = inputs[0]
for i in range(1, len(inputs)):
output += inputs[i]
return output
class StarTestReturnThree(torch.nn.Module):
def __init__(self):
super(TestScript.StarTestReturnThree, self).__init__()
def forward(self, rep):
return rep, rep, rep
def test_script_star_expr(self):
class M2(torch.jit.ScriptModule):
def __init__(self):
super(M2, self).__init__()
self.m = torch.jit.trace(TestScript.StarTestSumStarred(),
(torch.ones(4, 3), torch.ones(4, 3), torch.ones(4, 3)))
self.g = torch.jit.trace(TestScript.StarTestReturnThree(), torch.ones(4, 3))
@torch.jit.script_method
def forward(self, rep):
tup = self.g(rep)
return self.m(*tup)
m = M2()
self.assertEqual(m(torch.zeros(4, 3)), 3 * torch.zeros(4, 3))
def test_script_star_expr_string(self):
class M2(torch.jit.ScriptModule):
def __init__(self):
super(M2, self).__init__()
self.m = torch.jit.trace(TestScript.StarTestSumStarred(),
(torch.ones(4, 3), torch.ones(4, 3), torch.ones(4, 3)))
self.g = torch.jit.trace(TestScript.StarTestReturnThree(), torch.ones(4, 3))
self.define('''
def forward(self, rep):
tup = self.g(rep)
return self.m(*tup)
''')
m = M2()
self.assertEqual(m(torch.zeros(4, 3)), 3 * torch.zeros(4, 3))
class StarTestSumAndReturnThree(torch.nn.Module):
def __init__(self):
super(TestScript.StarTestSumAndReturnThree, self).__init__()
def forward(self, *inputs):
output = inputs[0]
for i in range(1, len(inputs)):
output += inputs[i]
return output, output, output
def test_script_star_assign(self):
class M2(torch.jit.ScriptModule):
def __init__(self):
super(M2, self).__init__()
self.g = torch.jit.trace(TestScript.StarTestSumAndReturnThree(), torch.ones(4, 3))
self.define('''
def forward(self, rep):
head, *tail = self.g(rep)
return head
''')
m = M2()
self.assertEqual(m(torch.zeros(4, 3)), 3 * torch.zeros(4, 3))
def test_script_module_star_assign2(self):
class M2(torch.jit.ScriptModule):
def __init__(self):
super(M2, self).__init__()
self.g = torch.jit.trace(
TestScript.StarTestSumAndReturnThree(),
(torch.ones(4, 3), torch.ones(4, 3), torch.ones(4, 3)),
_force_outplace=True)
self.define('''
def forward(self, rep):
*head, tail = self.g(rep, rep, rep)
return tail
''')
m = M2()
self.assertEqual(m(torch.ones(4, 3)), 3 * torch.ones(4, 3))
def test_script_module_star_assign2_inplace(self):
class M2(torch.jit.ScriptModule):
def __init__(self):
super(M2, self).__init__()
self.g = torch.jit.trace(
TestScript.StarTestSumAndReturnThree(),
(torch.ones(4, 3), torch.ones(4, 3), torch.ones(4, 3)),
_force_outplace=False)
self.define('''
def forward(self, rep):
*head, tail = self.g(rep, rep, rep)
return tail
''')
m = M2()
# since forward() makes three aliases to the input `rep` before passing
# it to StarTestSumAndReturnThree(), in-place behavior will be different
# than the above out of place.
self.assertEqual(m(torch.ones(4, 3)), 4 * torch.ones(4, 3))
def test_script_module_star_assign_fail_pythonop(self):
with self.assertRaisesRegex(RuntimeError, "cannot be used as a tuple"):
class M2(torch.jit.ScriptModule):
def __init__(self):
super(M2, self).__init__()
@torch.jit.ignore
def myfunc():
return torch.zeros(1, 2, 3), torch.zeros(1, 2, 3)
self.define('''
def forward(self, rep):
a, *b = myfunc()
return a
''')
m = M2()
m(torch.zeros(4, 3))
def test_script_module_star_assign_fail_builtin(self):
with self.assertRaisesRegex(RuntimeError, "cannot be used as a tuple"):
class M2(torch.jit.ScriptModule):
def __init__(self):
super(M2, self).__init__()
self.define('''
def forward(self, rep):
a, *b = torch.neg(rep)
return a
''')
m = M2()
m(torch.zeros(4, 3))
@skipIfCompiledWithoutNumpy
def test_pack_padded_pad_packed_trace(self):
from torch.nn.utils.rnn import pack_padded_sequence, pad_packed_sequence
T, B, C = 3, 5, 7
class PadPackedWrapper(torch.nn.Module):
def __init__(self):
super(PadPackedWrapper, self).__init__()
def forward(self, x, seq_lens):
x = pack_padded_sequence(x, seq_lens)
x, _ = pad_packed_sequence(x)
return x
x = np.ones((T, B, C))
seq_lens = np.array([3, 3, 2, 2, 1], dtype=np.int32)
# set padding value so we can test equivalence
for b in range(B):
if seq_lens[b] < T:
x[seq_lens[b]:, b, :] = 0
seq_lens = torch.from_numpy(seq_lens)
x = torch.autograd.Variable(torch.from_numpy(x), requires_grad=True)
m = PadPackedWrapper()
m_traced = torch.jit.trace(m, (x, seq_lens,))
y = m(x, seq_lens)
loss = torch.sum(y)
loss.backward()
grad = x.grad.clone()
x.grad.zero_()
y_traced = m_traced(x, seq_lens)
loss_traced = torch.sum(y_traced)
loss_traced.backward()
grad_traced = x.grad.clone()
self.assertEqual(y_traced, x)
self.assertEqual(y_traced, y)
self.assertEqual(grad, grad_traced)
f = io.BytesIO()
torch.onnx._export(m, (x, seq_lens), f, verbose=False)
def test_script_pack_padded_sequence(self):
from torch.nn.utils.rnn import pack_padded_sequence, pad_packed_sequence
def pack_padded_pad_packed_script(x, seq_lens):
x = pack_padded_sequence(x, seq_lens)
x, lengths = pad_packed_sequence(x)
return x, lengths
T, B, C = 3, 5, 7
x = torch.ones((T, B, C))
seq_lens = torch.tensor([3, 3, 2, 2, 1])
# set padding value so we can test equivalence
for b in range(B):
if seq_lens[b] < T:
x[seq_lens[b]:, b, :] = 0
eager_seq, eager_lengths = pack_padded_pad_packed_script(x, seq_lens)
with torch._jit_internal._disable_emit_hooks():
scripted_pack_padded_seq = torch.jit.script(pack_padded_pad_packed_script)
script_seq, script_lengths = scripted_pack_padded_seq(x, seq_lens)
self.assertEqual(eager_seq, script_seq)
self.assertEqual(eager_lengths, script_lengths)
class ExperimentalLSTM(torch.nn.Module):
def __init__(self, input_dim, hidden_dim):
super().__init__()
def forward(self, input):
# type: (Tensor)
packed = pack_padded_sequence(
input=input, lengths=torch.tensor([1, 2]), enforce_sorted=False
)
output, lengths = pad_packed_sequence(
sequence=packed, total_length=2
)
# lengths is flipped, so is output
return output[0]
lstm = ExperimentalLSTM(input_dim=2, hidden_dim=2)
with torch._jit_internal._disable_emit_hooks():
self.checkModule(lstm, [torch.ones(2, 2)])
def test_script_pad_sequence_pack_sequence(self):
from torch.nn.utils.rnn import pad_sequence, pack_sequence, pad_packed_sequence
def pad_sequence_func(tensor_list, batch_first=False, padding_value=0.0):
# type: (List[Tensor], bool, float) -> Tensor
return pad_sequence(tensor_list, batch_first, padding_value)
def pack_sequence_func(tensor_list, enforce_sorted=True):
# type: (List[Tensor], bool) -> Tensor
return pad_packed_sequence(pack_sequence(tensor_list, enforce_sorted))[0]
ones3 = torch.ones(3, 5)
ones4 = torch.ones(4, 5)
ones5 = torch.ones(5, 5)
tensor1 = torch.tensor([1, 2, 3])
tensor2 = torch.tensor([4, 5])
tensor3 = torch.tensor([6])
with torch._jit_internal._disable_emit_hooks():
self.checkScript(pad_sequence_func,
([ones3, ones4, ones5],))
self.checkScript(pad_sequence_func,
([ones3, ones4, ones5], True))
self.checkScript(pad_sequence_func,
([ones3, ones4, ones5], True, 2.5))
self.checkScript(pack_sequence_func,
([tensor1, tensor2, tensor3],))
self.checkScript(pack_sequence_func,
([tensor1, tensor2, tensor3], False))
def test_script_get_tracing_state(self):
def test_if_tracing(x):
if torch._C._get_tracing_state():
return x + 1
else:
return x - 1
inp = torch.randn(3, 3)
self.checkScript(test_if_tracing, (inp,))
def test_script_is_tracing(self):
def test_is_tracing(x):
if torch.jit.is_tracing():
return x + 1
else:
return x - 1
inp = torch.randn(3, 3)
self.checkScript(test_is_tracing, (inp,))
def test_is_scripting(self):
def foo():
return torch.jit.is_scripting()
self.assertFalse(foo())
scripted = torch.jit.script(foo)
self.assertTrue(scripted())
def test_comment_ignore_indent(self):
class Model(torch.nn.Module):
def __init__(self):
# useless comment that is not indented correctly # noqa: E115
super().__init__()
def forward(self):
return 5
# should compile without an error
self.checkModule(Model(), ())
def test_script_outputs(self):
with self.assertRaisesRegex(RuntimeError, "cannot be used as a tuple"):
@torch.jit.script
def foo(a):
c, d = a + a
return c + d
@torch.jit.script
def return3():
return 1, 2, 3
with self.assertRaisesRegex(RuntimeError, "too many values to unpack"):
@torch.jit.script
def bind2():
a, b = return3()
print(a)
print(b)
@unittest.skipIf(not RUN_CUDA, "requires CUDA")
def test_script_get_device_cuda(self):
@torch.jit.script
def foo(a):
return a.get_device()
v = torch.randn(1, device='cuda')
self.assertEqual(foo(v), 0)
def test_script_chunk(self):
@torch.jit.script
def foo(a):
b, c = torch.chunk(a, dim=0, chunks=2)
return b
v = torch.rand(10, 3)
self.assertEqual(torch.chunk(v, dim=0, chunks=2)[0], foo(v))
def test_script_copy(self):
class M(torch.nn.Module):
__annotations__ = {
"val": Optional[torch.Tensor]
}
def __init__(self):
super(M, self).__init__()
self.val = None
def some_method(self):
return 3
def forward(self, x):
# type: (Tensor) -> Tensor
self.val = x + self.some_method()
return x
m = torch.jit.script(M())
# test copy
copy.copy(m)
copy.deepcopy(m)
def test_script_forward_method_replacement(self):
# We want to support the use case of attaching a different `forward` method
class LowLevelModule(torch.nn.Module):
def __init__(self):
super(LowLevelModule, self).__init__()
def forward(self, input: torch.Tensor):
# Generic forward dispatch
return self.forward_pytorch(input) * 2
class TestModule(LowLevelModule):
def __init__(self):
super(TestModule, self).__init__()
# Replace the forward method
self.forward = types.MethodType(LowLevelModule.forward, self)
def forward_pytorch(self, input: torch.Tensor):
return torch.tensor(123)
def forward(self, input: torch.Tensor):
# Should not use this forward method
raise AssertionError("This method should not be used")
return self.forward_pytorch(input)
m = TestModule()
self.assertEqual(m(torch.tensor(1)), torch.tensor(246))
m_scripted = torch.jit.script(m)
self.assertEqual(m_scripted(torch.tensor(1)), torch.tensor(246))
# Suppression: ONNX warns when exporting RNNs because of potential batch size mismatch.
@suppress_warnings
@skipIfCompiledWithoutNumpy
def test_rnn_trace_override(self):
from torch.nn.utils.rnn import pack_padded_sequence, pad_packed_sequence
num_layers = 3
T, B, C = 11, 5, 7
class RNNTraceWrapper(torch.nn.Module):
def __init__(self, cell_type):
super(RNNTraceWrapper, self).__init__()
if cell_type == 'RNN':
self.rnn = torch.nn.RNN(input_size=C, hidden_size=C, num_layers=num_layers)
elif cell_type == 'LSTM':
self.rnn = torch.nn.LSTM(input_size=C, hidden_size=C, num_layers=num_layers)
elif cell_type == 'GRU':
self.rnn = torch.nn.GRU(input_size=C, hidden_size=C, num_layers=num_layers)
def forward(self, x, seq_lens):
x = pack_padded_sequence(x, seq_lens)
x, _ = self.rnn(x)
x, _ = pad_packed_sequence(x)
return x
for cell_type in ['RNN', 'LSTM', 'GRU']:
x = torch.ones(T, B, C, requires_grad=True)
seq_lens = torch.from_numpy(np.array([11, 3, 2, 2, 1], dtype=np.int32))
m = RNNTraceWrapper(cell_type)
m_traced = torch.jit.trace(m, (x, seq_lens,))
y = m(x, seq_lens)
loss = torch.sum(y)
loss.backward()
grad = x.grad.clone()
x.grad.zero_()
y_traced = m_traced(x, seq_lens)
loss_traced = torch.sum(y_traced)
loss_traced.backward()
grad_traced = x.grad.clone()
self.assertEqual(y_traced, y)
self.assertEqual(grad, grad_traced)
f = io.BytesIO()
torch.onnx._export(m, (x, seq_lens), f, verbose=False)
def test_python_call_non_tensor(self):
def foo(a, b, c):
# type: (Tensor, int, Tuple[Tensor, int]) -> Tuple[int, Tensor]
d, e = c
return b + e, a + d
@torch.jit.script
def bar():
x = torch.ones(3, 4)
a, b = foo(x, 3, (x, 3))
return a, b
self.assertEqual((6, torch.ones(3, 4) + 1), bar())
def test_python_call_non_tensor_wrong(self):
with self.assertRaisesRegex(RuntimeError, r"but instead got value of type tuple"):
@torch.jit.ignore
def foo():
# type: () -> Tensor
return ((3, 4),) # noqa: T484
@torch.jit.script
def bar():
return foo()
bar()
def test_if_different_type(self):
with self.assertRaisesRegex(RuntimeError, "c0 is set to type "
"int in the true branch and type "
"float in the false branch"):
@torch.jit.script
def diff_type_used():
if 1 == 2:
c0 = 1
else:
c0 = 1.0
return c0
with self.assertRaisesRegex(RuntimeError, "Variable 'c0' previously had type float"):
@torch.jit.script
def diff_existing_type(x):
c0 = 1.0
if 1 == 2:
c0 = 1
print(x)
return x
@torch.jit.script
def diff_type_unused():
if 1 == 1:
c0 = 1
print(c0)
else:
c0 = 1.0
print(c0)
return 1
def test_if_not_defined_error(self):
with self.assertRaisesRegex(RuntimeError, "c0 is not defined in the false branch"):
@torch.jit.script
def test():
if 1 == 1:
c0 = 1
return c0
with self.assertRaisesRegex(RuntimeError, "c0 is not defined in the true branch"):
@torch.jit.script
def test2():
if 1 == 1:
pass
else:
c0 = 1
return c0
def test_if_list_cat(self):
# testing that different length lists don't throw error on cat in shape prop
@torch.jit.script
def test_list(x):
if bool(x.sum() < 1):
c = [x, x]
else:
c = [x, x, x]
return torch.cat(c)
b = torch.zeros(2, 4)
_propagate_shapes(test_list.graph, (b,), False)
def test_if_supertype(self):
@torch.jit.script
def tensor_unifying(x, y, z):
# testing dynamic is appropriately set for y and z
if bool(x):
x, y, z = x + 1, y, z
else:
x, y, z = x + 1, x, y
return x, y, z
a = torch.zeros(2, 2, dtype=torch.float)
b = torch.zeros(2, 4, dtype=torch.long)
c = torch.zeros(2, 4, dtype=torch.float)
graph = _propagate_shapes(tensor_unifying.graph, (a, b, c), False)
if_outputs = list(graph.findNode("prim::If").outputs())
self.assertTrue(if_outputs[0].type().str() == "Float(*, *, requires_grad=0, device=cpu)")
self.assertTrue(if_outputs[1].type().str() == "Tensor(*, *, requires_grad=0, device=cpu)")
self.assertTrue(if_outputs[2].type().str() == "Tensor(*, *, requires_grad=0, device=cpu)")
def test_list_unify(self):
# allowing a unififed int?[] would cause a runtime error b/c
# the index operation expects int?[] to be a generic list,
# but in the true branch the IValue will be a int list
with self.assertRaisesRegex(RuntimeError, "int[] in the true branch and type None[]"):
@torch.jit.script
def list_optional_fails(x):
# type: (bool) -> Optional[int]
if x:
y = [1]
else:
y = [None] # noqa: T484
return y[0]
@torch.jit.script
def list_tensors(x):
# type: (bool) -> Tuple[Tensor, List[Tensor]]
if x:
a = torch.zeros([1, 1])
y = [a]
else:
a = torch.zeros([1, 2])
y = [a]
return a, y
self.run_pass('constant_propagation', list_tensors.graph)
m = self.createFunctionFromGraph(list_tensors.graph)
# testing that tensor type of lists is unified
self.getExportImportCopy(m)
@_inline_everything
def test_import_constants_not_specialized(self):
class Mod(torch.nn.Module):
def forward(self, x):
return torch.cat(2 * [x], dim=0)
class ScriptMod(torch.jit.ScriptModule):
def __init__(self, mod):
super(ScriptMod, self).__init__()
x = torch.zeros(1, 3)
mod_fn = lambda : mod(x) # noqa: E731
self.mod = torch.jit.trace(mod_fn, tuple())
@torch.jit.script_method
def forward(self):
return self.mod()
cm = ScriptMod(Mod())
# specialized tensor in graph
FileCheck().check("Double(1, 3, strides=[3, 1], requires_grad=0, device=cpu)").run(cm.forward.graph)
buffer = io.BytesIO()
torch.jit.save(cm, buffer)
buffer.seek(0)
# when tensor is loaded as constant it isnt specialized
cm_load = torch.jit.load(buffer)
FileCheck().check_not("Double(1, 3)").run(cm_load.forward.graph)
@skipIfTorchDynamo("TorchDynamo fails with unknown reason")
def test_type_annotations_repeated_list(self):
@torch.jit.script
def float_fn(x, y):
# type: (float, BroadcastingList3[float]) -> List[float]
return y
self.assertEqual(float_fn(2.0, 1.0), float_fn(2.0, [1.0, 1.0, 1.0]))
self.assertEqual(float_fn(2.0, 1.0), float_fn(2.0, (1.0, 1.0, 1.0)))
@torch.jit.script
def float_fn_call():
print(float_fn(1.0, 1.0))
print(float_fn(1.0, (1.0, 1.0, 1.0)))
@torch.jit.script
def int_fn(x):
# type: (BroadcastingList3[int]) -> List[int]
return x
self.assertEqual(int_fn(1), int_fn([1, 1, 1]))
self.assertEqual(int_fn(1), int_fn((1, 1, 1)))
@torch.jit.script
def int_fn_call():
print(int_fn(1))
print(int_fn((1, 1, 1)))
with self.assertRaisesRegex(RuntimeError, "must be a positive integer:"):
@torch.jit.script # noqa: T484
def fn(x):
# type: (BroadcastingListx[int]) -> List[int] # noqa: T484
return x
# using CU so that flake8 error on int[2] is not raised (noqa not working)
with self.assertRaisesRegex(RuntimeError, "Unknown type constructor"):
cu = torch.jit.CompilationUnit('''
def nested(x, y):
# type: (int, Tuple[int, int[2]]) -> List[int]
return x # noqa: T484
''')
@torch.jit.script
def f(x: BroadcastingList2[int]):
return x
out = f(1)
self.assertTrue(isinstance(out[0], int))
self.assertEqual(out, [1, 1])
def test_ntuple_builtins(self):
from torch.nn.modules.utils import _single, _pair, _triple, _quadruple
def test_ints():
return _single(1), _pair(2), _triple(3), _quadruple(4)
def test_floats():
return _single(1), _pair(2.1), _triple(3.1), _quadruple(4.1)
self.checkScript(test_ints, ())
self.checkScript(test_floats, ())
def test_embedding_renorm_grad_error(self):
# Testing that the builtin call to embedding_renorm_ correctly throws
# Error when .backward() is called on its input
def embedding_norm(input, embedding_matrix, max_norm):
F.embedding(input, embedding_matrix, max_norm=0.01)
@torch.jit.script
def embedding_norm_script(input, embedding_matrix, max_norm):
# type: (Tensor, Tensor, float) -> None
F.embedding(input, embedding_matrix, max_norm=0.01)
for _ in [embedding_norm, embedding_norm_script]:
input = torch.tensor([[1, 2, 4, 5], [4, 3, 2, 9]])
embedding_matrix = torch.randn(10, 3)
var1 = torch.randn(10, 3, requires_grad=True)
var2 = var1.detach().requires_grad_()
output1 = var1 * embedding_matrix
output2 = var2 * embedding_matrix
output1.sum().backward()
ignore = F.embedding(input, embedding_matrix, max_norm=0.01)
with self.assertRaisesRegex(RuntimeError, "modified"):
output2.sum().backward()
def test_type_annotations(self):
def fn(x, y):
# type: (Tensor, Tensor) -> Tuple[Tensor, Tensor, Tensor]
return x, x * 2, x * 3
with self.assertRaisesRegex(RuntimeError, r"need 4 values .* found only 3"):
@torch.jit.script
def script_fn(x):
x, y, z, w = fn(x, x)
with self.assertRaisesRegex(RuntimeError, r"too many values .* need 2 but found 3"):
@torch.jit.script
def script_fn2(x):
x, y = fn(x, x)
def fn_unpack(x):
y, z, w = fn(x, x)
return y
def fn_index(x):
q = fn(x, x)
return x
def fn_string(str, strpair):
# type: (str, Tuple[str, str]) -> Tuple[str, int, str, str]
str1, str2 = strpair
return str, 2, str1, str2
x = torch.ones(2, 2)
self.checkScript(fn_unpack, (x,), optimize=True)
self.checkScript(fn_index, (x,), optimize=True)
self.checkScript(fn_string, ("1", ("3", "4")), optimize=True)
def test_type_annotations_varargs(self):
@torch.jit.ignore
def fn_varargs(x, *args):
return args[0] if args else x
def fn1(x, y, z):
return fn_varargs(x)
def fn2(x, y, z):
return fn_varargs(x, y)
def fn3(x, y, z):
return fn_varargs(x, y, z)
x, y, z = [torch.randn(2, 2) for _ in range(3)]
self.checkScript(fn1, (x, y, z), optimize=True)
self.checkScript(fn2, (x, y, z), optimize=True)
self.checkScript(fn3, (x, y, z), optimize=True)
def test_type_annotation_py3(self):
code = dedent("""
import torch
from torch import Tensor
from typing import Tuple
def fn(x : torch.Tensor, y : Tensor, z) -> Tuple[Tensor, Tensor, Tensor]:
return (x, y + z, z)
""")
with tempfile.TemporaryDirectory() as tmp_dir:
script_path = os.path.join(tmp_dir, 'script.py')
with open(script_path, 'w') as f:
f.write(code)
fn = get_fn('test_type_annotation_py3', script_path)
fn = torch.jit.ignore(fn)
with self.assertRaisesRegex(RuntimeError, r"Expected a value of type 'Tensor' for argument"
r" 'x' but instead found type 'Tuple\[Tensor,"):
@torch.jit.script
def bad_fn(x):
x, y = fn((x, x), x, x)
return y
with self.assertRaisesRegex(RuntimeError, r"too many values .* need 2 but found 3"):
@torch.jit.script
def bad_fn2(x):
x, y = fn(x, x, x)
return y
with self.assertRaisesRegex(RuntimeError, r"need 4 values .* found only 3"):
@torch.jit.script
def bad_fn3(x):
x, y, z, w = fn(x, x, x)
return y
def good_fn(x):
y, z, w = fn(x, x, x)
return y, z, w
self.checkScript(good_fn, (torch.ones(2, 2),), optimize=True)
def test_type_annotation_module(self):
class BaseModule(torch.jit.ScriptModule):
@torch.jit.ignore
def foo(self, x):
# type: (Tensor) -> Tensor
return x + 1
@torch.jit.ignore
def bar(self, x, y):
# type: (Tensor, Tensor) -> Tuple[Tensor, Tensor]
return x + y, y
@torch.jit.ignore
def baz(self, x, y):
return x
class ModuleTooMany(BaseModule):
@torch.jit.script_method
def method(self, x):
return self.foo(x, x)
class ModuleTooFew(BaseModule):
@torch.jit.script_method
def method(self, x):
return self.bar(x)
class ModuleTooManyAssign(BaseModule):
@torch.jit.script_method
def method(self, x):
y, z, w = self.bar(x, x)
return x
class ModuleDefault(BaseModule):
@torch.jit.script_method
def method(self, x):
y = self.baz(x)
return x
with self.assertRaisesRegex(RuntimeError, "Expected at most 2 arguments but found 3"):
ModuleTooMany()
with self.assertRaisesRegex(RuntimeError, "Argument y not provided"):
ModuleTooFew()
with self.assertRaisesRegex(RuntimeError, "need 3 values .* found only 2"):
ModuleTooManyAssign()
with self.assertRaisesRegex(RuntimeError, "Argument y not provided."):
ModuleDefault()
def test_type_inferred_from_empty_annotation(self):
"""
Test that the type inferred from an empty or missing annotation is Torch.Tensor wtih `inferred=true`
"""
@torch.jit.script
def fn(x):
return x
graph = fn.graph
n = next(graph.inputs())
self.assertTrue(n.type() == torch._C.TensorType.getInferred())
with self.assertRaisesRegex(RuntimeError, "Inferred \'x\' to be of type \'Tensor"):
fn("1")
def test_script_define_order(self):
class M(torch.jit.ScriptModule):
@torch.jit.script_method
def call_foo(self, input):
return self.foo(input)
@torch.jit.script_method
def foo(self, input):
return input + 1
m = M()
self.assertEqual(2, m.call_foo(torch.ones((), dtype=torch.int64)))
def test_script_define_order_recursive_fail(self):
class M(torch.jit.ScriptModule):
@torch.jit.script_method
def call_foo(self, input):
return self.foo(input)
@torch.jit.script_method
def foo(self, input):
self.call_foo(input)
with self.assertRaisesRegex(RuntimeError, 'called recursively'):
M()
def test_script_kwargs_fn_call(self):
class M(torch.jit.ScriptModule):
@torch.jit.script_method
def call_foo(self, input):
return self.foo(input=input, bar=1)
@torch.jit.script_method
def foo(self, bar, input):
# type: (int, Tensor) -> Tensor
return input + bar
m = M()
self.assertEqual(2, m.call_foo(torch.ones((), dtype=torch.int64)))
def test_if_define(self):
@torch.jit.script
def foo(a):
if bool(a == 0):
b = 1
else:
b = 0
return b + 1
@torch.jit.script
def foo2(a):
b = 0
if bool(a == 0):
b = 1
return b + 1
@torch.jit.script
def foo3(a):
b = 1
if bool(a == 0):
c = 4
else:
b = 0
return b + 1
a = torch.ones(1, dtype=torch.long)
b = torch.zeros(1, dtype=torch.long)
self.assertEqual(1, foo(a))
self.assertEqual(2, foo(b))
self.assertEqual(1, foo2(a))
self.assertEqual(2, foo2(b))
self.assertEqual(1, foo3(a))
self.assertEqual(2, foo3(b))
def test_script_module_export_submodule(self):
class M1(torch.jit.ScriptModule):
def __init__(self):
super(M1, self).__init__()
self.weight = nn.Parameter(torch.randn(2))
@torch.jit.script_method
def forward(self, thing):
return self.weight + thing
class M2(torch.jit.ScriptModule):
def __init__(self):
super(M2, self).__init__()
# test submodule
self.sub = M1()
self.weight = nn.Parameter(torch.randn(2, 3))
self.bias = nn.Parameter(torch.randn(2))
self.define("""
def hi(self, a):
return self.weight.mm(a)
""")
@torch.jit.script_method
def doit(self, input):
return self.weight.mm(input)
@torch.jit.script_method
def doit2(self, input):
return self.weight.mm(input)
@torch.jit.script_method
def doit3(self, input):
return input + torch.ones([1], dtype=torch.double)
@torch.jit.script_method
def forward(self, input):
a = self.doit(input)
b = self.doit2(input)
c = self.hi(input)
return a + b + self.bias + c
with torch.jit.optimized_execution(False):
m_orig = M2()
m_import = self.getExportImportCopy(m_orig)
input = torch.randn(3, 2)
self.assertEqual(m_orig.doit(input), m_import.doit(input))
self.assertEqual(m_orig.hi(input), m_import.hi(input))
self.assertEqual(m_orig.doit3(input), m_import.doit3(input))
self.assertEqual(m_orig.forward(input), m_import.forward(input))
@slowTest
def test_compile_module_with_constant(self):
class Double(nn.Module):
def __init__(self, downsample=None):
super(Double, self).__init__()
def forward(self, input):
return input * 2
class Mod(nn.Module):
__constants__ = ['downsample']
def __init__(self, downsample=None):
super(Mod, self).__init__()
self.downsample = downsample
def forward(self, input):
if self.downsample is not None:
return self.downsample(input)
return input
none_mod = torch.jit.script(Mod(None))
double_mod = torch.jit.script(Mod(Double()))
self.assertEqual(none_mod(torch.tensor(1)), torch.tensor(1))
self.assertEqual(double_mod(torch.tensor(1)), torch.tensor(1) * 2)
def test_device_kwarg(self):
from torch import device
def f():
return device(type='cuda'), torch.device(type='cpu')
self.checkScript(f, ())
def test_script_module_export_tensor_type(self):
class M(torch.jit.ScriptModule):
def __init__(self, type):
super(M, self).__init__()
self.param = torch.nn.Parameter(torch.zeros((5, 5), dtype=type).random_())
@torch.jit.script_method
def foo(self):
return self.param
with torch.jit.optimized_execution(False):
for type in [torch.float, torch.double]:
m_orig = M(type)
m_import = self.getExportImportCopy(m_orig)
# check to make sure the storage wasn't resized
self.assertTrue(m_orig.param.storage().size() == 25)
self.assertEqual(m_orig.foo(), m_import.foo())
self.assertTrue(m_orig.foo().dtype == m_import.foo().dtype)
@unittest.skipIf(not RUN_CUDA, "testing cuda tensors require CUDA")
def test_script_module_export_tensor_cuda(self):
class M(torch.jit.ScriptModule):
def __init__(self):
super(M, self).__init__()
self.param = torch.nn.Parameter(torch.zeros((5, 5), device='cuda:0').random_())
@torch.jit.script_method
def foo(self):
return self.param
m_orig = M()
m_import = self.getExportImportCopy(m_orig)
# check to make sure the storage wasn't resized
self.assertTrue(m_orig.param.storage().size() == 25)
self.assertTrue(m_import.foo().device == torch.device('cuda:0'))
self.assertEqual(m_orig.foo(), m_import.foo())
self.assertTrue(m_orig.foo().dtype == m_import.foo().dtype)
def test_script_module_export_blocks(self):
class M(torch.jit.ScriptModule):
def __init__(self, n, m):
super(M, self).__init__()
self.weight = torch.nn.Parameter(torch.rand(n, m))
@torch.jit.script_method
def forward(self, input):
if bool(input.sum() > 0):
output = self.weight.mv(input)
else:
output = self.weight + input
return output
m_orig = M(200, 200)
m_import = self.getExportImportCopy(m_orig)
t = torch.rand(200)
self.assertEqual(m_orig(t), m_import(t))
def test_script_module_export_shared_storage(self):
class M(torch.jit.ScriptModule):
def __init__(self):
super(M, self).__init__()
self.param1 = torch.nn.Parameter(torch.rand(5, 5))
self.param2 = torch.nn.Parameter(self.param1[3])
self.param3 = torch.nn.Parameter(torch.rand(5, 5))
self.param4 = torch.nn.Parameter(torch.rand(11, 5)[1:6])
@torch.jit.script_method
def foo(self):
return self.param1 + self.param2 + self.param3 + self.param4
with torch.jit.optimized_execution(False):
m_orig = M()
m_import = self.getExportImportCopy(m_orig)
self.assertEqual(m_orig.foo(), m_import.foo())
self.assertTrue(m_import.param1.storage().data_ptr() == m_import.param2.storage().data_ptr())
self.assertTrue(m_import.param1.storage().data_ptr() != m_import.param3.storage().data_ptr())
def test_sequential_intermediary_types(self):
class A(torch.nn.Module):
def __init__(self):
super(A, self).__init__()
def forward(self, x):
return x + 3
class B(torch.nn.Module):
def __init__(self):
super(B, self).__init__()
def forward(self, x):
return {"1": x}
class C(torch.nn.Module):
def __init__(self):
super(C, self).__init__()
self.foo = torch.nn.Sequential(A(), B())
def forward(self, x):
return self.foo(x)
self.checkModule(C(), (torch.tensor(1),))
def test_ellipsis_const_mid(self):
def ellipsize(x):
# type: (Tensor) -> List[int]
return x[2, Ellipsis, 0:4, 4:8].size()
dummy = torch.zeros(8, 8, 8, 8, 8)
self.checkScript(ellipsize, (dummy,), optimize=True)
def test_ellipsis_const_mid_select(self):
def ellipsize(x):
# type: (Tensor) -> List[int]
return x[2, Ellipsis, 4, 4, 4:8, 2].size()
dummy = torch.zeros(8, 8, 8, 8, 8, 8, 8)
self.checkScript(ellipsize, (dummy,), optimize=True)
def test_ellipsis_const_start(self):
def ellipsize(x):
# type: (Tensor) -> List[int]
return x[Ellipsis, 0:4, 4:8].size()
dummy = torch.zeros(8, 8, 8, 8, 8)
self.checkScript(ellipsize, (dummy,), optimize=True)
def test_ellipsis_const_end(self):
def ellipsize(x):
# type: (Tensor) -> List[int]
return x[0:4, 2, Ellipsis].size()
dummy = torch.zeros(8, 8, 8, 8, 8)
self.checkScript(ellipsize, (dummy,), optimize=True)
def test_ellipsis_mid(self):
def ellipsize(x):
# type: (Tensor) -> List[int]
return x[2, ..., 0:4, 4:8].size()
dummy = torch.zeros(8, 8, 8, 8, 8)
self.checkScript(ellipsize, (dummy,), optimize=True)
def test_ellipsis_mid_select(self):
def ellipsize(x):
# type: (Tensor) -> List[int]
return x[2, ..., 4, 4, 4:8, 2].size()
dummy = torch.zeros(8, 8, 8, 8, 8, 8, 8)
self.checkScript(ellipsize, (dummy,), optimize=True)
def test_ellipsis_start(self):
def ellipsize(x):
# type: (Tensor) -> List[int]
return x[..., 0:4, 4:8].size()
dummy = torch.zeros(8, 8, 8, 8, 8)
self.checkScript(ellipsize, (dummy,), optimize=True)
def test_ellipsis_end(self):
def ellipsize(x):
# type: (Tensor) -> List[int]
return x[0:4, 2, ...].size()
dummy = torch.zeros(8, 8, 8, 8, 8)
self.checkScript(ellipsize, (dummy,), optimize=True)
def test_torch_manual_seed(self):
with freeze_rng_state():
def test():
torch.manual_seed(2)
return torch.rand(1)
script = torch.jit.script(test)
self.assertEqual(test(), script())
graph = script.graph_for()
FileCheck().check("aten::manual_seed").run(graph)
def test_index_select_shape_prop(self):
@torch.jit.script
def foo(x, y):
return torch.index_select(x, index=y, dim=1)
a = torch.zeros(2, 2)
b = torch.zeros(4, dtype=torch.long)
torch._C._jit_pass_complete_shape_analysis(foo.graph, (a, b), False)
FileCheck().check("Double(2, 4, strides=[4, 1], requires_grad=0, device=cpu)").run(str(foo.graph))
def test_shape_analysis_loop(self):
def foo(a, b, x):
c = a
# on the first iteration of the loop it appears that
# c should have a expand to the size of b
# but on the second+ iterations, there is no broadcast and the
# sizes are different.
# previously this would cause the compiler to (1) enter an infinite
# loop trying to compute the shape, and (2) insert invalid
# broadcasts.
# this test ensure we don't regress on these issues
for _ in range(2):
a = c + b
c = x
b = x
return a
self.checkScript(foo, (torch.zeros(1), torch.zeros(4), torch.zeros(5)), optimize=False)
def test_intlist_args(self):
def func_1(x):
return torch.nn.functional.adaptive_avg_pool1d(x, 1)
def func_2(x):
return torch.nn.functional.adaptive_avg_pool1d(x, output_size=1)
def func_3(x):
return torch.nn.functional.adaptive_avg_pool1d(x, output_size=[1])
x = torch.randn(8, 8, 8)
self.checkScript(func_1, [x], optimize=True)
self.checkScript(func_2, [x], optimize=True)
self.checkScript(func_3, [x], optimize=True)
def test_wrong_implicit_expand(self):
@_trace(torch.zeros(3), torch.zeros(1))
def foo(a, b):
return a + b
a = torch.rand(4)
b = torch.rand(4)
self.assertEqual(a + b, foo(a, b))
def test_builtin_args_fails(self):
with self.assertRaisesRegex(RuntimeError, 'Argument self not provided'):
@torch.jit.script
def f1(a):
torch.sum(foo=4)
with self.assertRaisesRegex(RuntimeError, 'specified twice'):
@torch.jit.script
def f2(a):
torch.sum(a, self=a)
with self.assertRaisesRegex(RuntimeError, 'not provided'):
@torch.jit.script
def f3(a):
torch.sum(dim=4)
with self.assertRaisesRegex(RuntimeError, 'for argument \'tensors\' but instead found type \'Tensor'):
@torch.jit.script
def f4(a):
torch.cat(a)
with self.assertRaisesRegex(RuntimeError, r'argument \'tensors\' but instead found type \'List\[int\]'):
@torch.jit.script
def f5(a):
torch.cat([3])
with self.assertRaisesRegex(RuntimeError, r'Expected a value of'
r' type \'List\[int\]\' for argument'
r' \'size\' but instead found type '
r'\'List\[Union\[List\[int\], int\]\]'):
@torch.jit.script
def f6(a):
a.expand(size=[3, [4]])
def test_builtin_args(self):
def t0(a):
# default arg dim
return torch.cat([a, a])
self.checkScript(t0, (torch.zeros(1, 1),))
def t1(a):
# keywords out of order
return torch.cat(dim=1, tensors=[a, a])
self.checkScript(t1, (torch.zeros(1, 1, 2),))
def t2(a):
# mix const/non-const attributes
if 1 == 1:
b = 1
else:
b = 0
return torch.sum(a, dim=b, keepdim=False)
self.checkScript(t2, (torch.zeros(1, 1, 2),))
def test_parser_type_annotations(self):
cu = torch.jit.CompilationUnit('''
def foo(x : Tensor, y : Tuple[Tuple[Tensor, Tensor], Tensor]) -> Tuple[Tensor, Tensor]:
return x, x
''')
self.assertExpected(str(cu.foo.schema))
def test_parser_type_annotations_comment(self):
cu = torch.jit.CompilationUnit('''
def foo(x, y):
# type: (Tensor, Tuple[Tuple[Tensor, Tensor], Tensor]) -> Tuple[Tensor, Tensor]
return x, x
''')
self.assertExpected(str(cu.foo.schema))
def test_parser_type_annotations_unknown_type(self):
with self.assertRaisesRegex(RuntimeError, "Unknown type name 'Foo'"):
cu = torch.jit.CompilationUnit('''
def foo(x : Tensor, y : Tuple[Tuple[Foo, Tensor], Tensor]) -> Tuple[Tensor, Tensor]:
return x, x
''')
def test_parser_type_annotations_subscript_non_ident(self):
with self.assertRaisesRegex(RuntimeError, r'Subscripted type must be a type identifier'):
cu = torch.jit.CompilationUnit('''
def foo(x : Tensor, y : Tuple[Tensor, Tensor][Tensor]) -> Tuple[Tensor, Tensor]:
return x, x
''')
def test_parser_type_annotations_subscript_tensor(self):
with self.assertRaisesRegex(RuntimeError, r'Unknown type constructor Tensor'):
cu = torch.jit.CompilationUnit('''
def foo(x : Tensor, y : Tensor[Tensor, Tensor]) -> Tuple[Tensor, Tensor]:
return x, x
''')
def test_parser_type_annotations_incompatible_expression(self):
with self.assertRaisesRegex(RuntimeError, r'Expression of type \+ cannot be used in a type expression'):
cu = torch.jit.CompilationUnit('''
def foo(x : Tensor, y : Tuple[3 + 4, Tensor]) -> Tuple[Tensor, Tensor]:
return x, x
''')
def test_gather_dynamic_index(self):
def t(x):
gather1 = x[0]
idx = 0 + 1
gather2 = x[idx]
return gather1 + gather2
self.checkScript(t, (torch.zeros(3, 2, 3),))
def test_torch_ignore_conversion_to_none(self):
class A(torch.nn.Module):
def __init__(self):
super(A, self).__init__()
@torch.jit.ignore
def ignored(self, a: int) -> None:
l: int = len([2 for i in range(a) if i > 2])
return
def forward(self) -> int:
a: int = 4
b: int = 5
self.ignored(a)
return a + b
class B(torch.nn.Module):
def __init__(self):
super(B, self).__init__()
@torch.jit.ignore
def ignored(self, a: int):
l: int = len([2 for i in range(a) if i > 2])
return
def forward(self) -> int:
a: int = 4
b: int = 5
self.ignored(a)
return a + b
modelA = torch.jit.script(A())
self.assertEqual(modelA(), 9)
modelB = torch.jit.script(B())
self.assertEqual(modelB(), 9)
def test_addmm_grad(self):
""" This test checks several things:
1. An expand node was inserted before the addmm operating on the
bias term.
2. The fused form of addmm appears in the ultimate graph that's
executed.
3. A sum op was emitted for accumulating gradients along the 0th
(expanded) dimension of the bias term.
4. The correct symbolic representation for the backward pass of the
mm operator was emitted (x.t() -> mm)
TODO: we should actually check these conditions once we have a way
to dump the GraphExecutor state. Namely the processed forward graph
and the backward graph.
"""
@torch.jit.script
def addmm_grad_test(b, x, w):
return torch.addmm(b, x, w)
# Initialize param and input values
w_init = torch.rand(2, 5)
b_init = torch.rand(5)
x = torch.rand(3, 2)
# Clone trainable params
b = b_init.clone()
b.requires_grad_()
w = w_init.clone()
w.requires_grad_()
# Test symbolic differentiation
y = addmm_grad_test(b, x, w)
y.sum().backward()
# clone params for autograd reference
b_ref = b_init.clone()
b_ref.requires_grad_()
w_ref = w_init.clone()
w_ref.requires_grad_()
y_ref = torch.addmm(b_ref, x, w_ref)
y_ref.sum().backward()
self.assertEqual(w.grad, w_ref.grad)
self.assertEqual(b.grad, b_ref.grad)
@unittest.skipIf(not RUN_CUDA, "running tests on cuda to verify cudnn fix")
def test_batch_norm_inference_backward_cuda(self):
with enable_profiling_mode_for_profiling_tests():
class MyBatchNorm(torch.nn.Module):
def __init__(self, num_features, affine, track_running_stats):
super(MyBatchNorm, self).__init__()
self.bn = torch.nn.BatchNorm2d(
num_features, 1e-5, affine=affine, track_running_stats=track_running_stats).float()
def forward(self, x: torch.Tensor):
o = self.bn(x)
o = torch.nn.functional.relu(o)
return o
batch = 4
c = 2
hw = 3
# Initialize param and input values
x_init = torch.randn(batch, c, hw, hw, dtype=torch.float).cuda()
grad = torch.randn(batch, c, hw, hw, dtype=torch.float).cuda()
training = False
affine = True
track_running_stats = True
module = torch.jit.script(MyBatchNorm(c, affine, track_running_stats)).cuda()
ref_module = MyBatchNorm(c, affine, track_running_stats).cuda()
module.eval()
ref_module.eval()
jit_module = torch.jit.script(module)
ref_module.load_state_dict(module.state_dict())
x = x_init.detach().clone()
x.requires_grad_()
x_ref = x_init.detach().clone()
x_ref.requires_grad_()
# Test symbolic differentiation
# Run Forward and Backward thrice to trigger autodiff graph
for i in range(0, 3):
y = jit_module(x)
y.backward(grad)
x.grad.zero_()
module.bn.running_mean.zero_()
module.bn.running_var.fill_(1.0)
ref_module.bn.running_mean.zero_()
ref_module.bn.running_var.fill_(1.0)
# run jitted module
y = jit_module(x)
y.backward(grad)
# reference computation
y_ref = ref_module(x_ref)
y_ref.backward(grad)
self.assertEqual(y_ref, y)
self.assertEqual(x.grad, x_ref.grad)
self.assertEqual(module.bn.running_mean, ref_module.bn.running_mean)
self.assertEqual(module.bn.running_var, ref_module.bn.running_var)
def test_zeros(self):
class M(torch.jit.ScriptModule):
__constants__ = ['d']
def __init__(self):
super(M, self).__init__()
self.d = torch.device('cpu')
@torch.jit.script_method
def create(self):
return torch.zeros([1, 1, 2], dtype=torch.float, device=self.d, layout=torch.strided)
r = M().create()
self.assertEqual(r.dtype, torch.float)
self.assertEqual(torch.zeros([1, 1, 2], dtype=torch.float), r)
def fn():
return torch.zeros((1, 2, 3))
self.checkScript(fn, ())
def test_vararg_zeros(self):
def foo():
return torch.zeros(3, 4, 5, dtype=torch.int)
self.checkScript(foo, ())
@unittest.skipIf(GRAPH_EXECUTOR != ProfilingMode.LEGACY, "the original version of test_rand")
def test_rand(self):
def test_rand():
a = torch.rand([3, 4])
return a + 1.0 - a
self.checkScript(test_rand, ())
fn = torch.jit.script(test_rand)
out = fn()
self.assertEqual(out.dtype, torch.double)
g = fn.graph_for()
# Testing shape analysis correctly setting type
if GRAPH_EXECUTOR != ProfilingMode.SIMPLE:
FileCheck().check("Double(*, *, requires_grad=0, device=cpu)") \
.check_not("Float(*, *, requires_grad=0, device=cpu)").run(g)
@torch.jit.script
def randint():
return torch.randint(0, 5, [1, 2])
out = randint()
self.assertEqual(out.dtype, torch.int64)
if GRAPH_EXECUTOR != ProfilingMode.SIMPLE:
FileCheck().check("Long(*, *, requires_grad=0, device=cpu)") \
.check_not("Float(*, *, requires_grad=0, device=cpu)") \
.check_not("Double(*, *, requires_grad=0, device=cpu)") \
.run(randint.graph_for())
@unittest.skipIf(not RUN_CUDA, "no CUDA")
@unittest.skipIf(GRAPH_EXECUTOR != ProfilingMode.PROFILING, "skip if profiling isn't enabled")
def test_autodiff_complex(self):
def foo(x: torch.Tensor, y: torch.Tensor, W: torch.Tensor):
return torch.exp(torch.mm(torch.complex(x, y), W.cfloat()))
@torch.jit.script
def jitted_foo(x: torch.Tensor, y: torch.Tensor, W: torch.Tensor):
return torch.exp(torch.mm(torch.complex(x, y), W.cfloat()))
x = torch.randn(128, 16, dtype=torch.float32, device='cuda:0')
y = torch.randn(128, 16, dtype=torch.float32, device='cuda:0')
W = torch.randn(16, 1, dtype=torch.float32, device='cuda:0', requires_grad=True)
W.data /= 4
with enable_profiling_mode_for_profiling_tests():
for i in range(4):
self.assertTrue((foo(x, y, W).grad_fn is None) == (jitted_foo(x, y, W).grad_fn is None))
def test_linear_grad(self):
with enable_profiling_mode_for_profiling_tests():
def t(x: torch.Tensor, w: torch.Tensor, b: Optional[torch.Tensor]):
return torch.nn.functional.linear(x, w, b)
x_init = torch.randn(4, 2)
w_init = torch.randn(3, 2)
b_init = torch.randn(3)
grad = torch.randn(4, 3)
with disable_autodiff_subgraph_inlining():
# script module
jit_t = torch.jit.script(t)
x = x_init.detach().requires_grad_()
w = w_init.detach().requires_grad_()
b = b_init.detach().requires_grad_()
x_ref = x_init.detach().requires_grad_()
w_ref = w_init.detach().requires_grad_()
b_ref = b_init.detach().requires_grad_()
# profiling/optimization runs
jit_o = jit_t(x, w, b)
jit_o.backward(grad)
jit_o = jit_t(x, w, b)
jit_o.backward(grad)
x.grad.zero_()
w.grad.zero_()
b.grad.zero_()
jit_o = jit_t(x, w, b)
jit_o.backward(grad)
o = t(x_ref, w_ref, b_ref)
o.backward(grad)
self.assertEqual(jit_o, o)
self.assertEqual(x.grad, x_ref.grad)
self.assertEqual(w.grad, w_ref.grad)
self.assertEqual(b.grad, b_ref.grad)
x.grad.zero_()
w.grad.zero_()
x_ref.grad.zero_()
w_ref.grad.zero_()
jit_o = jit_t(x, w, None)
jit_o.backward(grad)
o = t(x_ref, w_ref, None)
o.backward(grad)
self.assertEqual(jit_o, o)
self.assertEqual(x.grad, x_ref.grad)
self.assertEqual(w.grad, w_ref.grad)
@unittest.skipIf(GRAPH_EXECUTOR != ProfilingMode.PROFILING, "the profiling version of test_rand")
def test_rand_profiling(self):
def test_rand():
a = torch.rand([3, 4])
return a + 1.0 - a
# Testing shape analysis correctly setting type
with enable_profiling_mode_for_profiling_tests():
with num_profiled_runs(1):
fn = torch.jit.script(test_rand)
out = fn()
graph_str = torch.jit.last_executed_optimized_graph()
self.assertEqual(out.dtype, torch.double)
FileCheck().check("Double(3, 4, strides=[4, 1], requires_grad=0, device=cpu)") \
.check_not("Float(3, 4, strides=[4, 1], requires_grad=0, device=cpu)").run(graph_str)
# fn = self.checkScript(test_rand, ())
# out = fn()
# self.assertEqual(out.dtype, torch.double)
@torch.jit.script
def randint():
return torch.randint(0, 5, [1, 2])
with enable_profiling_mode_for_profiling_tests():
with num_profiled_runs(1):
out = randint()
graph_str = torch.jit.last_executed_optimized_graph()
self.assertEqual(out.dtype, torch.int64)
FileCheck().check("profiled_type=Long(1, 2, strides=[2, 1], requires_grad=0, device=cpu)").run(graph_str)
def test_erase_number_types(self):
def func(a):
b = 7 + 1 + 3
c = a + b
c += b
return c
graph = torch.jit.script(func).graph
FileCheck().check("int = prim::Constant").check("aten::add_").run(str(graph))
self.run_pass("erase_number_types", graph)
FileCheck().check_not("int = prim::Constant").run(str(graph))
def test_refine_tuple_types(self):
# TupleConstruct output type is not correct here.
graph_str = """
graph(%a : Float(123), %b : Float(4, 5, 6)):
%c : (Tensor, Tensor) = prim::TupleConstruct(%a, %b)
return (%c)
"""
graph = parse_ir(graph_str)
torch._C._jit_pass_refine_tuple_types(graph)
# After the pass, the output type should've been updated.
self.assertTrue('(Float(123), Float(4, 5, 6))' in str(graph.findNode('prim::TupleConstruct').output()))
# TODO(henrytu): Add test for RefineTypes for NamedTuple when it's supported by IR parser.
def test_remove_dropout(self):
weight_0_shape = (20, 5)
weight_1_shape = (20, 20)
input_shape = (10, 5)
class M(torch.nn.Module):
def __init__(self):
super(M, self).__init__()
self.weight_0 = torch.nn.Parameter(torch.rand(weight_0_shape))
self.weight_1 = torch.nn.Parameter(torch.rand(weight_1_shape))
def forward(self, x):
o = F.linear(x, self.weight_0)
o = F.dropout(o, training=self.training)
o = F.linear(o, self.weight_1)
return o
data = torch.rand(input_shape)
m = M()
m = torch.jit.script(m)
with self.assertRaisesRegex(RuntimeError, r'Dropout removal module in training mode is not yet supported'):
torch._C._jit_pass_remove_dropout(m._c)
m.eval()
ref_res = m(data)
# Need to inline otherwise we see instances of Function.
# We would have to use torch.linear/dropout to get around it otherwise.
from torch.jit._recursive import wrap_cpp_module
m = wrap_cpp_module(torch._C._freeze_module(m._c))
torch._C._jit_pass_remove_dropout(m._c)
res = m(data)
FileCheck().check_not("aten::dropout").run(str(m.graph))
torch.testing.assert_close(ref_res, res, rtol=1e-2, atol=1e-3)
def test_unfold_zero_dim(self):
def fn(x):
return x.unfold(0, 1, 1)
graph = torch.jit.script(fn).graph
torch._C._jit_pass_complete_shape_analysis(graph, (torch.tensor(0.39),), False)
out_dims = fn(torch.tensor(0.3923)).ndim
self.assertEqual(graph.findNode("aten::unfold").output().type().dim(), out_dims)
def test_mm_batching(self):
with enable_profiling_mode_for_profiling_tests():
lstm_cell = torch.jit.script(LSTMCellS)
def lstm(x, hx, cx, w_ih, w_hh, b_ih, b_hh):
for i in range(x.size(0)):
hx, cx = lstm_cell(x[i], hx, cx, w_ih, w_hh, b_ih, b_hh)
return hx
slstm = torch.jit.script(lstm)
inputs = get_lstm_inputs('cpu', training=True, seq_length=10)
slstm(*inputs, profile_and_replay=True).sum().backward(retain_graph=True)
if GRAPH_EXECUTOR == ProfilingMode.PROFILING:
slstm(*inputs, profile_and_replay=True).sum().backward()
fw_graph = slstm.graph_for(*inputs)
if GRAPH_EXECUTOR == ProfilingMode.LEGACY:
bw_graph = backward_graph(slstm, diff_graph_idx=0)
self.assertTrue('prim::MMBatchSide' in str(fw_graph))
self.assertTrue('prim::MMTreeReduce' in str(bw_graph))
sout = slstm(*inputs)
out = lstm(*inputs)
self.assertEqual(sout, out)
self.assertEqual(torch.autograd.grad(sout.sum(), inputs),
torch.autograd.grad(out.sum(), inputs))
def test_loop_unrolling(self):
def fn(x):
y = 0
for i in range(int(x)):
y -= i
return y
graph = torch.jit.script(fn).graph
self.run_pass('loop_unrolling', graph)
unroll_factor = 8
FileCheck().check("prim::Loop").check_count("aten::sub", unroll_factor) \
.check("prim::Loop").check("aten::sub").run(str(graph))
self.checkScript(fn, (torch.tensor(10),))
def test_loop_unrolling_const(self):
def fn():
y = 0
for _ in range(10):
y -= 1
return y
def fn2():
y = 0
for i in range(10):
y -= i
return y
def check(fn, name):
graph = torch.jit.script(fn).graph
self.run_pass('loop_unrolling', graph)
# entirely unrolled
FileCheck().check_not("prim::Loop'").run(str(graph))
self.checkScript(fn, ())
check(fn, 'add_const')
check(fn2, 'add_iter')
def test_loop_unrolling_nested(self):
def fn(x):
y = 0
for _ in range(10):
for j in range(int(x)):
y -= j
return y
graph = torch.jit.script(fn).graph
self.run_pass('loop_unrolling', graph)
# inner loop with 8 subs followed by loop epilogue
unroll_factor = 8
FileCheck().check("prim::Loop").check("prim::Loop").check_count('aten::sub', unroll_factor) \
.check("prim::Loop").check("aten::sub").run(str(graph))
self.checkScript(fn, (torch.tensor(10),))
def test_loop_unroll_unused_counter(self):
def fn(x):
y = 0
for _ in range(int(x)):
y -= 1
return y
graph = torch.jit.script(fn).graph
self.run_pass('loop_unrolling', graph)
FileCheck().check("prim::Loop").check_not("aten::add").check("return") \
.run(str(graph))
def test_loop_unroll_negative(self):
def fn(x):
y = 0
for _ in range(int(x)):
y += 1
return y
self.checkScript(fn, (torch.tensor(-20),))
self.checkScript(fn, (torch.tensor(-2),))
self.checkScript(fn, (torch.tensor(-1),))
self.checkScript(fn, (torch.tensor(0),))
self.checkScript(fn, (torch.tensor(1),))
self.checkScript(fn, (torch.tensor(2),))
def test_where(self):
def fn(x, y):
return torch.where(x > 0.0, x, y)
self.checkScript(fn, (torch.randn(3, 2, dtype=torch.float), torch.ones(3, 2, dtype=torch.float)))
def test_where_method(self):
def fn(x, y):
return x.where(x > 0.0, y)
self.checkScript(fn, (torch.randn(3, 2, dtype=torch.float), torch.ones(3, 2, dtype=torch.float)))
def test_union_to_number(self):
@torch.jit.script
def fn(x: Union[int, complex, float], y: Union[int, complex, float]):
return x + y
FileCheck().check(": Scalar):").run(fn.graph)
def test_reassign_module_lhs(self):
with self.assertRaisesRegex(RuntimeError, 'Cannot re-assign \'self\''):
class ReassignSelfLHS(torch.jit.ScriptModule):
@torch.jit.script_method
def forward(self, x):
for _ in range(20):
self = x
return self
ReassignSelfLHS()
def test_reassign_module_rhs(self):
with self.assertRaisesRegex(RuntimeError, 'Cannot re-assign \'x\' to a value of type module'):
class ReassignSelfRHS(torch.jit.ScriptModule):
@torch.jit.script_method
def forward(self, x):
for _ in range(20):
x = self
return self
ReassignSelfRHS()
def test_unknown_builtin(self):
with self.assertRaisesRegex(RuntimeError, 'object has no attribute or method'):
@torch.jit.script
def unknown_builtin(x):
return x.splork(3)
def test_return_tuple(self):
def return_tuple(x):
a = (x, x)
return a, x
self.checkScript(return_tuple, (torch.rand(4),))
def test_add_tuple_optional(self):
def foo(input: Tuple[torch.Tensor, Optional[torch.Tensor], Optional[torch.Tensor]]) -> Optional[torch.Tensor]:
changed_input = input[0] + 1
value: Tuple[torch.Tensor, Optional[torch.Tensor], Optional[torch.Tensor]] = (changed_input,) + input[1:]
return value[2]
inp: Tuple[torch.Tensor, Optional[torch.Tensor], Optional[torch.Tensor]] = (torch.rand(4), None, None)
self.checkScript(foo, (inp,))
def test_add_tuple_non_optional(self):
def foo(input: Tuple[torch.Tensor, torch.Tensor, torch.Tensor]) -> torch.Tensor:
changed_input = input[0] + 1
value: Tuple[torch.Tensor, torch.Tensor, torch.Tensor] = (changed_input,) + input[1:]
return torch.sum(value[2]) + 4
inp: Tuple[torch.Tensor, torch.Tensor, torch.Tensor] = (torch.rand(4), torch.rand(4), torch.rand(4))
self.checkScript(foo, (inp,))
def test_add_tuple_different_types(self):
def foo(a: Tuple[int, float], b: Tuple[int]) -> int:
c: Tuple[int, float, int] = a + b
d: Tuple[int, float, int, int] = c + b
return d[3] + 1
a = (1, 2.0)
b = (3,)
self.checkScript(foo, (a, b))
def test_add_tuple_same_types(self):
def foo(a: Tuple[int, int], b: Tuple[int, int, int]) -> int:
c: Tuple[int, int, int, int, int] = a + b
d: Tuple[int, int, int, int, int, int, int, int] = c + b
return d[6] - 2
a = (1, 2)
b = (3, 4, 5)
self.checkScript(foo, (a, b))
def test_method_no_self(self):
with self.assertRaisesRegex(RuntimeError, 'methods must have a self argument'):
class MethodNoSelf(torch.jit.ScriptModule):
@torch.jit.script_method # noqa: B902
def forward(): # noqa: B902
return torch.zeros(3, 4)
MethodNoSelf()
def test_return_stmt_not_at_end(self):
def return_stmt(x):
if bool(x > 3):
return x + 3
else:
return x
self.checkScript(return_stmt, (torch.rand(1),))
def test_for_in_range(self):
def fn():
c = 0
for i in range(100):
c += i
return c
self.checkScript(fn, ())
def test_for_in_range_dynamic(self):
def fn():
c = 0
for i in range(100):
acc = 0
for j in range(i):
acc += j
c += acc
return c
self.checkScript(fn, (), optimize=False)
def test_for_in_range_ast(self):
def test_script_for_in_range_ast():
c = 0
for i in range(100):
acc = 0
for j in range(i):
acc += j
c += acc
return c
self.checkScript(test_script_for_in_range_ast, ())
def test_for_in_range_if_ast(self):
@torch.jit.script
def test_script_for_in_range_if_ast(x):
output = x
for i in range(20):
if i == 0:
output = x.unsqueeze(0)
else:
output = torch.cat((output, x.unsqueeze(0)), dim=0)
return output
inputs = self._make_scalar_vars([0], torch.int64)
self.assertEqual(test_script_for_in_range_if_ast(*inputs).shape[0], 20)
def test_for_in_range_start_end(self):
def fn():
x = 0
for i in range(7, 100):
x += i
return x
self.checkScript(fn, ())
def test_for_in_range_start_end_step(self):
def fn(start, end, step):
# type: (int, int, int) -> int
x = 0
for i in range(start, end, step):
x += i
return x
self.checkScript(fn, (7, 100, 7))
self.checkScript(fn, (7, 100, -7))
self.checkScript(fn, (2, -11, -3))
self.checkScript(fn, (2, -11, 3))
self.checkScript(fn, (2, 10, 3))
self.checkScript(fn, (-2, -10, -10))
def test_for_in_range_zero_step(self):
@torch.jit.script
def fn():
x = 0
for i in range(2, -11, 0):
x += i
return x
with self.assertRaisesRegex(RuntimeError, "must not be zero"):
fn()
def test_range_args(self):
with self.assertRaisesRegex(RuntimeError, r'range expected at least 1 arguments, got 0'):
@torch.jit.script
def range_no_arg(x):
for _ in range():
x += 1
return x
with self.assertRaisesRegex(RuntimeError, r'found float'):
@torch.jit.script
def range_non_float():
for i in range(.5):
print(i)
def test_parse_empty_tuple_annotation(self):
cu = torch.jit.CompilationUnit('''
def foo(x : Tuple[()]) -> Tuple[()]:
return x
''')
foo_code = cu.find_function('foo').code
FileCheck().check("Tuple[()]").check("Tuple[()]").run(foo_code)
def test_parse_empty_tuple_annotation_element_error(self):
with self.assertRaisesRegex(
RuntimeError, 'Tuple literal in Tuple type annotation must not have any elements'):
cu = torch.jit.CompilationUnit('''
def foo(x : Tuple[(int,)]) -> Tuple[(int,)]:
return x
''')
def test_parse_none_type_annotation(self):
cu = torch.jit.CompilationUnit('''
def foo(x : NoneType) -> NoneType:
return x
''')
foo_code = cu.find_function('foo').code
FileCheck().check(": NoneType").check("-> NoneType").run(foo_code)
def test_empty_tuple_str(self):
empty_tuple_type = torch._C.TupleType([])
g = {'Tuple' : typing.Tuple}
python_type = eval(empty_tuple_type.annotation_str, g)
assert python_type is typing.Tuple[()]
def test_none_type_str(self):
none_type = torch._C.NoneType.get()
g = {'NoneType' : type(None)}
python_type = eval(none_type.annotation_str, g)
assert python_type is type(None)
@skipIfTorchDynamo("TorchDynamo fails with unknown reason")
def test_zip_enumerate_modulelist(self):
class Sub(torch.nn.Module):
def __init__(self):
super(Sub, self).__init__()
def forward(self, thing):
return thing - 2
class Double(torch.nn.Module):
def __init__(self):
super(Double, self).__init__()
def forward(self, thing):
return thing * 2
# zipping over two
class ZipModLists(torch.nn.Module):
def __init__(self, mods, mods2):
super(ZipModLists, self).__init__()
self.mods = mods
self.mods2 = mods2
def forward(self, x):
iter = 0
for mod1, mod2 in zip(self.mods, self.mods2):
x = mod2(mod1(x))
iter += 1
return x, iter
class ZipWithValues(torch.nn.Module):
__constants__ = ['tup_larger', 'tup_smaller']
def __init__(self, mods, mods2):
super(ZipWithValues, self).__init__()
self.mods = mods
self.mods2 = mods2
self.tup_larger = list(range(len(mods2) + 1))
self.tup_smaller = list(range(max(len(mods2) + 1, 1)))
def forward(self, x):
iter = 0
x2 = x
for val, mod1, mod2 in zip(self.tup_larger, self.mods, self.mods2):
x = mod2(mod1(x)) + val
iter += 1
for val, mod1, mod2 in zip(self.tup_smaller, self.mods, self.mods2):
x2 = mod2(mod1(x2)) + val
iter += 1
return x, iter
mods = nn.ModuleList([Double()]), nn.ModuleList([Double(), Sub(), Sub()]), nn.ModuleList([Sub(), Double()])
for i in range(len(mods)):
for j in range(len(mods)):
mod = ZipModLists(mods[i], mods[j])
self.checkModule(mod, (torch.tensor(.5),))
mod2 = ZipWithValues(mods[i], mods[j])
self.checkModule(mod2, (torch.tensor(.5),))
def test_enumerate_modlist_range(self):
class Double(torch.nn.Module):
def forward(self, thing):
return thing * 2
class Mod(torch.nn.Module):
def __init__(self):
super(Mod, self).__init__()
self.mods = nn.ModuleList([Double(), Double()])
def forward(self, x):
x2 = x
iter = 0
for val, mod in enumerate(self.mods):
x2 = mod(x2) * val
iter += 1
return iter, x, x2
self.checkModule(Mod(), (torch.tensor(.5),))
# variable length, modulelist
class Mod2(Mod):
def forward(self, x):
for val, mod in zip(range(int(x)), self.mods):
x = mod(x) * val
return x
with self.assertRaisesRegex(Exception, "that does not have a statically determinable length"):
torch.jit.script(Mod2())
# modulelist, variable length
class Mod3(Mod):
def forward(self, x):
for val, mod in zip(self.mods, range(int(x))):
x = mod(x) * val
return x
with self.assertRaisesRegex(Exception, "that does not have a statically determinable length"):
torch.jit.script(Mod3())
def test_for_in_enumerate(self):
def fn(x):
# type: (List[int]) -> int
sum = 0
for (i, v) in enumerate(x):
sum += i * v
return sum
self.checkScript(fn, ([1, 2, 3, 4, 5],))
def fn_enumerate_start_arg(x):
# type: (List[int]) -> int
sum = 0
for (i, v) in enumerate(x, 1):
sum += i * v
return sum
self.checkScript(fn_enumerate_start_arg, ([1, 2, 3, 4, 5],))
def fn_enumerate_start_kwarg(x):
# type: (List[int]) -> int
sum = 0
for (i, v) in enumerate(x, start=1):
sum += i * v
return sum
self.checkScript(fn_enumerate_start_kwarg, ([1, 2, 3, 4, 5],))
def fn_nested_enumerate(x):
# type: (List[int]) -> int
sum = 0
for (i, (j, v)) in enumerate(enumerate(x)):
sum += i * j * v
return sum
self.checkScript(fn_nested_enumerate, ([1, 2, 3, 4, 5],))
with self.assertRaisesRegex(RuntimeError, r'enumerate expected at least 1 arguments, got 0'):
@torch.jit.script
def enumerate_no_arg(x):
# type: (List[int]) -> int
sum = 0
for _ in enumerate():
sum += 1
return sum
with self.assertRaisesRegex(RuntimeError, r'enumerate expected at most 2 arguments, got 3'):
@torch.jit.script
def enumerate_too_many_args(x):
# type: (List[int]) -> int
sum = 0
for _ in enumerate(x, x, x):
sum += 1
return sum
def test_list_comprehension_modulelist(self):
class Inner(torch.nn.Module):
def forward(self, x):
return x + 10
class M(torch.nn.Module):
def __init__(self, mod_list):
super(M, self).__init__()
self.module_list = mod_list
def forward(self, x):
out = torch.jit.annotate(List[Tensor], [mod(x) for mod in self.module_list])
return out
mod = M(nn.ModuleList([Inner(), Inner()]))
self.checkModule(mod, (torch.tensor(3),))
mod = M(nn.ModuleList([]))
torch.jit.script(mod)
class M2(M):
def __init__(self, mod_list):
super(M2, self).__init__(mod_list)
def forward(self, x):
out = [mod(x) for mod in self.module_list]
return out
mod = M2(nn.ModuleList([Inner(), Inner()]))
self.checkModule(mod, (torch.tensor(3),))
mod = M2(nn.ModuleList([]))
# defaults to List of Tensor for empty modulelist
self.assertEqual(torch.jit.script(mod)(torch.tensor(.5)), [])
def bad_type_annotation():
out = torch.jit.annotate(int, [x for x in [1, 2, 3]]) # noqa: C416
return out
with self.assertRaisesRegex(Exception, "Expected an annotation"
" of type List"):
torch.jit.script(bad_type_annotation)
def test_list_comprehension_variable_write(self):
# i in comprehension doesn't write to function scope
def foo():
i = 1
x = [i if i != 5 else 3 for i in range(7)] # noqa: C416
return i, x
self.assertEqual(foo(), torch.jit.script(foo)())
def test_for_in_zip(self):
def fn(x, y):
# type: (List[int], List[int]) -> int
sum = 0
for (i, j) in zip(x, y):
sum += i * j
return sum
self.checkScript(fn, ([1, 2, 3, 4, 5], [2, 3, 4, 5, 6]))
def fn_multi_inputs(x, y, z):
# type: (List[int], List[int], List[int]) -> int
sum = 0
for (i, j, k) in zip(x, y, z):
sum += i * j * k
return sum
self.checkScript(fn_multi_inputs, ([1, 2, 3, 4], [2, 3, 4, 5], [3, 4, 5, 6]))
def fn_nested_zip(x, y, z):
# type: (List[int], List[int], List[int]) -> int
sum = 0
for (i, (j, k)) in zip(x, zip(y, z)):
sum += i * j * k
return sum
self.checkScript(fn_multi_inputs, ([1, 2, 3, 4], [2, 3, 4, 5], [3, 4, 5, 6]))
with self.assertRaisesRegex(RuntimeError, r'zip expected at least 1 arguments, got 0'):
@torch.jit.script
def zip_no_arg(x):
# type: (List[int]) -> int
sum = 0
for _ in zip():
sum += 1
return sum
with self.assertRaisesRegex(RuntimeError, r'too many values to unpack: need 2 but found 3'):
@torch.jit.script
def fn_nested_zip_wrong_target_assign(x, y, z):
# type: (List[int], List[int], List[int]) -> int
sum = 0
for (i, (j, k)) in zip(x, y, z):
sum += i * j * k
return sum
def test_for_in_zip_enumerate(self):
def fn_zip_enumerate(x, y):
# type: (List[int], List[int]) -> int
sum = 0
for (i, (j, v), k) in zip(x, enumerate(y), range(0, 100)):
sum += i * j * v * k
return sum
self.checkScript(fn_zip_enumerate, ([1, 2, 3, 4], [2, 3, 4, 5]))
def fn_enumerate_zip(x, y):
# type: (List[int], List[int]) -> int
sum = 0
for (i, (j, v)) in enumerate(zip(x, y)):
sum += i * j * v
return sum
self.checkScript(fn_enumerate_zip, ([1, 2, 3, 4], [2, 3, 4, 5]))
def test_for_in_tensors(self):
def test_sizes(x):
sumz = 0
for s in x:
sumz += 1
return sumz
self.checkScript(test_sizes, (torch.rand(5, 4, 3, 2, 1),))
self.checkScript(test_sizes, (torch.rand(777),))
self.checkScript(test_sizes, (torch.rand(0),))
def test_for_in_tensors_rank0(self):
with self.assertRaisesRegex(RuntimeError, "of a 0-d tensor"):
@torch.jit.script
def test_sizes(x):
sumz = 0
for s in x:
sumz += 1
return sumz
test_sizes(torch.tensor(1))
def test_for_in_tensors_fail_scalar(self):
with self.assertRaisesRegex(RuntimeError, "'float' object is not iterable"):
@torch.jit.script
def test_sizes(x):
# type: (float) -> int
sumz = 0
for s in x:
sumz += 1
return sumz
test_sizes(0.0)
def test_for_in_tensors_nested(self):
def test_sizes(x):
sumz = 0
for n in x:
for t in n:
sumz += 1
return sumz
self.checkScript(test_sizes, (torch.rand(5, 4, 3, 2, 1),))
# to avoid defining sum_list in multiple tests
def get_sum_list_fn(self):
def sum_list(a):
# type: (List[int]) -> int
sum = 0
for i in a:
sum += i
return sum
return sum_list
def test_sum_list_diff_elms(self):
self.checkScript(self.get_sum_list_fn(), ([1, 2, 3, 4, 5],))
def test_sum_list_empty(self):
self.checkScript(self.get_sum_list_fn(), ([],))
def test_sum_list_one(self):
self.checkScript(self.get_sum_list_fn(), ([1],))
def test_sum_list_literal(self):
def sum_list():
# type: () -> int
sum = 0
for i in [1, 2, 3, 4, 5]:
sum += i
return sum
self.checkScript(sum_list, ())
def test_sum_list_wrong_type(self):
with self.assertRaisesRegex(RuntimeError, "'int' object is not iterable"):
@torch.jit.script
def sum_list(a):
# type: (int) -> int
sum = 0
for i in a: # noqa: T484
sum += i
return sum
sum_list(1)
def test_list_iterables(self):
with self.assertRaisesRegex(RuntimeError, 'List of iterables is not supported currently'):
cu = torch.jit.CompilationUnit('''
def list_iterables(x):
for i, j in [2, 3, 4], [5, 6, 7]:
x += i
x += j
return x
''')
def test_for_in_string(self):
def test_strings(x):
# type: (str) -> str
reverse = ""
for c in x:
reverse = c + reverse
return reverse
self.checkScript(test_strings, ("hello",))
self.checkScript(test_strings, ("",))
def test_list_strings(x):
# type: (List[str]) -> str
result = ""
for sub_str in x:
result += sub_str
return result
self.checkScript(test_list_strings, (["hello", "world"],))
self.checkScript(test_list_strings, (["hello", " ", "world", ""],))
def test_for_in_dict(self):
def test_dicts(x):
# type: (Dict[str, int]) -> int
sum = 0
for key in x:
sum += x[key]
return sum
self.checkScript(test_dicts, ({"a": 1, "b": 2, "c": 3},))
def test_dict_keys_values(x):
# type: (Dict[str, int]) -> Tuple[str, int]
key_str = ""
sum = 0
for key in x.keys():
key_str += key
for val in x.values():
sum += val
return key_str, sum
self.checkScript(test_dicts, ({"a": 1, "b": 2, "c": 3},))
def test_for_tuple_unpack(self):
def for_tuple_unpack(x, y):
for i, j in [[3, 4], [5, 6], [7, 8]]:
x += i
y += j
return x, y
self.checkScript(for_tuple_unpack, (torch.tensor(3), torch.tensor(5)))
def nested_tuple_unpack(x, y):
# type: (List[int], List[int]) -> int
sum = 0
for i, (j, k), v in zip(x, enumerate(x), y):
sum += i + j + k + v
return sum
self.checkScript(nested_tuple_unpack, ([1, 3, 5], [2, 4, 6]))
def test_for_tuple_assign(self):
def test_simple_assign(x):
# type: (Tuple[int, float]) -> float
sum = 0.0
for a in x:
sum += float(a)
return sum
self.checkScript(test_simple_assign, ((1, 2.5),))
def test_tuple_assign(x):
# type: (Tuple[Tuple[int, int], Tuple[int, int]]) -> int
sum = 0
for a in x:
sum += a[0]
sum += a[1]
return sum
self.checkScript(test_tuple_assign, (((1, 2), (4, 7)), ))
def test_single_starred_lhs(self):
with self.assertRaisesRegex(RuntimeError, 'A Starred expression may only appear on the lhs within the presence'
' of another non-starred expression'):
cu = torch.jit.CompilationUnit('''
def single_starred_lhs(x):
a = (x, x, x)
*b, = a
return b
''')
def test_singleton_tuple_unpack(self):
def foo(a):
b, = (a,)
return b + 1
self.checkScript(foo, (torch.rand(3),))
def test_tuple_assignments(self):
def var_tuple_assign(x, y):
# type: (Tuple[Tensor, Tensor], Tensor) -> Tensor
(a, b), c = x, y
return a + b + c
tuple_inputs = (torch.randn(1, 4), torch.randn(3, 4))
self.checkScript(var_tuple_assign, (tuple_inputs, torch.randn(3, 4)))
def nested_tuple_assign(x, y, z):
# type: (int, Tuple[int, Tuple[int, int]], Tuple[int, int]) -> int
a, (b, (c, d)), (e, f) = x, y, z
return a + b + c + d + e + f
self.checkScript(nested_tuple_assign, ((1, (2, (3, 4)), (5, 6))))
def subscript_tuple_assign(a, x, i):
# type: (List[int], Tensor, int) -> Tuple[int, Tensor, int]
a[i], (x[i], b) = 1, (2, 3)
return a[i] + 1, x + 5, b
self.checkScript(subscript_tuple_assign, ([12, 7, 9, 11], torch.tensor((3, 13, 17)), 0))
def star_tuple_assign():
# type: () -> Tuple[int, int, Tuple[int, int], Tuple[int, int]]
a, (b, *c), *d = 1, (2, 3, 4), 5, 6
return a, b, c, d
self.checkScript(star_tuple_assign, ())
def subscript_tuple_augmented_assign(a):
# type: (Tuple[int, int]) -> Tuple[int, int]
a[0] += 1
return a
with self.assertRaisesRegex(RuntimeError, 'does not support augmented assign'):
scripted_aug_assign = torch.jit.script(subscript_tuple_augmented_assign)
class AttrTupleAssignmentTestClass:
def __init__(self, a: int, b: int):
self.a = a
self.b = b
def set_ab(self, a: int, b: int):
self.a, self.b = (a, b)
def get(self) -> Tuple[int, int]:
return (self.a, self.b)
make_global(AttrTupleAssignmentTestClass)
@torch.jit.script
def attr_tuple_assignment(o: AttrTupleAssignmentTestClass, a: int, b: int):
o.set_ab(a, b)
return o
o = AttrTupleAssignmentTestClass(1, 2)
self.assertEqual(attr_tuple_assignment(o, 3, 4).get(), (3, 4))
def test_multiple_assign(self):
def test():
a = b, c = d, f = (1, 1)
# side effect
ten = torch.tensor(1)
ten1 = ten2 = ten.add_(1)
# ordering
x = 1
y = 3
x, y = y, x + y
return a, b, c, d, f, ten, ten1, ten2, x, y
self.checkScript(test, ())
def test_multi_reduction(self):
with self.assertRaisesRegex(
RuntimeError,
'augmented assignment can only have one LHS expression'):
cu = torch.jit.CompilationUnit('''
def multi_reduction(x):
a, b += x
return a, b
''')
def test_invalid_call_arguments(self):
with self.assertRaisesRegex(RuntimeError, 'but instead found type '):
@torch.jit.script
def invalid_call_arguments(x):
return torch.unsqueeze(3, 4, 5, 6, 7, 8)
def test_invalid_lhs_assignment(self):
with self.assertRaisesRegex(RuntimeError, 'unexpected expression'):
cu = torch.jit.CompilationUnit('''
def invalid_lhs_assignment(x):
x + 1 = x
return x
''')
def test_multi_starred_expr_lhs(self):
with self.assertRaisesRegex(RuntimeError, 'Only one starred expression is allowed on the lhs'):
cu = torch.jit.CompilationUnit('''
def multi_starred_expr_lhs():
a, *b, *c = [1, 2, 3, 4, 5, 6]
return a
''')
def test_pack_tuple_into_non_var(self):
with self.assertRaisesRegex(RuntimeError, 'Cannot pack a tuple into a non-variable'):
cu = torch.jit.CompilationUnit('''
def pack_tuple_into_non_var(x):
a, *1 = (3, 4, 5)
return x
''')
def test_print_kwargs(self):
with self.assertRaisesRegex(RuntimeError, 'print doesn\'t accept any keyword arguments'):
cu = torch.jit.CompilationUnit('''
def print_kwargs(x):
print(x, flush=True)
return x
''')
def test_builtin_use_as_value(self):
with self.assertRaisesRegex(RuntimeError, 'builtin cannot be used as a value'):
@torch.jit.script
def builtin_use_as_value(x):
return x.unsqueeze
def test_wrong_use_as_tuple(self):
with self.assertRaisesRegex(RuntimeError, 'cannot be used as a tuple'):
def test_fn():
return 3
@torch.jit.script
def wrong_use_as_tuple(self):
a, b = test_fn
return a
def test_wrong_attr_lookup(self):
with self.assertRaisesRegex(RuntimeError, 'attribute lookup is not defined on builtin'):
@torch.jit.script
def wrong_attr_lookup(self, x):
a = x.unsqueeze.myattr
return a
def test_wrong_use_as_callable(self):
with self.assertRaisesRegex(RuntimeError, 'cannot call a value'):
@torch.jit.script
def wrong_use_as_callable(x):
return x(3, 4, 5)
def test_python_val_doesnt_have_attr(self):
with self.assertRaisesRegex(RuntimeError, 'object has no attribute abcd'):
@torch.jit.script
def python_val_doesnt_have_attr():
# this has to be a module otherwise attr lookup would not be
# allowed in the first place
return shutil.abcd
def test_wrong_module_attr_lookup(self):
with self.assertRaisesRegex(RuntimeError, 'python value of type \'type\' cannot be used as a value'):
import io
@torch.jit.script
def wrong_module_attr_lookup():
return io.BytesIO
def test_wrong_method_call_inputs(self):
with self.assertRaisesRegex(RuntimeError, 'Argument y not provided'):
class SomeModule(torch.jit.ScriptModule):
@torch.jit.script_method
def foo(self, x, y):
return x
@torch.jit.script_method
def forward(self, x, y):
return self.foo(x)
SomeModule()
def test_single_starred_expr_for_loop(self):
with self.assertRaisesRegex(RuntimeError, 'A Starred expression may only appear'):
cu = torch.jit.CompilationUnit('''
def test():
x = 0
for *a in [1, 2, 3]:
x = x + 1
return x
''')
def test_call_ge(self):
with self.assertRaisesRegex(RuntimeError, 'Expected at most 1 arguments but found 3'):
@_trace(torch.zeros(1, 2, 3))
def foo(x):
return x
@torch.jit.script
def test_fn():
return foo(torch.full([1], 1), torch.full([1], 2), torch.full([1], 3))
def test_wrong_return_type(self):
with self.assertRaisesRegex(RuntimeError, 'but instead got value of type tuple'):
@torch.jit.ignore
def somefunc():
# type: () -> Tuple[Tuple[Tensor, Tensor]]
return torch.zeros(3, 4), torch.zeros(4, 5) # noqa: T484
@torch.jit.script
def wrong_return_type():
return somefunc()
wrong_return_type()
# Tests for calling between different front-end modes
def test_call_python_fn_from_tracing_fn(self):
def python_fn(x):
return torch.neg(x)
@_trace(torch.rand(3, 4))
def traced_fn(x):
return python_fn(x) + 1
# The neg op in the python function should be properly inlined to the
# graph
FileCheck().check("aten::neg").run(str(traced_fn.graph))
def test_call_python_mod_from_tracing_fn(self):
class PythonMod(torch.nn.Module):
def __init__(self):
super(PythonMod, self).__init__()
self.param = torch.nn.Parameter(torch.rand(4, 3), requires_grad=False)
def forward(self, x):
return torch.mm(x, self.param)
pm = PythonMod()
@_trace(torch.rand(3, 4))
def traced_fn(x):
return pm(x) + 1.0
# Note: the parameter self.param from the Python module is inlined
# into the graph
self.assertTrue(len(list(traced_fn.graph.inputs())) == 1)
FileCheck().check("aten::mm").check("aten::add").run(str(traced_fn.graph))
@_tmp_donotuse_dont_inline_everything
def test_call_traced_fn_from_tracing_fn(self):
@_trace(torch.rand(3, 4))
def traced_fn1(x):
return torch.neg(x)
@_trace(torch.rand(3, 4))
def traced_fn(x):
return traced_fn1(x) + 1
FileCheck().check("traced_fn").check("prim::CallFunction").check("aten::add") \
.run(str(traced_fn.graph))
@unittest.skip("error in first class mode")
def test_call_traced_mod_from_tracing_fn(self):
class TracedModule(torch.nn.Module):
def __init__(self):
super(TracedModule, self).__init__()
self.param = torch.nn.Parameter(torch.rand(4, 3), requires_grad=False)
def forward(self, x):
return torch.mm(x, self.param)
tm = torch.jit.trace(TracedModule(), torch.rand(3, 4))
with self.assertRaisesRegex(RuntimeError, "must be registered as submodules"):
@_trace(torch.rand(3, 4))
def traced_fn(x):
return tm(x) + 1.0
@_tmp_donotuse_dont_inline_everything
def test_call_script_fn_from_tracing_fn(self):
@torch.jit.script
def script_fn(x):
return torch.neg(x)
@_trace(torch.rand(3, 4))
def traced_fn(x):
return script_fn(x) + 1
FileCheck().check("prim::CallFunction").check("aten::add").run(str(traced_fn.graph))
@unittest.skip("error in first class mode")
def test_call_script_mod_from_tracing_fn(self):
with self.assertRaisesRegex(RuntimeError, "must be registered as submodules"):
class ScriptMod(torch.jit.ScriptModule):
def __init__(self):
super(ScriptMod, self).__init__()
self.param = torch.nn.Parameter(torch.rand(3, 4), requires_grad=False)
@torch.jit.script_method
def forward(self, x):
for _i in range(4):
x += self.param
return x
sm = ScriptMod()
@_trace(torch.rand(3, 4))
def traced_fn(x):
return sm(x) + 1.0
def test_call_python_fn_from_traced_module(self):
def python_fn(x):
return torch.neg(x)
class TracedModule(torch.nn.Module):
def __init__(self):
super(TracedModule, self).__init__()
self.param = torch.nn.Parameter(torch.rand(4, 3))
def forward(self, x):
return torch.mm(python_fn(x), self.param)
tm = torch.jit.trace(TracedModule(), torch.rand(3, 4))
# Note: parameter self.param from the traced module should appear as
# an input to the graph and the neg op from the Python function should
# be properly inlined
self.assertTrue(len(list(tm.graph.inputs())) == 2)
FileCheck().check("aten::neg").check("aten::mm").run(str(tm.graph))
def test_call_python_mod_from_traced_module(self):
class PythonModule(torch.nn.Module):
def __init__(self):
super(PythonModule, self).__init__()
self.param = torch.nn.Parameter(torch.rand(5, 7))
def forward(self, x):
return torch.mm(x, self.param)
class TracedModule(torch.nn.Module):
def __init__(self):
super(TracedModule, self).__init__()
self.param = torch.nn.Parameter(torch.rand(4, 5))
self.mod = PythonModule()
def forward(self, x):
return self.mod(torch.mm(x, self.param)) + 1.0
tm = torch.jit.trace(TracedModule(), torch.rand(3, 4))
FileCheck().check_not("value=<Tensor>").check("aten::mm")\
.check("prim::CallMethod[name=\"forward\"]").check("aten::add") \
.run(str(tm.graph))
FileCheck().check("aten::mm").run(str(tm.mod.graph))
def test_op_dtype(self):
def check_equal_and_dtype(a, b):
self.assertEqual(a, b)
self.assertEqual(a.dtype, b.dtype)
def fn():
a = torch.arange(10)
b = torch.arange(10, dtype=torch.float)
c = torch.arange(1, 10, 2)
d = torch.arange(1, 10, 2, dtype=torch.float)
e = torch.arange(1, 10., 2)
f = torch.arange(1, 10., 2, dtype=torch.float)
return a, b, c, d, e, f
scripted_fn = torch.jit.script(fn)
eager_out = fn()
script_out = scripted_fn()
for a, b in zip(eager_out, script_out):
check_equal_and_dtype(a, b)
def test_floor_div(self):
@torch.jit.script
def foo(a, b):
# type: (int, int) -> int
return a // b
for i in range(-8, 8):
for j in range(-8, 8):
if j != 0:
self.assertEqual(foo(i, j), i // j)
def test_floordiv(self):
funcs_template = dedent('''
def fn():
ten = {a_construct}
ten_or_scalar = {b_construct}
return ten // ten_or_scalar, torch.floor_divide(ten, ten_or_scalar)
''')
lhs = ["torch.tensor([5.5, 3.2])", "torch.tensor([2, 2])", "torch.tensor([3, 2])"]
rhs = ["1.5", "2", "4", "1.1"] + lhs
for tensor in lhs:
for tensor_or_scalar in rhs:
funcs_str = funcs_template.format(a_construct=tensor, b_construct=tensor_or_scalar)
scope = {}
execWrapper(funcs_str, globals(), scope)
cu = torch.jit.CompilationUnit(funcs_str)
f_script = cu.fn
f = scope['fn']
self.assertEqual(f_script(), f())
def test_call_python_fn_from_script_fn(self):
@torch.jit.ignore
def python_fn(x):
return torch.neg(x)
@torch.jit.script
def script_fn(x):
return python_fn(x) + 1
# Note: the call to python_fn appears as `^python_fn()` and is called
# as a PythonOp in the interpreter
a = torch.tensor(1)
self.assertEqual(script_fn(a), torch.tensor(0))
FileCheck().check("python_fn").run(str(script_fn.graph))
def test_call_python_mod_from_script_fn(self):
class PythonModule(torch.nn.Module):
def __init__(self):
super(PythonModule, self).__init__()
self.param = torch.nn.Parameter(torch.rand(5, 7))
def forward(self, x):
return torch.mm(x, self.param)
pm = PythonModule()
@torch.jit.script
def script_fn(x):
return pm(x) + 1
# Note: call to pm(x) appears as ^<python_value>() in the trace.
# Parameters are NOT inlined.
FileCheck().check("python_value").check("aten::add").run(str(script_fn.graph))
@_tmp_donotuse_dont_inline_everything
def test_call_script_fn_from_script_fn(self):
@torch.jit.script
def script_fn1(x):
return torch.neg(x)
@torch.jit.script
def script_fn(x):
return script_fn1(x) + 1
FileCheck().check("prim::CallFunction").run(str(script_fn.graph))
def test_call_script_mod_from_script_fn(self):
with self.assertRaisesRegex(RuntimeError, "Cannot call a ScriptModule that is not a submodule of the caller"):
class ScriptMod(torch.jit.ScriptModule):
def __init__(self):
super(ScriptMod, self).__init__()
@torch.jit.script_method
def forward(self, x):
return torch.mm(x, torch.zeros([4, 3]))
sm = ScriptMod()
@torch.jit.script
def script_fn(x):
return sm(x) + 1
def test_call_python_fn_from_script_module(self):
@torch.jit.ignore
def python_fn(x):
return torch.neg(x)
class ScriptMod(torch.jit.ScriptModule):
def __init__(self):
super(ScriptMod, self).__init__()
self.param = torch.nn.Parameter(torch.rand(4, 3))
@torch.jit.script_method
def forward(self, x):
return python_fn(torch.mm(x, self.param))
sm = ScriptMod()
FileCheck().check("aten::mm").check("python_fn") \
.run(str(sm.forward.graph))
def test_call_python_mod_from_script_module(self):
class PythonMod(torch.nn.Module):
def __init__(self):
super(PythonMod, self).__init__()
self.param = torch.nn.Parameter(torch.rand(3, 5))
@torch.jit.ignore
def forward(self, x):
return torch.mm(x, self.param)
class ScriptMod(torch.jit.ScriptModule):
def __init__(self):
super(ScriptMod, self).__init__()
self.param = torch.nn.Parameter(torch.rand(4, 3))
self.pm = PythonMod()
@torch.jit.script_method
def forward(self, x):
return self.pm(torch.mm(x, self.param))
sm = ScriptMod()
# Note: the call into PythonMod appears as ^forward(). Parameters
# are NOT inlined
FileCheck().check("aten::mm").check("forward").run(str(sm.graph))
@_tmp_donotuse_dont_inline_everything
def test_call_script_fn_from_script_module(self):
@torch.jit.script
def script_fn(x):
return torch.neg(x)
class ScriptMod(torch.jit.ScriptModule):
def __init__(self):
super(ScriptMod, self).__init__()
self.param = torch.nn.Parameter(torch.rand(4, 3))
@torch.jit.script_method
def forward(self, x):
return script_fn(torch.mm(x, self.param))
sm = ScriptMod()
graph = (sm.forward.graph)
FileCheck().check("aten::mm").check("prim::CallFunction").run(str(graph))
@_tmp_donotuse_dont_inline_everything
def test_call_script_mod_from_script_module(self):
class ScriptMod1(torch.jit.ScriptModule):
def __init__(self):
super(ScriptMod1, self).__init__()
self.param = torch.nn.Parameter(torch.rand(3, 5))
@torch.jit.script_method
def forward(self, x):
return torch.mm(x, self.param)
class ScriptMod(torch.jit.ScriptModule):
def __init__(self):
super(ScriptMod, self).__init__()
self.param = torch.nn.Parameter(torch.rand(4, 3))
self.tm = ScriptMod1()
@torch.jit.script_method
def forward(self, x):
return self.tm(torch.mm(x, self.param))
sm = ScriptMod()
# Note: the parameters from both modules should appear in the flattened
# input list to the graph. The mm op from ScriptMod1 should be properly
# inlined
# 3 % values in graph input lists, two mms in body
FileCheck().check_count('%', 3).check(":").check_count("mm", 1).check("prim::CallMethod").run(str(sm.graph))
def test_module_with_params_called_fails(self):
with self.assertRaisesRegex(RuntimeError, "Cannot call a ScriptModule that is not a submodule of the caller"):
class ScriptMod(torch.jit.ScriptModule):
def __init__(self):
super(ScriptMod, self).__init__()
self.param = torch.nn.Parameter(torch.rand(3, 3))
@torch.jit.script_method
def forward(self, x):
return torch.mm(x, self.param)
sm = ScriptMod()
@torch.jit.script
def some_func(x):
return sm(x)
def test_tuple_index_to_list(self):
def test_non_constant_input(a):
# type: (bool) -> int
if a:
b = 1
else:
b = 0
c = (0, 1)
return c[b]
self.checkScript(test_non_constant_input, (True,))
self.checkScript(test_non_constant_input, (False,))
with self.assertRaisesRegex(RuntimeError, "because we cannot resolve the output type"):
@torch.jit.script
def test_non_constant_input(a):
# type: (bool) -> None
if a:
b = 1
else:
b = 0
c = (0, 1.1)
print(c[b])
def test_tuple_indexing(self):
def tuple_index(a):
if bool(a):
b = (1, 2)
else:
b = (0, 2)
return b[-2], b[1]
self.checkScript(tuple_index, (torch.tensor([0]),))
self.checkScript(tuple_index, (torch.tensor([1]),))
self.checkScript(tuple_index, (torch.tensor([1]),), optimize=True)
tuple_comp = torch.jit.script(tuple_index)
FileCheck().check_count("TupleIndex", 2, exactly=True).run(str(tuple_comp.graph))
with self.assertRaisesRegex(RuntimeError, "index must be an integer"):
@torch.jit.script
def test_indexing_float():
c = (1, 2)
return c[0.1]
def test_indexing_out_of_bounds_pos():
c = (1, 2)
return c[2]
self.checkScriptRaisesRegex(test_indexing_out_of_bounds_pos, (), Exception,
"out of range")
def test_indexing_out_of_bounds_neg():
c = (1, 2)
return c[-3]
self.checkScriptRaisesRegex(test_indexing_out_of_bounds_pos, (), Exception,
"out of range")
def negative_index():
tup = (1, 2, 3, 4)
return tup[-1]
self.checkScript(negative_index, [])
def really_negative_index():
tup = (1, 2, 3, 4)
return tup[-100]
self.checkScriptRaisesRegex(really_negative_index, [], Exception, "index out of range")
def negative_slice():
tup = (1, 2, 3, 4)
return tup[-3:4]
self.checkScript(negative_slice, [])
def really_slice_out_of_bounds():
tup = (1, 2, 3, 4)
return tup[-300:4000]
self.checkScript(really_slice_out_of_bounds, [])
def test_namedtuple_attr(self):
def f(x):
return x.max(dim=1).indices + torch.max(x, dim=1).indices
self.checkScript(f, (torch.rand(20, 20, 20),), optimize=True)
with self.assertRaisesRegex(RuntimeError, "object has no attribute or method"):
@torch.jit.script
def g1(x):
return x.max(dim=1).unknown_symbol
with self.assertRaisesRegex(RuntimeError, "object has no attribute or method"):
@torch.jit.script
def g2(x):
print((x, x, x).__doc__)
return x
def test_tuple_len(self):
@torch.jit.script
def foo():
return len((1, "str", None))
self.assertEqual(foo(), 3)
@torch.jit.script
def test_indexing_end_out_of_bounds():
c = (1, 2)
return c[2:10]
self.assertEqual(test_indexing_end_out_of_bounds(), ())
def test_lower_nested_tuples(self):
@torch.jit.script
def test():
return ((1, 2), 3)
self.run_pass('constant_propagation', test.graph)
FileCheck().check("prim::Constant").check_not("TupleConstruct").run(test.graph)
# fails if a tuple can't be lowered
self.run_pass('lower_all_tuples', test.graph)
def test_unwrap_optional_builtin(self):
def test(x):
# type: (Optional[int]) -> int
x = torch.jit._unwrap_optional(x)
x = x + x # noqa: T484
return x
self.checkScript(test, (3,))
with self.assertRaisesRegex(AssertionError, "Unwrapping null optional"):
test(None)
test_script = torch.jit.script(test)
with self.assertRaisesRegex(RuntimeError, "Unwrapping null optional"):
test_script(None)
@torch.jit.script
def test_test():
return torch.jit._unwrap_optional(1)
with self.assertRaisesRegex(RuntimeError, r"could not be inferred from actual type None"):
@torch.jit.script
def test_no_type():
# type: () -> int
return torch.jit._unwrap_optional(None)
def test_indexing_error(self):
with self.assertRaisesRegex(RuntimeError, "'int' object is not subscriptable"):
@torch.jit.script
def test_wrong_type():
a = 8
return a[0]
def test_unsupported_builtin_error(self):
with self.assertRaisesRegex(RuntimeError,
"Python builtin <built-in function hypot> is currently"):
@torch.jit.script
def test_unsupported(a):
return math.hypot(a, 2.0)
def test_annotated_script_fn(self):
@torch.jit.script
def foo(x, y, z):
# type: (Tensor, Tuple[Tensor, Tensor, Tensor], Tuple[Tensor, Tuple[Tensor, Tensor]]) -> Tensor
return x
self.assertExpected(str(foo.schema))
def test_annotated_script_method(self):
class SM(torch.jit.ScriptModule):
@torch.jit.script_method
def forward(self, x, y):
# type: (Tuple[Tensor, Tensor], Tensor) -> Tuple[Tensor, Tensor, Tensor]
return y, y, y
sm = SM()
self.assertExpectedStripMangled(str(sm.forward.schema))
def test_annotated_script_fn_return_mismatch(self):
with self.assertRaisesRegex(RuntimeError, "but is actually of type"):
@torch.jit.script
def return_tup(x):
# type: (Tensor) -> Tuple[Tuple[Tensor, Tensor], Tensor]
return x, x # noqa: T484
def test_annotated_script_fn_arg_mismatch(self):
with self.assertRaisesRegex(RuntimeError, r"Arguments for call are not valid"):
@torch.jit.script
def tuple_arg(x):
# type: (Tuple[Tensor, Tensor]) -> Tensor
return x + 1 # noqa: T484
def test_script_non_tensor_args_outputs(self):
@torch.jit.script
def fn(x, y):
# type: (Tensor, float) -> float
return float((x + y).sum())
x = torch.ones(2, 2)
z = fn(x, 1)
self.assertIsInstance(z, float)
self.assertEqual(z, 8.)
@unittest.skip('https://github.com/pytorch/pytorch/issues/9595')
def test_inline_and_run_annotated_script_fn(self):
@torch.jit.script
def to_inline(x, y):
# type: (Tuple[Tensor, Tensor], Tensor) -> Tensor
return y
@torch.jit.script
def some_func(x):
return to_inline((x, x), x)
x = torch.rand(3, 4)
self.assertEqual(some_func(x), x)
def test_file_format_serialization(self):
filename = tempfile.mktemp()
writer = torch._C.PyTorchFileWriter(filename)
buffers = [os.urandom(size) for size in [random.randint(1, 100) for i in range(20)]]
offsets = []
for i, buf in enumerate(buffers):
writer.write_record(str(i), buf, len(buf))
offsets.append(i)
serialized_offsets = pickle.dumps(offsets)
writer.write_record("meta", serialized_offsets, len(serialized_offsets))
writer.write_end_of_file()
reader = torch._C.PyTorchFileReader(filename)
serialized_offsets_read = reader.get_record("meta")
parsed_serialized_offsets = pickle.loads(serialized_offsets)
for i, offset in enumerate(parsed_serialized_offsets):
data = reader.get_record(str(offset))
assert(data == buffers[i])
# for each type, the input type annotation and corresponding return type annotation
def type_input_return_pairs(self):
return [
('Tensor', 'Tensor'),
('torch.Tensor', 'Tensor'),
('str', 'str'),
('int', 'int'),
('bool', 'bool'),
('BroadcastingList3[float]', 'List[float]'),
('BroadcastingList2[int]', 'List[int]'),
('List[int]', 'List[int]'),
('Optional[int]', 'Optional[int]'),
]
# replacing code input & return type pair
def format_code(self, code, pair):
return code.format(input=pair[0], output=pair[1])
# ***** Type annotation tests ****
# Test combinations of:
# {String frontend, Python AST Frontend}
# {Python 3-style type annotations, MyPy-style type comments}
# {Script method, Script function}
# String frontend , Python 3-style type annotations , Script function
def test_annot_string_py3_fn(self):
code = '''
def foo(x : {input}, y : Tuple[Tensor, Tensor]) -> Tuple[{output}, {output}]:
return x, x
'''
test_str = []
for pair in self.type_input_return_pairs():
cu = torch.jit.CompilationUnit(self.format_code(code, pair))
test_str.append(str(cu.foo.schema))
self.assertExpected("\n".join(test_str) + "\n")
# String frontend , Python 3-style type annotations , Script method
def test_annot_string_py3_method(self):
class TestModule(torch.jit.ScriptModule):
def __init__(self):
super(TestModule, self).__init__()
code = '''
def foo(self, x : {input}, y : Tuple[Tensor, Tensor]) -> Tuple[{output}, {output}]:
return x, x
'''
test_str = []
for pair in self.type_input_return_pairs():
# clear the class registry as we will be defining foo multiple times
jit_utils.clear_class_registry()
tm = TestModule()
tm.define(self.format_code(code, pair))
test_str.append(str(tm.foo.schema))
self.assertExpectedStripMangled("\n".join(test_str) + "\n")
# String frontend , MyPy-style type comments , Script function
def test_annot_string_mypy_fn(self):
code = '''
def foo(x, y):
# type: ({input}, Tuple[Tensor, Tensor]) -> Tuple[{output}, {output}]
return x, x
'''
test_str = []
for pair in self.type_input_return_pairs():
cu = torch.jit.CompilationUnit(self.format_code(code, pair))
test_str.append(str(cu.foo.schema))
self.assertExpectedStripMangled("\n".join(test_str) + "\n")
# String frontend , MyPy-style type comments , Script method
def test_annot_string_mypy_method(self):
class TestModule(torch.jit.ScriptModule):
def __init__(self):
super(TestModule, self).__init__()
code = '''
def foo(self, x, y):
# type: ({input}, Tuple[Tensor, Tensor]) -> Tuple[{output}, {output}]
return x, x
'''
test_str = []
for pair in self.type_input_return_pairs():
# clear the class registry as we will be defining foo multiple times
jit_utils.clear_class_registry()
tm = TestModule()
tm.define(self.format_code(code, pair))
test_str.append(str(tm.foo.schema))
self.assertExpectedStripMangled("\n".join(test_str) + "\n")
# Python AST Frontend , Python 3-style type annotations , Script function
def test_annot_ast_py3_fn(self):
code = dedent('''
from typing import Tuple, List, Optional
from torch import Tensor
from torch.jit.annotations import BroadcastingList2, BroadcastingList3
import torch
@torch.jit.script
def foo(x : {input}, y : Tuple[Tensor, Tensor]) -> Tuple[{output}, {output}]:
return x, x
''')
test_str = []
for pair in self.type_input_return_pairs():
fn = jit_utils._get_py3_code(self.format_code(code, pair), 'foo')
test_str.append(str(fn.schema))
self.assertExpectedStripMangled("\n".join(test_str) + "\n")
def test_multiline_annot_ast_py3_fn(self):
code = dedent('''
from typing import Tuple, List, Optional
from torch import Tensor
from torch.jit.annotations import BroadcastingList2, BroadcastingList3
import torch
@torch.jit.script
def foo(x, # type: {input}
y # type: Tuple[Tensor, Tensor]
):
# type: (...) -> Tuple[{output}, {output}]
return x, x
''')
test_str = []
for pair in self.type_input_return_pairs():
fn = jit_utils._get_py3_code(self.format_code(code, pair), 'foo')
args = fn.schema.arguments
returns = fn.schema.returns
self.assertEqual(str(args[0].type), pair[1])
self.assertEqual(str(args[1].type), "Tuple[Tensor, Tensor]")
self.assertEqual(str(returns[0].type), "Tuple[{}, {}]".format(pair[1], pair[1]))
def test_bad_multiline_annotations(self):
with self.assertRaisesRegex(RuntimeError, "Return type line"):
@torch.jit.script
def bad_type_line(a, # type: Tensor
b, # type: Tensor
c # type: Tensor
):
# type: (int, int, int) -> Tensor
# type: bad type line # noqa: F723
return a + b + c
with self.assertRaisesRegex(RuntimeError, "Return type line"):
@torch.jit.script
def bad_return_line(a, # type: Tensor
b,
c # type: Tensor
):
# type: (int, int, int) -> Tensor
return a + b + c
# TODO: this should be supported but is difficult to parse
with self.assertRaisesRegex(RuntimeError, "Number of type annotations"):
@torch.jit.script
def missing_type(a, # type: Tensor
b,
c # type: Tensor
):
# type: (...) -> Tensor
return a + b + c
# Python AST Frontend , Python 3-style type annotations , Script method
def test_annot_ast_py3_method(self):
code = dedent('''
from typing import Tuple, List, Optional
from torch import Tensor
from torch.jit.annotations import BroadcastingList2, \\
BroadcastingList3
import torch
class FooModule(torch.jit.ScriptModule):
@torch.jit.script_method
def foo(self, x : {input}, y : Tuple[Tensor, Tensor]) -> Tuple[{output}, {output}]:
return x, x
instance = FooModule()
''')
test_str = []
for pair in self.type_input_return_pairs():
fn = jit_utils._get_py3_code(self.format_code(code, pair), 'instance')
test_str.append(str(fn.foo.schema))
self.assertExpectedStripMangled("\n".join(test_str) + "\n")
# Python AST Frontend , MyPy-style type comments , Script function
def test_annot_ast_mypy_fn(self):
code = dedent('''
import torch
@torch.jit.script
def foo(x, y):
# type: ({input}, Tuple[Tensor, Tensor]) -> Tuple[{output}, {output}]
return x, x
''')
test_str = []
for pair in self.type_input_return_pairs():
fn = jit_utils._get_py3_code(self.format_code(code, pair), 'foo')
test_str.append(str(fn.schema))
self.assertExpected("\n".join(test_str) + "\n")
# Python AST Frontend , MyPy-style type comments , Script method
def test_annot_ast_mypy_method(self):
code = dedent('''
import torch
class FooModule(torch.jit.ScriptModule):
@torch.jit.script_method
def foo(self, x, y):
# type: ({input}, Tuple[Tensor, Tensor]) -> Tuple[{output}, {output}]
return x, x
instance = FooModule()
''')
test_str = []
for pair in self.type_input_return_pairs():
fn = jit_utils._get_py3_code(self.format_code(code, pair), 'instance')
test_str.append(str(fn.foo.schema))
self.assertExpectedStripMangled("\n".join(test_str) + "\n")
# Tests that "# type: ignore[*]" is supported in type lines and is
# properly ignored.
def test_mypy_type_ignore(self):
@torch.jit.script
def foo(x): # type: ignore
return x
@torch.jit.script
def bar(x): # type: ignore[no-redef]
return x
def test_method_casts_script(self):
cast_types = [
'byte', 'char', 'double', 'float', 'int', 'long', 'short'
]
for cast_type in cast_types:
cu = torch.jit.CompilationUnit('''
def cast_to(x):
return x.{cast_type}()
'''.format(cast_type=cast_type))
x = torch.rand(3, 4, 5) * 128
cu_result = cu.cast_to(x)
reference = getattr(x, cast_type)()
self.assertEqual(cu_result, reference)
def test_string_frontend_elif(self):
code = '''
def func(niter):
# type: (int)
rv = 0
for i in range(niter):
if i % 3 == 0 and i % 5 == 0:
rv += 35
elif i % 3 == 0:
rv += 3
elif i % 5 == 0:
rv += 5
else:
rv += i
return rv
'''
self.checkScript(dedent(code), (101,))
def test_module_parameters_and_buffers(self):
weights = torch.randn(10, 10)
bias = torch.randn(10)
weights2 = torch.randn(10, 10)
bias2 = torch.randn(10)
class TestLinear(torch.nn.Module):
def __init__(self, in_features, out_features):
super(TestLinear, self).__init__()
self.in_features = in_features
self.out_features = out_features
self.weight = torch.nn.Parameter(torch.empty(out_features, in_features))
self.bias = torch.nn.Parameter(torch.empty(out_features))
self.register_buffer('counter', torch.ones(out_features))
self.reset_parameters()
def reset_parameters(self):
torch.nn.init.kaiming_uniform_(self.weight, a=math.sqrt(5))
if self.bias is not None:
fan_in, _ = torch.nn.init._calculate_fan_in_and_fan_out(self.weight)
bound = 1 / math.sqrt(fan_in)
torch.nn.init.uniform_(self.bias, -bound, bound)
def forward(self, input):
return F.linear(input, self.weight, self.bias) + self.counter
# Initialize a ScriptModule that uses the weak module above multiple times
class Strong(torch.jit.ScriptModule):
def __init__(self):
super(Strong, self).__init__()
self.fc1 = TestLinear(10, 10)
self.fc1.weight = torch.nn.Parameter(weights)
self.fc1.bias = torch.nn.Parameter(bias)
self.fc2 = TestLinear(10, 10)
self.fc2.weight = torch.nn.Parameter(weights2)
self.fc2.bias = torch.nn.Parameter(bias2)
@torch.jit.script_method
def forward(self, x):
return x + self.fc1(x) + self.fc1(x) + self.fc2(x)
strong_mod = Strong()
# Run same calculation as module
inp = torch.ones(10)
lin = torch.nn.Linear(10, 10)
lin.weight = torch.nn.Parameter(weights)
lin.bias = torch.nn.Parameter(bias)
lin2 = torch.nn.Linear(10, 10)
lin2.weight = torch.nn.Parameter(weights2)
lin2.bias = torch.nn.Parameter(bias2)
expected_result = inp + (lin(inp) + torch.ones(10)) * 2 + lin2(inp) + torch.ones(10)
self.assertEqual(strong_mod(inp), expected_result)
self.assertExportImportModule(strong_mod, (inp,))
def test_module_copying(self):
class Submodule(torch.nn.Module):
def __init__(self):
super(Submodule, self).__init__()
def forward(self, x):
return x + 100
class Weak(torch.nn.Module):
def __init__(self, in_features, out_features):
super(Weak, self).__init__()
self.weight = torch.nn.Parameter(torch.ones(out_features, in_features))
self.bias = torch.nn.Parameter(torch.ones(out_features))
self.register_buffer("buffer", torch.ones(out_features))
self.submodule = Submodule()
def forward(self, x):
return F.linear(x, self.weight, self.bias) \
+ self.buffer + self.submodule(x)
class Strong(torch.jit.ScriptModule):
def __init__(self, weak):
super(Strong, self).__init__()
self.weak = weak
@torch.jit.script_method
def forward(self, x):
return self.weak(x)
inp = torch.ones(5, 5) * 5
weak_mod = Weak(5, 5)
strong_mod = Strong(weak_mod)
self.assertTrue(isinstance(strong_mod.weak, torch.jit.ScriptModule))
self.assertFalse(isinstance(weak_mod, torch.jit.ScriptModule))
self.assertIs(strong_mod.weak.weight, weak_mod.weight)
self.assertIs(strong_mod.weak.buffer, weak_mod.buffer)
# strong_mod.weak.submodule has been recursively scripted
self.assertIsNot(strong_mod.weak.submodule, weak_mod.submodule)
weak_mod.weight.data += torch.ones(5, 5) * 100
self.assertTrue(strong_mod(inp).allclose(weak_mod(inp)))
# Re-assignment is not tracked
weak_mod.weight = torch.nn.Parameter(torch.ones(5, 5) * 100)
self.assertFalse(strong_mod(inp).allclose(weak_mod(inp)))
def test_backend_cudnn_enabled(self):
# Only test that this compiles
@torch.jit.script
def fn(x):
if torch.backends.cudnn.enabled:
x = x + 2
else:
x = x + 3
return x
def test_inplace_add(self):
def foo(a, b):
c = a + b
c.add_(b)
return c
self.checkScript(foo, (torch.rand(3), torch.rand(3)))
def test_add_out(self):
def foo(a, b):
c = a + b
e = 2 * a
torch.add(c, b, out=e)
return e
self.checkScript(foo, (torch.rand(3), torch.rand(3)))
def test_tuple_error_msg(self):
def fn(t: Any):
if isinstance(t, tuple):
a, b = t
return a + b
with self.assertRaisesRegexWithHighlight(RuntimeError, "Provided tuple is not fully defined/refined", "t"):
s = torch.jit.script(fn)
def test_augmented_assign(self):
def foo(a, b):
a += b
a -= b
a /= b
a *= b
return a, b
self.checkScript(foo, (torch.rand(3), torch.rand(3)))
def test_ignored_props(self):
class A(nn.Module):
__jit_ignored_attributes__ = ["ignored", "ignored_return_val"]
def __init__(self):
super().__init__()
@property
def ignored(self):
raise ValueError("shouldn't be called")
@property
def ignored_return_val(self):
return 1
@torch.jit.ignore
def call(self):
return self.ignored_return_val
f = torch.jit.script(A())
# jank way to test if there is no error
self.assertTrue(isinstance(f, torch.jit.ScriptModule))
self.assertTrue(isinstance(f.call(), property))
def test_pass(self):
def foo(x):
# type: (bool) -> int
for _i in range(3):
pass
if x:
pass
else:
pass
return 3
self.checkScript(foo, (True,))
def test_lhs_indexing(self):
def foo(a, b):
a = a.clone()
a[0] = b
return a
self.checkScript(foo, (torch.rand(2, 3), torch.rand(3)))
def test_lhs_advanced_indexing_assignment(self):
def foo(x, y):
a = torch.exp(x)
b = x == 1
a[b] = y[b]
return a
self.checkScript(foo, (torch.ones(4, 3), torch.ones(4, 3)))
def test_lhs_advanced_indexing_augmented_assignment(self):
def foo(x, y):
a = torch.exp(x)
b = x == 1
a[b] += y[b]
return a
self.checkScript(foo, (torch.ones(4, 3), torch.ones(4, 3)))
def test_lhs_indexing_list(self):
def foo(a, b):
ls = [a]
ls[0] = b
return ls
self.checkScript(foo, (torch.rand(2, 3), torch.rand(3)))
def test_inplace_copy_script(self):
def foo(x):
a = torch.rand(3, 4)
a.copy_(x)
return a
self.checkScript(foo, (torch.rand(3, 4),))
def test_lhs_indexing_increment(self):
def foo(a, b):
a[0] += b
return a
self.checkScript(foo, (torch.rand(2, 3), torch.rand(3)))
def test_lhs_indexing_increment_list(self):
def foo(a, b):
a = a.clone()
ls = [a, b]
ls[0] += b
return ls
self.checkScript(foo, (torch.rand(2, 3), torch.rand(3)))
def test_lhs_indexing_increment_list_prim(self):
def foo():
ls = [1, 2, 3]
ls[0] += 5
return ls
self.checkScript(foo, ())
def test_lhs_indexing_multi(self):
def foo(a, b):
a = a.clone()
foo, a[0], bar = (1, b, 3)
return foo, a, bar
self.checkScript(foo, (torch.rand(2, 3), torch.rand(3)))
def test_bool_dispatch(self):
with torch._jit_internal._disable_emit_hooks(): # TODO: Python print broadcasting list
def kwarg_false(x):
# type: (Tensor) -> Tensor
return F.max_pool1d(x, 1, 1, return_indices=False)
self.checkScript(kwarg_false, (torch.randn(3, 3, 3),))
def kwarg_true(x):
# type: (Tensor) -> Tuple[Tensor, Tensor]
return F.max_pool1d(x, 1, 1, return_indices=True)
self.checkScript(kwarg_true, (torch.randn(3, 3, 3),))
def full_kwarg_false(x):
# type: (Tensor) -> Tensor
return F.max_pool1d(x, 1, 1, ceil_mode=False, return_indices=False)
self.checkScript(full_kwarg_false, (torch.randn(3, 3, 3),))
def full_kwarg_true(x):
# type: (Tensor) -> Tuple[Tensor, Tensor]
return F.max_pool1d(x, 1, 1, ceil_mode=False, return_indices=True)
self.checkScript(full_kwarg_true, (torch.randn(3, 3, 3),))
def use_default(x):
# type: (Tensor) -> Tensor
return F.max_pool1d(x, 1, 1)
self.checkScript(use_default, (torch.randn(3, 3, 3),))
def arg_false(x):
# type: (Tensor) -> Tensor
return F.max_pool1d(x, 1, 1, 0, 1, False, False)
self.checkScript(arg_false, (torch.randn(3, 3, 3),))
def arg_true(x):
# type: (Tensor) -> Tuple[Tensor, Tensor]
return F.max_pool1d(x, 1, 1, 0, 1, False, True)
self.checkScript(arg_true, (torch.randn(3, 3, 3),))
def test_infer_size(self):
from torch._C import _infer_size
def fn(x, y):
# type: (Tensor, Tensor) -> List[int]
return _infer_size(x.size(), y.size())
self.checkScript(fn, (torch.ones(2, 4, 2), torch.ones(2, 4, 2)))
def test_hash(self):
def tester(fn, inputs):
for x in inputs:
for y in inputs:
if x == y:
self.assertEqual(fn(x), fn(y))
else:
self.assertNotEqual(fn(x), fn(y))
@torch.jit.script
def int_hash(x):
# type: (int) -> int
return hash(x)
@torch.jit.script
def float_hash(x):
# type: (float) -> int
return hash(x)
@torch.jit.script
def str_hash(x):
# type: (str) -> int
return hash(x)
tester(int_hash, (20, 21, 22))
tester(float_hash, (20.0, 21.00001, 22.443))
tester(str_hash, ("", "hello", "a"))
def test_id(self):
with self.assertRaisesRegex(RuntimeError, "Expected a value"):
@torch.jit.script
def test_id_scalars():
return id(2) == id(None)
@torch.jit.script
class FooTest(object):
def __init__(self, x):
self.foo = x
def getFooTest(self):
return self.foo
@torch.jit.script
def test_id_class_types():
obj1 = FooTest(torch.tensor(3))
obj2 = FooTest(torch.tensor(2))
assert obj1 is not obj2
assert id(obj1) != id(obj2)
assert id(obj1) != id(None)
return True
self.assertTrue(test_id_class_types())
def test_mutable_dce(self):
@torch.jit.script
def foo():
a = torch.rand(2, 3)
a += torch.rand(2, 3)
b = torch.rand(2, 3)
b += torch.rand(2, 3)
# b should be cleaned up but not a
return a
FileCheck().check_count("aten::rand", 2, exactly=True) \
.check_count("aten::add", 1, exactly=True).run(str(foo.graph))
def test_mutable_dce_block(self):
@torch.jit.script
def foo():
a = torch.rand(2, 3)
a += torch.rand(2, 3)
b = torch.rand(2, 3)
if bool(a > torch.zeros(2, 3)):
b += torch.rand(2, 3)
a += torch.rand(2, 3)
# a should be cleaned up but not b
return b
FileCheck().check("prim::If").check_count("aten::rand", 1, exactly=True) \
.run(str(foo.graph))
def test_mutable_dce_graph_input(self):
@torch.jit.script
def foo(a):
a += torch.rand(2, 3)
# shouldn't clean up `a` even though it's not used in the output
FileCheck().check("aten::rand").check("aten::add").run(str(foo.graph))
def test_mutable_dce_list(self):
@torch.jit.script
def foo(a):
l = []
l.append(a)
c = l[0]
b = torch.rand(2, 3)
c += torch.rand(2, 3)
return b
# c does not get cleaned up because there is a wildcard + mutation
FileCheck().check_count("aten::rand", 2, exactly=True).run(str(foo.graph))
def test_mutable_dce_loop(self):
@torch.jit.script
def foo(a):
l = []
l.append(a)
i = 0
b = torch.rand(2, 3)
while i < 1:
dead = torch.rand(2, 3)
c = l[0]
c += torch.rand(2, 3)
i += 1
return b
FileCheck().check("prim::Loop").check_not("aten::rand").check("aten::__getitem__") \
.check_count("aten::rand", 1, exactly=True).run(str(foo.graph))
def test_mutable_dce_indirect_wildcards(self):
def fn():
x = torch.ones(2, 3)
x_1 = x.view(-1)
l = []
l.append(x_1)
x_view = l[0]
x.add_(torch.ones(2, 3))
return x_view
self.checkScript(fn, ())
def test_mutable_dce_indirect_wildcard_write(self):
def fn():
indexes = torch.jit.annotate(List[Tensor], [])
word_ids = torch.zeros(10, dtype=torch.int32)
word_ids[1] = 1
indexes.append(word_ids)
return word_ids
self.checkScript(fn, ())
def test_mutable_dce_wildcards(self):
def fn():
x = torch.ones(2, 3)
l = []
l.append(x)
x_view = l[0]
x.add_(torch.ones(2, 3))
return x_view
self.checkScript(fn, (), profiling=ProfilingMode.SIMPLE)
def test_cpp_function_tensor_str(self):
x = torch.randn(2, 2)
scale = torch.randn(2, 2, requires_grad=True)
shift = torch.randn(2, 2, requires_grad=True)
@torch.jit.script
def fn(x, scale, shift):
return scale * x + shift
with self.capture_stdout() as captured:
print(fn(x, scale, shift))
def test_string_index(self):
def fn(x):
# type: (str)
return x[2], x[-1]
self.checkScript(fn, ("abcde",))
def test_ord(self):
def fn(x):
# type: (str) -> int
return ord(x)
self.checkScript(fn, ("h"))
self.checkScript(fn, ("y"))
def index_str_to_tensor(s):
# type: (str) -> Tensor
return torch.tensor(ord(s)) # noqa: T484
s = u'\u00a3'.encode('utf8')[:1]
self.checkScript(index_str_to_tensor, (s,))
def test_chr(self):
def fn(x):
# type: (int) -> str
return chr(x)
self.checkScript(fn, (1,))
self.checkScript(fn, (97,))
def test_round(self):
def round_float(x):
# type: (float) -> float
return round(x)
def round_int(x):
# type: (int) -> float
return round(x)
self.checkScript(round_float, (1.5,))
self.checkScript(round_int, (2,))
def test_convert_base(self):
def test_hex(x):
# type: (int) -> str
return hex(x)
def test_oct(x):
# type: (int) -> str
return oct(x)
def test_bin(x):
# type: (int) -> str
return bin(x)
numbers = [-1000, -10, 0, 1, 10, 2343]
for n in numbers:
self.checkScript(test_bin, (n,))
self.checkScript(test_oct, (n,))
self.checkScript(test_hex, (n,))
@unittest.skipIf(IS_WINDOWS or IS_SANDCASTLE, "NYI: TemporaryFileName support for Windows or Sandcastle")
def test_get_set_state(self):
class Root(torch.jit.ScriptModule):
__constants__ = ['number']
def __init__(self, number):
super(Root, self).__init__()
self.register_buffer('buffer1', torch.ones(2, 2))
self.register_buffer('buffer2', torch.ones(2, 2))
self.number = number
@torch.jit.script_method
def __getstate__(self):
return (self.buffer1, self.buffer2, 74, self.training)
@torch.jit.script_method
def __setstate__(self, state):
self.buffer1 = state[0] + 10
self.buffer2 = state[1] + 10
self.training = state[3]
class M(torch.jit.ScriptModule):
__constants__ = ['number']
def __init__(self, number, submodule):
super(M, self).__init__()
self.register_buffer('buffer1', torch.ones(2, 2))
self.register_buffer('buffer2', torch.ones(2, 2))
self.number = number
self.submodule = submodule
@torch.jit.script_method
def __getstate__(self):
return (self.buffer1, self.buffer2, 74, self.submodule, self.training)
@torch.jit.script_method
def __setstate__(self, state):
self.buffer1 = state[0] + 10
self.buffer2 = state[1] + 10
self.submodule = state[3]
self.training = state[4]
with TemporaryFileName() as fname:
m = M(23, submodule=Root(99))
m.save(fname)
loaded = torch.jit.load(fname)
# Check original module
self.assertEqual(m.buffer1, torch.ones(2, 2))
self.assertEqual(m.buffer2, torch.ones(2, 2))
# Check top level module
self.assertEqual(loaded.buffer1, torch.ones(2, 2) + 10)
self.assertEqual(loaded.buffer2, torch.ones(2, 2) + 10)
# Check submodule
self.assertEqual(loaded.submodule.buffer1, torch.ones(2, 2) + 10)
self.assertEqual(loaded.submodule.buffer2, torch.ones(2, 2) + 10)
# Check simpler module
class NoArgState(torch.nn.Module):
def __init__(self):
super(NoArgState, self).__init__()
self.register_buffer('buffer1', torch.ones(2, 2))
self.register_buffer('buffer2', torch.ones(2, 2))
def forward(self):
pass
@torch.jit.export
def __getstate__(self):
return 5, self.training
@torch.jit.export
def __setstate__(self, state):
self.buffer1 = torch.ones(2, 2) + state[0]
self.buffer2 = torch.ones(2, 2) + 10
self.training = state[1]
with TemporaryFileName() as fname:
m = torch.jit.script(NoArgState())
m.save(fname)
loaded = torch.jit.load(fname)
self.assertEqual(loaded.buffer1, torch.ones(2, 2) + 5)
self.assertEqual(loaded.buffer2, torch.ones(2, 2) + 10)
def test_string_slicing(self):
def fn1(x):
# type: (str) -> str
return x[1:3]
def fn2(x):
# type: (str) -> str
return x[-1:3]
def fn3(x):
# type: (str) -> str
return x[3:1]
def fn4(x):
# type: (str) -> str
return x[3:100]
self.checkScript(fn1, ("abcdefghi",))
self.checkScript(fn2, ("abcdefghi",))
self.checkScript(fn3, ("abcdefghi",))
self.checkScript(fn4, ("abcdefghi",))
def test_early_return_closure(self):
code = dedent('''
def tanh(self):
output = torch.tanh(self)
def backward(grad_output):
pass
return output, backward
''')
cu = torch.jit.CompilationUnit(code)
g = cu.tanh.graph
FileCheck().check_count("prim::Closure_0", 2).check("NoneType = prim::Constant") \
.check_next("return").run(g)
code = dedent('''
def tanh(self):
output = torch.tanh(self)
def backward(grad_output):
a = 1
if output:
return 1
else:
a = 2
return a
return output, backward
''')
cu = torch.jit.CompilationUnit(code)
g = cu.tanh.graph
FileCheck().check_count("prim::Closure_0", 2).check("int = prim::If") \
.run(g)
code = dedent('''
def loop_in_closure(self):
output = torch.tanh(self)
def backward(grad_output):
for i in range(3):
return 1
return 4
return output, backward
''')
cu = torch.jit.CompilationUnit(code)
fc = FileCheck()
fc.check("prim::Closure").check("(Tensor, NoneType) = prim::TupleConstruct")
# Loop then two if's added in exit transform
fc.check("prim::Closure").check("prim::Loop").check_count("prim::If", 2)
fc.run(cu.loop_in_closure.graph)
code = dedent('''
def tanh(self):
output = torch.tanh(self)
def backward(grad_output):
if 1 == 1:
return 1
else:
return 1.
return output, backward
''')
with self.assertRaisesRegex(RuntimeError, "returned a value of type int but"):
cu = torch.jit.CompilationUnit(code)
@_inline_everything
def test_early_return_fork_join(self):
@torch.jit.script
def foo(x):
if x.dim() == 2:
return torch.neg(x), x
else:
return torch.neg(x), x + 1
x = torch.rand(3, 4)
@torch.jit.script
def wait_script(x):
fut = torch.jit._fork(foo, x)
y_hat = foo(x)
y = torch.jit._wait(fut)
return y, y_hat
FileCheck().check("with prim::fork").check("prim::If").check("return")\
.run(wait_script.graph)
def test_early_return_type_refinement(self):
@torch.jit.script
def test(x):
# type: (Optional[int]) -> int
if x is None:
return 1
else:
return x
self.assertEqual(test(None), 1)
self.assertEqual(test(2), 2)
def test_exceptions_with_control_flow(self):
def test_num_ifs(func, num_ifs):
g = torch.jit.script(func).graph
FileCheck().check_count("prim::If", num_ifs, exactly=True).run(g)
def no_guard_ifs_added(x):
# type: (int) -> int
if x == 1:
return 1
else:
if x == 2:
raise RuntimeError("hi")
else:
raise RuntimeError("hi")
self.checkScript(no_guard_ifs_added, (1,))
self.checkScriptRaisesRegex(no_guard_ifs_added, (2,), Exception, "")
test_num_ifs(no_guard_ifs_added, 2)
# FUNCTION LOOKS LIKE:
# graph(%x.1 : int):
# %7 : str = prim::Constant[value="Exception"]()
# %2 : int = prim::Constant[value=1]()
# %5 : int = prim::Constant[value=2]()
# %19 : int = prim::Uninitialized()
# %3 : bool = aten::eq(%x.1, %2)
# %20 : int = prim::If(%3)
# block0():
# -> (%2)
# block1():
# %6 : bool = aten::eq(%x.1, %5)
# = prim::If(%6)
# block0():
# = prim::RaiseException(%7)
# -> ()
# block1():
# = prim::RaiseException(%7)
# -> ()
# -> (%19)
# return (%20)
def no_ifs_added(x):
# type: (int) -> int
if x < 0:
raise RuntimeError("hi")
return x
self.checkScript(no_ifs_added, (1,))
self.checkScriptRaisesRegex(no_ifs_added, (-2,), Exception, "")
test_num_ifs(no_ifs_added, 1)
def test_if_might(x):
# type: (int)
if x > 0:
if x == 1:
return 1
else:
a = 2
else:
raise RuntimeError("hi")
return a + 2
self.checkScript(test_if_might, (1,))
self.checkScript(test_if_might, (3,))
self.checkScriptRaisesRegex(no_ifs_added, (-2,), Exception, "")
test_num_ifs(test_if_might, 3) # one if added to guard a + 2
def test_loop_no_escape(x):
# type: (int)
if x >= 0:
for i in range(x):
raise RuntimeError("hi")
else:
return 5
return x + 3
self.checkScript(test_loop_no_escape, (0,))
self.checkScript(test_loop_no_escape, (-1,))
self.checkScriptRaisesRegex(test_loop_no_escape, (1,), Exception, "")
# if guard gets optimized away
test_num_ifs(test_loop_no_escape, 1)
def test_loop_exception_with_continue(x):
# type: (int)
i = 0
for i in range(5):
if i == x:
raise RuntimeError("hi")
else:
continue
print(i)
return i + 5
self.checkScript(test_loop_exception_with_continue, (-1,))
self.checkScriptRaisesRegex(test_loop_exception_with_continue, (1,), Exception, "")
test_num_ifs(test_loop_exception_with_continue, 1) # no ifs added to guard print
def test_exception_exits_closure(self):
code = dedent('''
def no_return_func(self):
# type: (Tensor) -> Tensor
output = torch.tanh(self)
def backward(grad_output):
raise RuntimeError("Hi")
''')
with self.assertRaisesRegex(RuntimeError, "does not return along all"):
cu = torch.jit.CompilationUnit(code)
code = dedent('''
def test_exit_pair_reset(x):
# type: (int) -> int
if x > 0:
a = 0
def backward(grad_output):
raise RuntimeError("Hi")
a = a + 1
else:
return x
return a + 1
''')
func = torch.jit.CompilationUnit(code).test_exit_pair_reset
self.assertEqual(func(1,), 2)
self.assertEqual(func(-1,), -1)
# final a + 1 gets inlined into the first branch and optimized away
FileCheck().check_count("prim::If", 1, exactly=True).run(func.graph)
def test_non_final_return(self):
def simple(x):
if bool(x > 3):
return x + 1
else:
return x + 2
raise RuntimeError("nope")
def nest(x):
x = x + 1
if bool(x > 3):
if bool(x > 4):
x += 1
return x + 1
else:
return x + 2
def early_ret(x):
x = x + 1
if bool(x > 3):
return x + 1
x = x + 1
return x + 2
def nest_early_ret(x):
x = x + 1
if bool(x > 3):
if bool(x > 4):
return x + 2
return x + 1
x = x + 1
return x + 2
def not_early_ret(x):
s = ""
if bool(x > 3):
if bool(x > 4):
return 1, s
s += "foo"
else:
s += "5"
s += "hi"
return 7, s
def not_total_ret(x):
s = ""
if bool(x > 3):
if bool(x > 4):
return 1, s
else:
return 2, s
else:
s += "5"
return 7, s
for i in range(3):
for func in [simple, nest, early_ret, nest_early_ret, not_early_ret,
not_total_ret]:
self.checkScript(func, (torch.tensor(2.5 + i),))
def vars_used_after_ret(x):
# type: (int) -> int
if x == 0:
return x
else:
y = 2
z = 3
return x + y * z
self.checkScript(vars_used_after_ret, (1,))
self.checkScript(vars_used_after_ret, (0,))
def complicated(x):
# type: (int) -> int
if x:
if x == 2:
return 1
assert 1 == 2
else:
if x == 3:
return 2
assert 1 == 2
else:
a = 2
b = 3
else:
a = 4
b = 1
return a + b
assert 1 == 2
for i in range(4):
self.checkScript(complicated, (i,))
def test_partial_returns(self):
with self.assertRaisesRegex(RuntimeError, "does not return along all"):
@torch.jit.script
def no_ret():
# type: () -> int
pass
with self.assertRaisesRegex(RuntimeError, "does not return along all"):
@torch.jit.script
def partial(x):
# type: (Tensor) -> int
if x:
return 1
with self.assertRaisesRegex(RuntimeError, "does not return along all"):
@torch.jit.script
def typed_none():
# type: () -> Optional[int]
pass
@torch.jit.script
def none_ret():
pass
self.assertIs(none_ret(), None)
FileCheck().check(": None").run(none_ret.graph)
def test_early_returns_loops(self):
def nest_while_ret(x):
# type: (int) -> int
y = 4
while x < 4:
if x < 3:
return y
else:
y = y + 1
break
y = y + 2
y = y + 1
return y
self.checkScript(nest_while_ret, (2,))
self.checkScript(nest_while_ret, (3,))
self.checkScript(nest_while_ret, (4,))
def loop_ret(x, y):
# type: (int, int) -> (int)
i = 0
for i in range(x):
if x == y:
return x + y
i = i + y
i = i - 1
return i
self.checkScript(loop_ret, (3, 3))
self.checkScript(loop_ret, (2, 3))
self.checkScript(loop_ret, (3, 1))
def test_will_ret(y):
# type: (int) -> int
for i in range(y):
return 2
return 1
self.checkScript(test_will_ret, (0,))
self.checkScript(test_will_ret, (1,))
def test_loop_nest_ret(y):
# type: (int) -> int
for i in range(y):
for i in range(y - 2):
return 10
return 5
return 0
self.checkScript(test_loop_nest_ret, (0,))
self.checkScript(test_loop_nest_ret, (1,))
self.checkScript(test_loop_nest_ret, (2,))
def test_nn_init(self):
tests = (
('constant_', (lambda: (torch.ones(2, 2), 2.5)), "Tensor, float"),
('ones_', (lambda: (torch.ones(2, 2),)), "Tensor"),
('zeros_', (lambda: (torch.ones(2, 2),)), "Tensor"),
('uniform_', (lambda: (torch.ones(2, 2),)), "Tensor"),
('normal_', (lambda: (torch.ones(2, 2),)), "Tensor"),
('xavier_normal_', (lambda: (torch.ones(2, 2),)), "Tensor"),
('xavier_uniform_', (lambda: (torch.ones(2, 2),)), "Tensor"),
)
for name, args_fn, type_str in tests:
# Build test code
arg_str = ', '.join([chr(i + ord('a')) for i in range(len(args_fn()))])
code = dedent('''
def test({arg_str}):
# type: ({type_str})
return torch.nn.init.{name}({arg_str})
''').format(arg_str=arg_str, type_str=type_str, name=name)
cu = torch.jit.CompilationUnit(code)
# Compare functions
init_fn = getattr(torch.nn.init, name)
script_out = self.runAndSaveRNG(cu.test, args_fn())
eager_out = self.runAndSaveRNG(init_fn, args_fn())
self.assertEqual(script_out, eager_out)
FileCheck().check_not("prim::PythonOp").run(cu.test.graph)
def test_early_return_rewrite(self):
def test_foo(x: bool):
if x:
return 1
return 2
self.checkScript(test_foo, (True,))
self.checkScript(test_foo, (False,))
FileCheck().check_count("prim::If", 1, exactly=True).run(torch.jit.script(test_foo).graph)
def test_multiple(x: int):
if x == 5:
return x * x
else:
y = 2 * x
z = y * 2
if z == 8:
return 1
if z != 16:
z = z - 2
abc = 4
else:
return 3
z = z * abc
return z * z * z
self.checkScript(test_multiple, (5,))
self.checkScript(test_multiple, (2,))
self.checkScript(test_multiple, (4,))
self.checkScript(test_multiple, (3,))
self.checkScript(test_multiple, (10,))
graph = torch.jit.script(test_multiple).graph
FileCheck().check_count("prim::If", 3, exactly=True).run(graph)
def test_is_scripting_metacompile(self):
@torch.jit.script
def foo():
if torch.jit.is_scripting():
return 1
else:
print("hello") + 2 # will not be compiled
self.assertEqual(foo(), 1)
def test_boolean_literal_constant_metacompile(self):
class Mod(torch.nn.Module):
__constants__ = ['val']
def __init__(self, val):
super(Mod, self).__init__()
self.val = val
def forward(self):
if self.val:
return 1
else:
return "2"
self.checkModule(Mod(True), ())
self.checkModule(Mod(False), ())
@torch.jit.script
def foo():
if True:
return 1
else:
return "2"
self.assertEqual(foo(), 1)
def test_assert_is_scripting_metacompile(self):
def foo():
assert not torch.jit.is_scripting(), "TestErrorMsg"
print("hello") + 2 # will not be compiled
f = torch.jit.script(foo)
with self.assertRaisesRegex(torch.jit.Error, "TestErrorMsg"):
f()
def test_isinstance_metacompile(self):
@torch.jit.script
def test_primitive_type(x):
# type: (int) -> int
if isinstance(x, int):
return x + 1
else:
return x - 1
self.assertEqual(test_primitive_type(1), 2)
with self.assertRaisesRegex(Exception, "Expected a value of type"):
test_primitive_type(1.5)
_MyNamedTuple = namedtuple('_MyNamedTuple', ['value'])
@torch.jit.script
def test_non_primitive_types(x):
# type: (_MyNamedTuple) -> Tensor
if isinstance(1, _MyNamedTuple):
return 10
if isinstance(x, _MyNamedTuple):
return x.value + 1
else:
return 1
out = test_non_primitive_types(_MyNamedTuple(value=torch.tensor(5.0)))
self.assertEqual(out, torch.tensor(6.0))
def test_namedtuple_type_inference(self):
_AnnotatedNamedTuple = NamedTuple('_NamedTupleAnnotated', [('value', int)])
_UnannotatedNamedTuple = namedtuple('_NamedTupleUnAnnotated', ['value'])
def test_check_named_tuple_value():
named_tuple = _AnnotatedNamedTuple(1)
return named_tuple.value
self.checkScript(test_check_named_tuple_value, ())
def test_error():
return _UnannotatedNamedTuple(1)
with self.assertRaisesRegex(RuntimeError, r"Expected a value of type \'Tensor \(inferred\)\' "
r"for argument \'value\' but instead found type \'int\'."):
torch.jit.script(test_error)
def test_namedtuple_default_values_simple_type(self):
class Point(NamedTuple):
x: Optional[int] = None
y: int = 2
make_global(Point)
class M(torch.nn.Module):
def __init__(self):
super(M, self).__init__()
def forward(self, point: Point):
return point
p = Point(x=3, y=2)
self.checkModule(M(), (p,))
self.checkModule(M(), (Point(),))
m = torch.jit.script(M())
FileCheck().check(r"NamedTuple(x : int? = None, y : int = 2))") \
.run(m.graph)
def test_namedtuple_default_values_missing(self):
class Point(NamedTuple):
x: Optional[int]
y: int
z: int = 3
make_global(Point)
class M(torch.nn.Module):
def __init__(self):
super(M, self).__init__()
def forward(self, point: Point):
return point
p1 = Point(x=3, y=2)
p2 = Point(x=3, y=2, z=1)
self.checkModule(M(), (p1,))
self.checkModule(M(), (p2,))
m = torch.jit.script(M())
FileCheck().check(r"NamedTuple(x : int?, y : int, z : int = 3))") \
.run(m.graph)
def test_namedtuple_default_values_container_type(self):
class Point(NamedTuple):
x: Optional[List[int]] = None
y: List[int] = [1, 2, 3]
z: Optional[Dict[str, int]] = {"a": 1}
make_global(Point)
class M(torch.nn.Module):
def __init__(self):
super(M, self).__init__()
def forward(self, point: Point):
return point
p = Point(x=[4, 5, 6], y=[3, 2, 1], z={"b": 2})
self.checkModule(M(), (p,))
self.checkModule(M(), (Point(),))
m = torch.jit.script(M())
first_line = r"NamedTuple(x : int[]? = None, y : int[] = " \
r"[1, 2, 3], z : Dict(str, int)? = {a: 1}))"
FileCheck().check(first_line) \
.run(m.graph)
def test_namedtuple_default_values_Tensor_type(self):
class Point(NamedTuple):
x: torch.Tensor = torch.rand(2, 3)
make_global(Point)
class M(torch.nn.Module):
def __init__(self):
super(M, self).__init__()
def forward(self, point: Point):
return point
p = Point(x=torch.rand(2, 3))
with self.assertRaisesRegex(RuntimeError, "Tensors are not "
"supported as default NamedTuple "
"fields"):
m = torch.jit.script(M())
m(p)
@unittest.skipIf(sys.version_info < (3, 7, 0), "defaults keyword added in Python 3.8")
def test_namedtuple_default_values_using_factory_constructor(self):
Pair = namedtuple("Pair", ["x", "y"], defaults=(1, 2))
make_global(Pair)
@torch.jit.script
def fn(x: Pair) -> Pair:
return x
# TODO: We can't use `checkScript` with the NamedTuple factory
# constructor. Using the factory constructor with TorchScript
# TorchScript creates an anonymous `NamedTuple` class instead of
# preserving the actual name. For example, the actual generated
# signature in this case is:
# graph(%x.1 : NamedTuple(x : Tensor, y : Tensor))
# It looks like similar test cases have had this issue as well
# (see: `test_namedtuple_python`).
FileCheck().check(r"NamedTuple(x : Tensor = 1, y : Tensor = 2))") \
.check_next(r"return (%x.1)") \
.run(fn.graph)
def test_isinstance_dynamic(self):
@torch.jit.script
def foo(a):
# type: (Optional[List[int]]) -> int
b = 0
if isinstance(a, (int, (float,), list, str)):
b += 1
if isinstance(a, (int, str)):
b += 1
if isinstance(a, List[int]):
b += 1
return b
self.assertEqual(foo([3, 4]), 2)
self.assertEqual(foo(None), 0)
def test_function_overloads(self):
# TODO: pyflakes currently does not compose @overload annotation with other
# decorators. This is fixed on master but not on version 2.1.1.
# Next version update remove noqa and add @typing.overload annotation
@torch.jit._overload # noqa: F811
def test_simple(x1): # noqa: F811
# type: (int) -> int
pass
@torch.jit._overload # noqa: F811
def test_simple(x1): # noqa: F811
# type: (float) -> float
pass
def test_simple(x1): # noqa: F811
return x1
def invoke_function():
return test_simple(1.0), test_simple(.5)
self.checkScript(invoke_function, ())
# testing that the functions are cached
compiled_fns_1 = torch.jit._script._get_overloads(test_simple)
compiled_fns_2 = torch.jit._script._get_overloads(test_simple)
for a, b in zip(compiled_fns_1, compiled_fns_2):
self.assertIs(a.graph, b.graph)
old_func = test_simple
# testing that new functions added work with caching
@torch.jit._overload # noqa: F811
def test_simple(x1): # noqa: F811
# type: (str) -> str
pass
@torch.jit.script
def my_func():
return old_func("hi")
# testing new function same qualified name
@torch.jit._overload # noqa: F811
def test_simple(a, b): # noqa: F811
# type: (int, int) -> int
pass
def test_simple(a, b):
return a + b
@torch.jit.script
def fn():
return test_simple(3, 4)
self.assertEqual(fn(), 7)
# currently we take the default values have to be specified in the
# overload as well - TODO take them from implementation and apply
# where the type is valid.
@torch.jit._overload # noqa: F811
def identity(x1): # noqa: F811
# type: (str) -> str
pass
@torch.jit._overload # noqa: F811
def identity(x1): # noqa: F811
# type: (float) -> float
pass
def identity(x1=1.0): # noqa: F811
return x1
def invoke():
return identity(), identity(.5), identity("hi")
self.checkScript(invoke, ())
def schema_match_failure():
return identity((1, 2))
thrown = False
try:
torch.jit.script(schema_match_failure)
except Exception as e:
thrown = True
self.assertTrue(r"of type 'str'" in str(e) and r"of type 'float" in str(e))
self.assertTrue(thrown)
with self.assertRaisesRegex(Exception, "cannot be directly compiled"):
torch.jit.script(identity)
@torch.jit._overload # noqa: F811
def impl_compile_failure(x, y): # noqa: F811
# type: (str, str) -> (str)
pass
@torch.jit._overload # noqa: F811
def impl_compile_failure(x, y): # noqa: F811
# type: (int, int) -> (int)
pass
def impl_compile_failure(x, y): # noqa: F811
return x - y
def test():
impl_compile_failure("one", "two")
with self.assertRaisesRegex(Exception, "Arguments for call are not valid"):
torch.jit.script(test)
@torch.jit._overload # noqa: F811
def good_overload(x=1): # noqa: F811
# type: (int) -> (int)
pass
def good_overload(x=1): # noqa: F811
return x
@torch.jit.script
def foo():
return good_overload()
self.assertEqual(foo(), 1)
with self.assertRaisesRegex(Exception, "must equal to the default parameter"):
@torch.jit._overload # noqa: F811
def bad_default_on_overload(x, y=2): # noqa: F811
# type: (int, int) -> (int)
pass
def bad_default_on_overload(x, y=1): # noqa: F811
# type: (int, int) -> (int)
pass
@torch.jit.script
def test():
return bad_default_on_overload(1, 2)
@torch.jit._overload # noqa: F811
def diff_default(x): # noqa: F811
# type: (int) -> int
pass
@torch.jit._overload # noqa: F811
def diff_default(x): # noqa: F811
# type: (str) -> str
pass
def diff_default(x="hi"): # noqa: F811
return x
def test():
return diff_default(), diff_default(2), diff_default("abc")
self.assertEqual(test(), torch.jit.script(test)())
@torch.jit._overload # noqa: F811
def diff_num_params(x): # noqa: F811
# type: (float) -> float
pass
@torch.jit._overload # noqa: F811
def diff_num_params(x, y): # noqa: F811
# type: (int, int) -> int
pass
def diff_num_params(x, y=2, z=3): # noqa: F811
# type: (Union[float, int], int, int)
return x + y + z
def test():
return diff_num_params(1.0), diff_num_params(1, 2), diff_num_params(1), diff_num_params(1, 2, 3)
self.assertEqual(test(), torch.jit.script(test)())
@torch.jit._overload # noqa: F811
def diff_num_params_no_annot():
# type: () -> int
pass
def diff_num_params_no_annot(x=1): # noqa: F811
return x
def test():
return diff_num_params_no_annot(1.0)
with self.assertRaisesRegex(Exception, "Parameters not specified"):
torch.jit.script(test)
def test_function_overload_misuse(self):
with self.assertRaisesRegex(RuntimeError, "Only `pass` statement or `...` can be the body"):
@torch.jit._overload
def wrong_decl_body(x: str) -> str:
return x + "0"
with self.assertRaisesRegex(RuntimeError, "Only `pass` statement or `...` can be the body"):
class MyClass:
@torch.jit._overload_method
def method(self):
return 0
@torch.jit._overload
def null_overload(x: int) -> int: ... # noqa: E704
@torch.jit._overload # noqa: F811
def null_overload(x: str) -> str: # noqa: F811
pass
def null_overload_driver():
return null_overload(0)
with self.assertRaisesRegex(RuntimeError, 'Implementation for the function ".+" is missing.'):
torch.jit.script(null_overload_driver)
class OverloadMisuse(torch.nn.Module):
def __init__(self):
super().__init__()
@torch.jit._overload_method
def forward(self, x: int):
pass
@torch.jit._overload_method # noqa: F811
def forward(self, x: Tensor): # noqa: F811
pass
with self.assertRaisesRegex(RuntimeError, 'Implementation for the method ".+" is missing.'):
m = torch.jit.script(OverloadMisuse())
def test_script_method_torch_function_overload(self):
class MyCustomTensor(torch.Tensor):
pass
class MyCustomModule(torch.nn.Module):
def forward(self, x):
return torch.relu(x)
scripted_mod = torch.jit.script(MyCustomModule())
t = torch.tensor([3.0])
ref_out = scripted_mod(t)
t_custom = MyCustomTensor([3.0])
out1 = scripted_mod(t_custom)
self.assertEqual(out1, ref_out)
out2 = scripted_mod.forward(t_custom)
self.assertEqual(out2, ref_out)
def test_function_overloading_isinstance(self):
@torch.jit._overload # noqa: F811
def my_conv(x, y): # noqa: F811
# type: (float, str) -> (float)
pass
@torch.jit._overload # noqa: F811
def my_conv(x, y): # noqa: F811
# type: (float, float) -> (float)
pass
def my_conv(x, y=2.0): # noqa: F811
if isinstance(y, str):
if y == "hi":
return 4.0 - x
else:
return 5.0 - x
else:
return 2.0 + x
def test_uses():
return my_conv(1.5), my_conv(1.5, "hi"), my_conv(1.5, 5.0)
self.checkScript(test_uses, ())
def test_method_overloading(self):
class Over(torch.nn.Module):
def __init__(self):
super(Over, self).__init__()
@torch.jit._overload_method # noqa: F811
def forward(self, x): # noqa: F811
# type: (Tuple[Tensor, Tensor]) -> Tensor
pass
@torch.jit._overload_method # noqa: F811
def forward(self, x): # noqa: F811
# type: (Tensor) -> Tensor
pass
def forward(self, x): # noqa: F811
if isinstance(x, Tensor):
return x + 20
else:
return x[0] + 5
class S(torch.jit.ScriptModule):
def __init__(self):
super(S, self).__init__()
self.weak = Over()
@torch.jit.script_method
def forward(self, x):
return self.weak(x) + self.weak((x, x))
s_mod = S()
x = torch.ones(1)
self.assertEqual(s_mod(x), x + 20 + 5 + x)
over = Over()
self.assertEqual(over((x, x)), x + 5)
self.assertEqual(over((x)), x + 20)
class Unannotated(torch.nn.Module):
def __init__(self):
super(Unannotated, self).__init__()
@torch.jit._overload_method # noqa: F811
def hello(self, x): # noqa: F811
pass
@torch.jit._overload_method # noqa: F811
def hello(self, x): # noqa: F811
# type: (int) -> (int)
pass
def hello(self, x): # noqa: F811
return x + 3
def forward(self):
return self.hello(1), self.hello(.5)
w = Unannotated()
with self.assertRaisesRegex(Exception, "explicitly add type annotations to overloaded functions"):
torch.jit.script(w)
class CompileOverloadError(torch.nn.Module):
def __init__(self):
super(CompileOverloadError, self).__init__()
@torch.jit._overload_method # noqa: F811
def hello(self, x): # noqa: F811
# type: (str) -> (int)
pass
@torch.jit._overload_method # noqa: F811
def hello(self, x): # noqa: F811
# type: (int) -> (int)
pass
def hello(self, x): # noqa: F811
return x + 1
def forward(self):
return self.hello("hi"), self.hello(.5)
w = CompileOverloadError()
with self.assertRaisesRegex(Exception, "but instead found type \'str\'"):
torch.jit.script(w)
# testing overload declared first, then non-overload
with self.assertRaisesRegex(Exception, "Overloads are not useable when a module"):
class W3(torch.nn.Module):
def __init__(self):
super(W3, self).__init__()
@torch.jit._overload_method # noqa: F811
def forward(self, x): # noqa: F811
# type: (int) -> int
pass
@torch.jit._overload_method # noqa: F811
def forward(self, x): # noqa: F811
# type: (Tensor) -> Tensor
pass
def forward(self, x): # noqa: F811
return x + 5
a = W3()
b = torch.jit.script(a)
class W3(torch.nn.Module):
def __init__(self):
super(W3, self).__init__()
def forward(self, x): # noqa: F811
return x + 5 + 10
a = W3()
b = torch.jit.script(a)
# testing non-overload declared first, then overload
class W2(torch.nn.Module):
def __init__(self):
super(W2, self).__init__()
def hello(self, x1, x2):
return x1 + x2
def forward(self, x):
return self.hello(x, x)
a = torch.jit.script(W2())
self.assertEqual(a(torch.tensor(1)), torch.tensor(2))
class W2(torch.nn.Module):
def __init__(self):
super(W2, self).__init__()
@torch.jit._overload_method # noqa: F811
def hello(self, x): # noqa: F811
pass
@torch.jit._overload_method # noqa: F811
def hello(self, x): # noqa: F811
# type: (int) -> (int)
pass
def hello(self, x): # noqa: F811
return x + 5 + 10
def forward(self, x):
return self.hello(1), self.hello(x)
with self.assertRaisesRegex(Exception, "Overloads are not useable when a module"):
a = torch.jit.script(W2())
def test_narrow_copy(self):
def foo(a):
return a.narrow_copy(0, 0, 5)
self.checkScript(foo, [torch.rand(10)])
def test_select_after_chunk(self):
def foo(x):
chunked = torch.chunk(x, 1)
foo = chunked[0]
foo.add_(5)
return x
self.checkScript(foo, [torch.rand(2, 3)])
def test_nn_LSTM_with_layers(self):
class M(torch.jit.ScriptModule):
def __init__(self):
super(M, self).__init__()
self.rnn = nn.LSTM(2, 3, 2, dropout=0)
@torch.jit.script_method
def forward(self, x, lengths, h0, c0):
return self.rnn(x, (h0, c0))[0]
class Eager(torch.nn.Module):
def __init__(self):
super(Eager, self).__init__()
self.rnn = nn.LSTM(2, 3, 2, dropout=0)
def forward(self, x, lengths, h0, c0):
return self.rnn(x, (h0, c0))[0]
inputs = (torch.randn(1, 1, 2), torch.LongTensor([7]), torch.randn(2, 1, 3), torch.randn(2, 1, 3))
eager_out = self.runAndSaveRNG(lambda: Eager()(*inputs), ())[0]
script_out = self.runAndSaveRNG(lambda: M()(*inputs), ())[0]
self.assertEqual(eager_out, script_out)
def test_nn_LSTM(self):
input = torch.nn.utils.rnn.pack_sequence([torch.randn(5, 5)])
class S(torch.jit.ScriptModule):
def __init__(self):
super(S, self).__init__()
self.x = torch.nn.LSTM(5, 5)
@torch.jit.script_method
def forward(self, input: PackedSequence) -> Tuple[PackedSequence, Tuple[torch.Tensor, torch.Tensor]]:
return self.x(input)
eager_out = self.runAndSaveRNG(lambda x: torch.nn.LSTM(5, 5)(x), (input,))[0]
script_out = self.runAndSaveRNG(lambda x: S()(x), (input,))[0]
self.assertEqual(eager_out, script_out)
def test_nn_GRU(self):
seq_input = torch.nn.utils.rnn.pack_sequence([torch.randn(5, 5)])
tensor_input = torch.randn(5, 5, 5)
class SeqLengthGRU(torch.jit.ScriptModule):
def __init__(self):
super(SeqLengthGRU, self).__init__()
self.x = torch.nn.GRU(5, 5)
@torch.jit.script_method
def forward(self, input: PackedSequence) -> Tuple[PackedSequence, torch.Tensor]:
return self.x(input)
class TensorGRU(torch.jit.ScriptModule):
def __init__(self):
super(TensorGRU, self).__init__()
self.x = torch.nn.GRU(5, 5)
@torch.jit.script_method
def forward(self, input: torch.Tensor) -> Tuple[torch.Tensor, torch.Tensor]:
return self.x(input)
seq_eager_out = self.runAndSaveRNG(lambda x: torch.nn.GRU(5, 5)(x), (seq_input,))[0]
seq_script_out = self.runAndSaveRNG(lambda x: SeqLengthGRU()(x), (seq_input,))[0]
tensor_eager_out = self.runAndSaveRNG(lambda x: torch.nn.GRU(5, 5)(x), (tensor_input,))[0]
tensor_script_out = self.runAndSaveRNG(lambda x: TensorGRU()(x), (tensor_input,))[0]
self.assertEqual(seq_eager_out, seq_script_out)
self.assertEqual(tensor_eager_out, tensor_script_out)
def test_torchscript_memoryformat(self):
@torch.jit.script
def fn(x):
return x.contiguous(memory_format=torch.channels_last)
x = torch.randn(4, 3, 6, 6)
y = fn(x)
self.assertTrue(y.is_contiguous(memory_format=torch.channels_last))
def test_torchscript_multi_head_attn(self):
@torch.jit.script
def jit_multihead_attn_forward(query, # type: Tensor
key, # type: Tensor
value, # type: Tensor
embed_dim_to_check, # type: int
num_heads, # type: int
in_proj_weight, # type: Tensor
in_proj_bias, # type: Tensor
bias_k, # type: Optional[Tensor]
bias_v, # type: Optional[Tensor]
add_zero_attn, # type: bool
dropout, # type: float
out_proj_weight, # type: Tensor
out_proj_bias, # type: Tensor
training=True, # type: bool
key_padding_mask=None, # type: Optional[Tensor]
need_weights=True, # type: bool
attn_mask=None # type: Optional[Tensor]
):
# type: (...) -> Tuple[Tensor, Optional[Tensor]]
return torch.nn.functional.multi_head_attention_forward(query, key, value,
embed_dim_to_check, num_heads,
in_proj_weight, in_proj_bias,
bias_k, bias_v,
add_zero_attn, dropout,
out_proj_weight, out_proj_bias,
training, key_padding_mask,
need_weights, attn_mask)
src_l = 3
bsz = 5
embed_size = 8
nhead = 2
multi_head_attn = torch.nn.MultiheadAttention(embed_size, nhead)
query = torch.rand((src_l, bsz, embed_size))
key = torch.rand((src_l, bsz, embed_size))
value = torch.rand((src_l, bsz, embed_size))
mask = (torch.triu(torch.ones(src_l, src_l)) == 1).transpose(0, 1)
mask = mask.float().masked_fill(mask == 0, float('-inf')).masked_fill(mask == 1, float(0.0)).double()
jit_out = jit_multihead_attn_forward(query, key, value,
embed_size, nhead,
multi_head_attn.in_proj_weight,
multi_head_attn.in_proj_bias,
multi_head_attn.bias_k, multi_head_attn.bias_v,
multi_head_attn.add_zero_attn, multi_head_attn.dropout,
multi_head_attn.out_proj.weight,
multi_head_attn.out_proj.bias, attn_mask=mask)[0]
py_out = torch.nn.functional.multi_head_attention_forward(query, key, value,
embed_size, nhead,
multi_head_attn.in_proj_weight,
multi_head_attn.in_proj_bias,
multi_head_attn.bias_k,
multi_head_attn.bias_v,
multi_head_attn.add_zero_attn,
multi_head_attn.dropout,
multi_head_attn.out_proj.weight,
multi_head_attn.out_proj.bias,
attn_mask=mask)[0]
# print("rel. error: ")
# print(jit_out / py_out - 1)
self.assertEqual(jit_out, py_out, atol=5e-4, rtol=1e-4)
def test_torchscript_multi_head_attn_fast_path(self):
src_l = 3
bsz = 5
embed_size = 8
nhead = 2
multi_head_attn = torch.nn.MultiheadAttention(embed_size, nhead, batch_first=True)
multi_head_attn = multi_head_attn.eval()
query = key = value = torch.rand((bsz, src_l, embed_size))
with torch.no_grad():
py_out = multi_head_attn(query, key, value)
mha = torch.jit.script(multi_head_attn)
jit_out = mha(query, key, value)
torch.testing.assert_close(jit_out, py_out)
@unittest.skipIf(not RUN_CUDA, "no CUDA")
def test_scriptmodule_multi_head_attn_cuda(self):
class MyModule(torch.jit.ScriptModule):
def __init__(self, embed_dim, num_heads):
super(MyModule, self).__init__()
sample_q = torch.randn(3, 2, embed_dim)
sample_kv = torch.randn(3, 2, embed_dim)
attention = nn.MultiheadAttention(embed_dim, num_heads)
attention.eval()
self.mod = torch.jit.trace(attention,
(sample_q, sample_kv, sample_kv))
@torch.jit.script_method
def forward(self, q, k, v):
return self.mod(q, k, v)
embed_dim = 8
num_heads = 2
sl = 3
bs = 2
model = MyModule(embed_dim, num_heads).cuda()
q = torch.randn(sl, bs, embed_dim, device="cuda")
kv = torch.randn(sl, bs, embed_dim, device="cuda")
jit_out = model(q, kv, kv)[0]
py_out = torch.nn.functional.multi_head_attention_forward(q, kv, kv,
embed_dim, num_heads,
model.mod.in_proj_weight,
model.mod.in_proj_bias,
None, None, None, 0.0,
model.mod.out_proj.weight,
model.mod.out_proj.bias)[0]
self.assertEqual(jit_out, py_out, atol=5e-4, rtol=1e-4)
@unittest.skipIf(not RUN_CUDA, "no CUDA")
def test_scriptmodule_transformer_cuda(self):
class MyModule(torch.jit.ScriptModule):
def __init__(self, transformer, sample_q, sample_kv):
super(MyModule, self).__init__()
transformer.eval()
self.mod = torch.jit.trace(transformer,
(sample_q, sample_kv))
@torch.jit.script_method
def forward(self, q, k):
return self.mod(q, k)
d_model = 8
nhead = 2
num_encoder_layers = 2
num_decoder_layers = 2
dim_feedforward = 16
bsz = 2
seq_length = 5
tgt_length = 3
src = torch.randn(seq_length, bsz, d_model)
tgt = torch.randn(tgt_length, bsz, d_model)
transformer = nn.Transformer(d_model, nhead, num_encoder_layers,
num_decoder_layers, dim_feedforward, dropout=0.0)
model = MyModule(transformer, tgt, src)
src = torch.randn(seq_length, bsz, d_model)
tgt = torch.randn(tgt_length, bsz, d_model)
jit_out = model(tgt, src)
py_out = transformer(tgt, src)
# print(jit_out/py_out-1)
# print(torch.allclose(jit_out, py_out, atol=5e-4, rtol=1e-4))
self.assertEqual(jit_out, py_out, atol=5e-4, rtol=1e-4)
def test_list_python_op(self):
def python_list_op(lst):
# type: (List[Tensor]) -> Tensor
return lst[0]
def fn(lst):
# type: (List[Tensor]) -> Tensor
return python_list_op(lst)
self.checkScript(fn, ([torch.ones(2) + 2, torch.ones(2)],))
@unittest.skipIf(not RUN_CUDA, "no CUDA")
def test_weak_cuda(self):
class M(torch.jit.ScriptModule):
def __init__(self):
super(M, self).__init__()
self.lstm = torch.nn.LSTM(5, 5)
self.lstm.cuda()
@torch.jit.script_method
def forward(self, x):
return self.lstm(x)
m = M()
m.cuda()
out = m(torch.ones(5, 5, 5).cuda())
self.assertTrue(out[0].is_cuda)
def test_ignore_decorator(self):
with warnings.catch_warnings(record=True) as warns:
class M(torch.jit.ScriptModule):
def __init__(self):
super(M, self).__init__()
tensor = torch.zeros(1, requires_grad=False)
self.register_buffer('some_state', torch.nn.Parameter(tensor))
@torch.jit.script_method
def forward(self, x):
self.ignored_code(x)
return x
@torch.jit.ignore(drop_on_export=True)
def ignored_code(self, x):
self.some_state = torch.tensor((100,))
FileCheck().check("TorchScript will now drop the function").run(str(warns[0]))
# Assert ignored code is run
m = M()
m2 = self.getExportImportCopy(m)
pp = str(m2.forward.code)
self.assertNotIn('ignored_code', pp)
with self.assertRaisesRegex(torch.jit.Error, "annotated to be ignored and cannot be run"):
m2.forward(torch.ones(1))
def test_ignored_as_value(self):
class Model(nn.Module):
def __init__(self):
super(Model, self).__init__()
@torch.jit.unused
def tuple_ignored(self, x):
# type: (Tensor) -> Tuple[Tensor, Tensor]
return x, x
@torch.jit.unused
def single_val_ignored(self, x, y):
# type: (Tensor, Tensor) -> Tensor
return x
def forward(self, x, use_ignore_path):
# type: (Tensor, bool) -> Tuple[Tensor, Tensor]
if 1 == 2:
return self.tuple_ignored(x)
if use_ignore_path:
return self.single_val_ignored(x, x), self.single_val_ignored(x, x)
return x, x
original = Model()
scripted = torch.jit.script(original)
self.assertEqual(scripted(torch.tensor(.5), False), (torch.tensor(.5), torch.tensor(.5)))
buffer = io.BytesIO()
torch.jit.save(scripted, buffer)
buffer.seek(0)
loaded = torch.jit.load(buffer)
with self.assertRaisesRegex(torch.jit.Error, "annotated to be ignored and cannot be run"):
loaded(torch.tensor(.5), True)
def test_module_error(self):
class MyModule(torch.nn.Module):
def __init__(self):
super(MyModule, self).__init__()
def forward(self, foo):
return foo
with self.assertRaisesRegex(RuntimeError, "cannot be compiled since it inherits from nn.Module"):
torch.jit.script(MyModule)
def test_view_write(self):
def fn(x, y):
l = []
l.append(x)
x_view = l[0]
a = x + x
x_view.add_(y)
b = x + x
return a == b
self.checkScript(fn, (torch.rand(2, 3), torch.rand(2, 3)))
def test_module_attrs(self):
class M(torch.jit.ScriptModule):
def __init__(self, table):
super(M, self).__init__()
self.table = torch.jit.Attribute(table, Dict[str, torch.Tensor])
self.x = torch.nn.Parameter(torch.tensor([100.0]))
@torch.jit.script_method
def forward(self, key):
# type: (str) -> Tensor
return self.table[key] + self.x
with torch._jit_internal._disable_emit_hooks():
# TODO: re-enable module hook when Python printing of attributes is
# supported
m = M({char : torch.ones(1) + ord(char) - ord("a") for char in "abcdefg"})
self.assertEqual(m("c"), torch.tensor([103.]))
def test_module_none_attrs(self):
class MyMod(torch.jit.ScriptModule):
def __init__(self):
super(MyMod, self).__init__()
self.optional_value = None
@torch.jit.script_method
def forward(self):
return self.optional_value
graph = MyMod().forward.graph
FileCheck().check("prim::GetAttr").run(graph)
self.run_pass('peephole', graph)
FileCheck().check_not("prim::GetAttr").run(graph)
def test_tensor_import_export(self):
@torch.jit.script
def foo(x):
a = torch.tensor(1)
b = torch.tensor([1, 2])
c = [a, b]
return c
self.run_pass('constant_propagation', foo.graph)
m = self.createFunctionFromGraph(foo.graph)
self.getExportImportCopy(m)
def get_pickle_values(self):
return (('dict', {"I": "am", "a test": "test"}, Dict[str, str]),
('float', 2.3, float),
('int', 99, int),
('bool', False, bool),
('tuple', (1, 2, 3, 4), Tuple[int, int, int, int]),
('list', [(1, 2), (3, 4)], List[Tuple[int, int]]),
('tensor', torch.randn(2, 2), torch.Tensor),
('int_list', [1, 2, 3, 4], List[int]),
('tensor_list', [torch.ones(2, 2) + i for i in range(4)], List[torch.Tensor]),
('bool_list', [True, True, False, True], List[bool]),
('float_list', [1., 2., 3., 4.], List[float]),
('str_list', ['hello', 'bye'], List[str]),
('none', None, Optional[int]),
('a_device', torch.device('cpu'), torch.device),
('another_device', torch.device('cuda:1'), torch.device))
def test_attribute_serialization(self):
tester = self
class M(torch.jit.ScriptModule):
def __init__(self):
super(M, self).__init__()
for name, value, the_type in tester.get_pickle_values():
setattr(self, name, torch.jit.Attribute(value, the_type))
@torch.jit.script_method
def forward(self):
return (self.dict, self.float, self.int, self.bool, self.tuple,
self.list, self.int_list, self.tensor_list, self.bool_list,
self.float_list, self.str_list, self.none)
m = M()
imported_m = self.getExportImportCopy(m)
self.assertEqual(m(), imported_m())
def test_string_len(self):
def fn(x):
# type: (str) -> int
return len(x)
self.checkScript(fn, ("",))
self.checkScript(fn, ("h",))
self.checkScript(fn, ("hello",))
def test_multiline_optional_future_refinement(self):
@torch.jit.script
def fun() -> int:
future: Optional[
torch.jit.Future[Tuple[torch.Tensor]]
] = None
return 1
self.assertEqual(fun(), 1)
@unittest.skipIf(IS_WINDOWS or IS_SANDCASTLE, "NYI: TemporaryFileName support for Windows or Sandcastle")
def test_attribute_unpickling(self):
tensor = torch.randn(2, 2)
tester = self
class M(torch.jit.ScriptModule):
def __init__(self):
super(M, self).__init__()
for name, value, the_type in tester.get_pickle_values():
setattr(self, "_" + name, torch.jit.Attribute(value, the_type))
@torch.jit.script_method
def forward(self):
return (self._dict, self._float, self._int, self._bool, self._tuple,
self._list, self._int_list, self._tensor_list, self._bool_list,
self._float_list, self._str_list, self._none)
with TemporaryFileName() as fname:
M().save(fname)
loaded = torch.jit.load(fname)
def is_tensor_value(item):
if isinstance(item, torch.Tensor):
return True
if isinstance(item, list):
return is_tensor_value(item[0])
return False
for name, value, the_type in self.get_pickle_values():
if is_tensor_value(value):
continue
self.assertEqual(value, getattr(loaded, "_" + name))
@unittest.skipIf(IS_WINDOWS or IS_SANDCASTLE, "NYI: TemporaryFileName support for Windows or Sandcastle")
@unittest.skipIf(not BUILD_WITH_CAFFE2, "PyTorch is build without Caffe2 support")
def test_old_models_bc(self):
model = {
'archive/version': b'1',
'archive/code/archive.py':
b'''
op_version_set = 0
def forward(self,
_0: Tensor) -> Tensor:
_1 = torch.zeros([10], dtype=6, layout=0, device=torch.device("cpu"))
result = torch.to(torch.fill_(_1, 5), dtype=6, layout=0, device=torch.device("cpu"),
non_blocking=False, copy=False)
result2 = torch.rand([10], dtype=6, layout=0, device=torch.device("cpu"))
result3 = torch.rand_like(result2, dtype=6, layout=0, device=torch.device("cpu"))
_2 = torch.add(torch.add(result, result2, alpha=1), result3, alpha=1)
return _2
''',
'archive/attributes.pkl': b'\x80\x02](e.',
'archive/libs.py': b'op_version_set = 0\n',
'archive/model.json':
b'''
{
"protoVersion":"2",
"mainModule":{
"torchscriptArena":{
"key":"code/archive.py"
},
"name":"archive",
"optimize":true
},
"producerName":"pytorch",
"producerVersion":"1.0",
"libs":{
"torchscriptArena":{
"key":"libs.py"
}
}
}'''}
with TemporaryFileName() as fname:
archive_name = os.path.basename(os.path.normpath(fname))
with zipfile.ZipFile(fname, 'w') as archive:
for k, v in model.items():
archive.writestr(k, v)
with open(fname, "rb") as f:
fn = torch.jit.load(f)
x = torch.zeros(10)
fn(x)
def test_submodule_attribute_serialization(self):
class S(torch.jit.ScriptModule):
def __init__(self, list_data):
super(S, self).__init__()
self.table = torch.jit.Attribute({"I": "am", "a test": "test"}, Dict[str, str])
self.list = torch.jit.Attribute(list_data, List[Tuple[int, int]])
@torch.jit.script_method
def forward(self):
return (self.table, self.list)
class M(torch.jit.ScriptModule):
def __init__(self):
super(M, self).__init__()
self.table = torch.jit.Attribute({"this": "is", "a different": "dict"}, Dict[str, str])
self.tensor = torch.jit.Attribute(torch.randn(2, 2), torch.Tensor)
self.s1 = S([(1, 2)])
self.s2 = S([(4, 5)])
@torch.jit.script_method
def forward(self):
return (self.table, self.tensor, self.s1.table, self.s2.list, self.s1.list)
m = M()
imported_m = self.getExportImportCopy(m)
self.assertEqual(m(), imported_m())
def test_serialization_big_ints(self):
class M(torch.jit.ScriptModule):
def __init__(self):
super(M, self).__init__()
self.int32_max = torch.jit.Attribute(2**31 - 1, int)
self.int32_min = torch.jit.Attribute(-2**31, int)
self.uint32_max = torch.jit.Attribute(2**32, int)
self.int64_max = torch.jit.Attribute(2**63 - 1, int)
self.int64_min = torch.jit.Attribute(-2**63, int)
self.tensor = torch.nn.Parameter(torch.ones(2, 2))
@torch.jit.script_method
def forward(self, x):
# type: (int) -> (int)
return x + (self.int32_max + self.int32_min) + (self.int64_max + self.int64_min)
m = M()
imported = self.getExportImportCopy(m)
self.assertEqual(m(10), imported(10))
self.assertEqual(m.int32_max, imported.int32_max)
self.assertEqual(m.int32_min, imported.int32_min)
self.assertEqual(m.uint32_max, imported.uint32_max)
self.assertEqual(m.int64_max, imported.int64_max)
self.assertEqual(m.int64_min, imported.int64_min)
def test_script_scope(self):
scripted = torch.jit.script(torch.nn.functional.triplet_margin_loss)
@unittest.skipIf(IS_WINDOWS, "NYI: TemporaryFileName on Windows")
def test_serialization_sharing(self):
class M(torch.jit.ScriptModule):
def __init__(self):
super(M, self).__init__()
self.list = torch.jit.Attribute([], List[str])
@torch.jit.script_method
def forward(self, key):
# type: (str) -> List[str]
self.list.append(key)
self.list.append(key)
self.list.append(key)
return self.list
# the text of the string should only appear once in the pickling
m = M()
s1 = "a long string"
s2 = "a different, even longer string"
self.assertEqual(m(s1), [s1] * 3)
self.assertEqual(m(s2), [s1] * 3 + [s2] * 3)
with TemporaryFileName() as fname:
m.save(fname)
archive_name = os.path.basename(os.path.normpath(fname))
archive = zipfile.ZipFile(fname, 'r')
pickled_data = archive.read(os.path.join(archive_name, 'data.pkl'))
out = io.StringIO()
pickletools.dis(pickled_data, out=out)
disassembled = out.getvalue()
FileCheck().check_count(s1, 1, exactly=True) \
.check_count("BINGET", 2, exactly=True) \
.check_count(s2, 1, exactly=True) \
.check_count("BINGET", 2, exactly=True).run(out.getvalue())
def test_sys_stdout_override(self):
@torch.jit.script
def foo():
print('foo')
class Redirect(object):
def __init__(self):
self.s = ''
def write(self, s):
self.s += s
old_stdout = sys.stdout
redirect = Redirect()
try:
sys.stdout = redirect
foo()
finally:
sys.stdout = old_stdout
FileCheck().check('foo').run(redirect.s)
def test_dtype_attr(self):
class Foo(torch.nn.Module):
def __init__(self):
super(Foo, self).__init__()
self.dtype = torch.zeros([]).dtype
def forward(self):
return torch.zeros(3, 4, dtype=self.dtype)
f = Foo()
torch.jit.script(f)
def test_named_buffers_are_iterable(self):
class MyMod(torch.nn.Module):
def __init__(self):
super(MyMod, self).__init__()
self.mod = (torch.nn.ReLU())
self.mod2 = (torch.nn.ReLU())
self.mod3 = torch.nn.Sequential(torch.nn.Sequential(torch.nn.ReLU()))
self.register_buffer('x', torch.zeros(3))
self.register_buffer('y', torch.zeros(3))
self.z = torch.zeros(3)
def bleh(self):
return self.z + 4
@torch.jit.export
def method(self):
names = [""]
vals = []
for name, buffer in self.named_buffers():
names.append(name)
vals.append(buffer + 2)
return names, vals
def forward(self, x):
return x
model = MyMod()
x = torch.jit.script(model)
z = self.getExportImportCopy(x)
self.assertEqual(z.method(), x.method())
self.assertEqual(z.method(), model.method())
self.assertEqual(x.method(), model.method())
names = x.method()
for name in names:
self.assertNotEqual('z', name)
def test_static_if_prop(self):
class MaybeHasAttr(torch.nn.Module):
def __init__(self, add_attr):
super(MaybeHasAttr, self).__init__()
if add_attr:
self.maybe_attr = 1
def forward(self):
if hasattr(self, "maybe_attr") and True:
return self.maybe_attr
else:
return 0
class MaybeHasAttr2(torch.nn.Module):
def __init__(self, add_attr):
super(MaybeHasAttr2, self).__init__()
if add_attr:
self.maybe_attr = 1
def forward(self):
if not hasattr(self, "maybe_attr") or False:
return 0
else:
return self.maybe_attr
torch.jit.script(MaybeHasAttr(True))
torch.jit.script(MaybeHasAttr(False))
torch.jit.script(MaybeHasAttr2(True))
torch.jit.script(MaybeHasAttr2(False))
class MyMod(torch.nn.Module):
def forward(self):
if hasattr(self, "foo"):
return 1
else:
return 0
@torch.jit.export
def fee(self):
return 1
self.checkModule(MyMod(), ())
class HasAttrMod(torch.nn.Module):
__constants__ = ["fee"]
def __init__(self):
super().__init__()
self.fee = 3
def forward(self):
a = hasattr(self, "fee")
b = hasattr(self, "foo")
c = hasattr(self, "hi")
d = hasattr(self, "nonexistant")
return (a, b, c, d)
def foo(self):
return 1
@torch.jit._overload_method
def hi(self, x: Tensor): ... # noqa: E704
def hi(self, x): # noqa: F811
return 2
self.checkModule(HasAttrMod(), ())
@torch.jit.script
class FooTest(object):
def __init__(self):
self.x = 1
def foo(self, y):
return self.x + y
def foo():
a = FooTest()
val1 = hasattr(a, "foo"), hasattr(a, "x"), hasattr(a, "bla")
val2 = hasattr(FooTest, "foo"), hasattr(FooTest, "a")
return val1, val2
self.assertEqual(foo(), torch.jit.script(foo)())
def _test_pickle_checkpoint(self, device):
with TemporaryFileName() as fname:
class M(torch.jit.ScriptModule):
__constants__ = ['fname']
def __init__(self, tensor):
super(M, self).__init__()
self.fname = fname
self.tensor = torch.nn.Parameter(tensor)
@torch.jit.script_method
def forward(self, x):
y = self.tensor + x
torch.save(y, self.fname)
return y
param = torch.randn(2, 2).to(device)
input = torch.randn(2, 2).to(device)
m = M(param)
m(input)
with open(fname, "rb") as handle:
loaded_tensor = torch.load(fname)
self.assertEqual(loaded_tensor, input + param)
def _test_pickle_checkpoint_views(self, device):
with TemporaryFileName() as fname:
class M(torch.jit.ScriptModule):
__constants__ = ['fname']
def __init__(self, tensor):
super(M, self).__init__()
self.fname = fname
self.tensor = torch.nn.Parameter(tensor)
@torch.jit.script_method
def forward(self, x):
y = self.tensor + x
y_view = y.view(4)
torch.save((y, y_view, y), self.fname)
return y
param = torch.randn(2, 2).to(device)
input = torch.randn(2, 2).to(device)
m = M(param)
m(input)
with open(fname, "rb") as handle:
loaded_y, loaded_y_view, loaded_y_2 = torch.load(fname)
self.assertEqual(loaded_y, input + param)
with torch.no_grad():
loaded_y_view[1] += 20
# assert that loaded_y changed as well
self.assertEqual(loaded_y.view(4), loaded_y_view)
self.assertEqual(loaded_y_2.view(4), loaded_y_view)
@unittest.skipIf(not RUN_CUDA, "no CUDA")
def test_pickle_checkpoint_cuda(self):
self._test_pickle_checkpoint('cuda')
self._test_pickle_checkpoint_views('cuda')
def test_pickle_checkpoint(self):
self._test_pickle_checkpoint('cpu')
self._test_pickle_checkpoint_views('cpu')
def test_pickle_checkpoint_tup(self):
@torch.jit.script
def foo(fname):
# type: (str) -> None
torch.save((3, 4), fname)
with TemporaryFileName() as name:
foo(name)
self.assertEqual(torch.load(name), (3, 4))
def test_string_list(self):
def fn(string):
# type: (str) -> List[str]
return list(string)
self.checkScript(fn, ("abcdefgh",))
def test_unicode_comments(self):
@torch.jit.script
def test(self, a):
# 🤷🤷🤷🤷
return torch.nn.functional.relu(a)
def test_get_set_state_with_tensors(self):
class M(torch.nn.Module):
def __init__(self):
super(M, self).__init__()
self.tensor = torch.randn(2, 2)
@torch.jit.export
def __getstate__(self):
return (self.tensor, self.training)
@torch.jit.export
def __setstate__(self, state):
self.tensor = state[0]
self.training = state[1]
def forward(self, x):
return x + self.tensor
with TemporaryFileName() as fname:
m = torch.jit.script(M())
m.save(fname)
loaded = torch.jit.load(fname)
self.assertEqual(loaded.tensor, m.tensor)
def test_in_for_and_comp_expr(self):
def fn(d):
# type: (Dict[str, int]) -> List[int]
out = [1]
for i in range(d["hi"] if "hi" in d else 6):
out.append(i)
return out
self.checkScript(fn, ({'hi': 2, 'bye': 3},))
self.checkScript(fn, ({'bye': 3},))
def test_for_else(self):
def fn():
c = 0
for i in range(4):
c += 10
else:
print("In else block of for...else")
with self.assertRaisesRegex(torch.jit.frontend.NotSupportedError, "else branches of for loops aren't supported"):
torch.jit.script(fn)
def test_split(self):
def split_two(tensor):
a, b, c = torch.split(tensor, 2, dim=1)
return a, b, c
x = torch.randn(3, 6)
y = torch.randn(3, 6)
self.checkScript(split_two, [(x + y)])
def test_conv_error(self):
@torch.jit.script
def fn(x, y):
return F.conv2d(x, y)
try:
fn(torch.ones(2, 2), torch.ones(4, 4))
except RuntimeError as e:
self.assertFalse('frame' in str(e))
def test_python_op_name(self):
import random
with self.assertRaisesRegex(RuntimeError, "randint"):
@torch.jit.script
def fn():
return random.randint()
def test_dir(self):
class M(torch.jit.ScriptModule):
def forward(self, t):
return t
self.assertTrue('forward' in dir(M()))
def test_kwarg_expansion_error(self):
@torch.jit.ignore
def something_else(h, i):
pass
def fn(x):
something_else(**x)
with self.assertRaisesRegex(torch.jit.frontend.NotSupportedError, "keyword-arg expansion is not supported"):
torch.jit.script(fn)
def test_kwargs_error_msg(self):
def other(**kwargs):
print(kwargs)
def fn():
return other()
with self.assertRaisesRegex(torch.jit.frontend.NotSupportedError, 'variable number'):
torch.jit.script(fn)
def another_other(*args):
print(args)
def another_fn():
return another_other()
with self.assertRaisesRegex(torch.jit.frontend.NotSupportedError, 'variable number'):
torch.jit.script(another_fn)
def test_inferred_error_msg(self):
"""
Test that when we get a type mismatch on a function where we inferred
the type to be tensor, a good error message is given.
"""
@torch.jit.script
def foo(a):
return a
with self.assertRaisesRegex(RuntimeError, (r"Expected a value of type \'Tensor \(inferred\)\'"
r"[\S\s]*Inferred \'a\' to be of type \'Tensor\'")):
foo("1")
def test_type_comments_in_body(self):
@torch.jit.script
def foo(a, # type: int
b, # type: int
):
# type: (...) -> int
# type: int
return a + b
class M(torch.nn.Module):
def __init__(self,
a, # type: int
b # type: int
):
# type: (...) -> None
super(M, self).__init__()
self.a = a # type: int
self.b = b # type: int
torch.jit.script(M(2, 3))
def test_input_keyword_in_schema(self):
def f(x):
return torch.ceil(input=x)
inp = torch.randn(10)
self.checkScript(f, (inp, ))
def test_module_method_reassignment(self):
class Foo(torch.nn.Module):
def __init__(self):
super().__init__()
def _forward(self, x):
return x
forward = _forward
sm = torch.jit.script(Foo())
input = torch.ones(2, 2)
self.assertEqual(input, sm(input))
# Tests the case where a torch.Tensor subclass (like Parameter) is used as
# input.
def test_script_module_tensor_subclass_argument(self):
@torch.jit.script
def parameter_script(x: torch.nn.Parameter):
return x
input = torch.ones(2, 2)
self.assertEqual(input, parameter_script(input))
def test_save_load_attr_error(self):
class Inner(nn.Module):
def __init__(self):
super().__init__()
def forward(self, x):
return x
class Wrapper(nn.Module):
def __init__(self, inner):
super().__init__()
self.inner = inner
def forward(self, x):
# this attribute doesn't exist on `Inner`
return self.inner.b(x)
inner_module = torch.jit.script(Inner())
inner_module = self.getExportImportCopy(inner_module)
wrapped = Wrapper(inner_module)
# This should properly complain that `self.inner` doesn't have the attribute `b`
with self.assertRaisesRegex(RuntimeError, 'has no attribute'):
torch.jit.script(wrapped)
def test_rescripting_loaded_modules(self):
class InnerSubmod(nn.Module):
__constants__ = ['my_constant']
def __init__(self):
super().__init__()
self.register_buffer("foo", torch.ones(1))
self.register_parameter("bar", torch.nn.Parameter(torch.ones(1)))
self.baz = torch.ones(1)
self.my_constant = 1
def forward(self, x):
return x + x
class Inner(nn.Module):
def __init__(self):
super().__init__()
self.submod = InnerSubmod()
def forward(self, x):
return self.submod(x)
class Wrapper(nn.Module):
def __init__(self, inner):
super().__init__()
self.inner = inner
def forward(self, x):
# access inner elements
ret = self.inner.submod(x) + self.inner.submod.foo + self.inner.submod.bar + self.inner.submod.baz
ret = ret + self.inner.submod.my_constant
return ret
inner_module = torch.jit.script(Inner())
wrapped = Wrapper(inner_module)
self.checkModule(wrapped, torch.ones(1))
inner_module_loaded = self.getExportImportCopy(inner_module)
wrapped_loaded = Wrapper(inner_module_loaded)
self.assertEqual(wrapped(torch.ones(1)), wrapped_loaded(torch.ones(1)))
def test_interpret_graph(self):
def fn(x):
return x.unfold(0, 1, 1)
graph_str = """
graph(%a : Tensor, %b : Tensor):
%c : Tensor = aten::mul(%a, %b)
return (%c)
"""
graph = parse_ir(graph_str)
a = torch.rand(10)
b = torch.rand(10)
test = torch._C._jit_interpret_graph(graph, (a, b))
ref = a * b
self.assertEqual(test, ref)
def test_signed_float_zero(self):
class MyModule(torch.nn.Module):
def __init__(self):
super(MyModule, self).__init__()
def forward(self, x):
return torch.div(x, -0.)
inp = torch.ones(1)
self.checkModule(MyModule(), inp)
def test_index_with_tuple(self):
class MyModule(torch.nn.Module):
def __init__(self):
super(MyModule, self).__init__()
def forward(self, x):
return x[(1,)]
self.checkModule(MyModule(), (torch.ones(2, 3),))
def test_context_manager(self):
class MyModule(torch.nn.Module):
def __init__(self):
super(MyModule, self).__init__()
def forward(self, x, y):
p = x + y
q = p + 2.0
return q
x = torch.randn(3, 2, dtype=torch.float)
y = torch.randn(3, 2, dtype=torch.float)
for fuser_name in ['fuser0', 'fuser1', 'none']:
with torch.jit.fuser(fuser_name):
self.checkModule(MyModule(), (x, y))
# known to be failing in tracer
EXCLUDE_TRACED = {
# The following fail due to #12024.
# A prim::ListConstruct is involved and the indices get traced as TensorType,
# which always require_grad. This causes a crash in autodiff.
'test___getitem___adv_index',
'test___getitem___adv_index_beg',
'test___getitem___adv_index_comb',
'test___getitem___adv_index_dup',
'test___getitem___adv_index_sub',
'test___getitem___adv_index_sub_2',
'test___getitem___adv_index_sub_3',
'test___getitem___adv_index_var',
# jit doesn't support sparse tensors.
'test_to_sparse',
'test_to_sparse_dim',
}
EXCLUDE_TYPE_CHECK = {
# slogdet tests use itemgetter to select its only differentiable output,
# but this happens outside of the graph we handle, so there are fewer
# reference outputs than graph outputs.
'test_slogdet_1x1_neg_det',
'test_slogdet_1x1_pos_det',
'test_slogdet_distinct_singular_values',
'test_slogdet_neg_det',
'test_slogdet_pos_det',
'test_slogdet_symmetric',
'test_slogdet_symmetric_pd',
'test_slogdet_batched_1x1_neg_det',
'test_slogdet_batched_pos_det',
'test_slogdet_batched_symmetric',
'test_slogdet_batched_symmetric_pd',
'test_slogdet_batched_distinct_singular_values'
}
# chunk returns a list in scripting and we don't unpack the list,
# Thus it won't be replaced by ConstantChunk and run AD.
# It's explicitly checked in test_chunk_constant_script_ad
# Similary for split, it's replaced by split_with_sizes in tracing,
# but we don't have AD formula for aten::split(Tensor, int[], int),
# an op registered in JIT so AD is not triggered in scripting.
EXCLUDE_SCRIPT_AD_CHECK = {
'test_chunk',
'test_chunk_dim',
'test_chunk_dim_neg0',
'test_split_size_list',
'test_split_size_list_dim',
'test_split_size_list_dim_neg0',
'test_tensor_indices_sections',
'test_tensor_indices_sections_dim',
'test_tensor_indices_sections_dim_neg0',
'test_tensor_split_sections',
'test_tensor_split_sections_dim',
'test_tensor_split_sections_dim_neg0'
}
EXCLUDE_PYTHON_PRINT = {
# no support for BroadcastingList in python printer
'test_nn_max_unpool1d',
'test_nn_max_unpool2d',
'test_nn_max_unpool3d',
'test_nn_max_pool1d',
'test_nn_max_pool2d',
'test_nn_max_pool3d',
'test_nn_max_pool1d_with_indices',
}
EXCLUDE_ALIAS = {
# aliases, which may appear in method_tests but are tested elsewhere
'true_divide',
# Disable tests for lu from common_methods_invocations.py
# TODO(@nikitaved) Enable jit tests once autograd.Function does support scripting
'lu'
}
class TestJitGeneratedModule(JitTestCase):
pass
class TestJitGeneratedFunctional(JitTestCase):
pass
# UBSAN per-function exclusions don't seem to work with OpenMP pragmas,
# and we have to disable the failing tests here instead.
UBSAN_DISABLED_TESTS = [
"test___rdiv___constant",
"test___rdiv___scalar_constant",
"test_addcdiv",
"test_addcdiv_broadcast_all",
"test_addcdiv_broadcast_rhs",
"test_addcdiv_scalar",
"test_addcdiv_scalar_broadcast_lhs",
"test_addcdiv_scalar_broadcast_rhs",
"test_addcdiv_scalar_scale",
"test_addcdiv_scalar_scale_broadcast_lhs",
"test_addcdiv_scalar_scale_broadcast_rhs",
"test_addcdiv_scale",
"test_addcdiv_scale_broadcast_all",
"test_addcdiv_scale_broadcast_rhs",
"test_add_broadcast_all",
"test_add_broadcast_lhs",
"test_add_broadcast_rhs",
"test_add_constant",
"test_add_scalar",
"test_add_scalar_broadcast_lhs",
"test_add_scalar_broadcast_rhs",
"test_div",
"test_div_broadcast_all",
"test_div_broadcast_lhs",
"test_div_broadcast_rhs",
"test_div_scalar",
"test_div_scalar_broadcast_lhs",
"test_div_scalar_broadcast_rhs",
"test_rsqrt",
"test_rsqrt_scalar",
"test_add",
"test_reciprocal",
"test_reciprocal_scalar",
]
L = 20
M = 10
S = 5
def add_nn_module_test(*args, **kwargs):
no_grad = False if 'no_grad' not in kwargs else kwargs['no_grad']
if 'desc' in kwargs and 'eval' in kwargs['desc']:
# eval() is not supported, so skip these tests
return
test_name = get_nn_mod_test_name(**kwargs)
@suppress_warnings
def do_test(self):
if test_name in EXCLUDE_SCRIPT_MODULES:
return
if not kwargs.get('check_jit', True):
raise unittest.SkipTest('module test skipped on JIT')
module_name = get_nn_module_name_from_kwargs(**kwargs)
if 'constructor' in kwargs:
nn_module = kwargs['constructor']
else:
nn_module = getattr(torch.nn, module_name)
if "FunctionalModule" in str(nn_module):
return
if 'constructor_args_fn' in kwargs:
constructor_args = kwargs['constructor_args_fn']()
else:
constructor_args = kwargs.get('constructor_args', ())
def create_script_module(*args, **kwargs):
"""Construct a script module that passes arguments through to self.submodule"""
formals, tensors, actuals = get_script_args(args)
method_args = ', '.join(['self'] + actuals)
call_args_str = ', '.join(actuals)
call = "self.submodule({})".format(call_args_str)
script = script_method_template.format(method_args, call)
submodule_constants = []
if kwargs.get('is_constant'):
submodule_constants = ['submodule']
# Create module to use the script method
class TheModule(torch.jit.ScriptModule):
__constants__ = submodule_constants
def __init__(self):
super(TheModule, self).__init__()
self.submodule = nn_module(*constructor_args)
def make_module(script):
module = TheModule()
# check __repr__
str(module)
module.define(script)
return module
module = make_module(script)
self.assertExportImportModule(module, tensors)
create_script_module.last_graph = module.graph
mod = module(*args)
return mod
# Construct a normal nn module to stay consistent with create_script_module
# and make use of a single global rng_state in module initialization
def create_nn_module(*args, **kwargs):
module = nn_module(*constructor_args)
return module(*args)
# Set up inputs from tuple of sizes or constructor fn
dtype = torch.double
if 'input_fn' in kwargs:
input = kwargs['input_fn']()
if isinstance(input, Tensor):
input = (input,)
if all(tensor.is_complex() for tensor in input):
dtype = torch.cdouble
else:
input = (kwargs['input_size'],)
if 'target_size' in kwargs:
input = input + (kwargs['target_size'],)
elif 'target_fn' in kwargs:
if torch.is_tensor(input):
input = (input,)
input = input + (kwargs['target_fn'](),)
elif 'target' in kwargs:
input = input + (kwargs['target'],)
# Extra parameters to forward()
if 'extra_args' in kwargs:
input = input + kwargs['extra_args']
args_variable, kwargs_variable = create_input(input, dtype=dtype)
f_args_variable = deepcopy(unpack_variables(args_variable))
# TODO(issue#52052) Neither this nor no_grad should be required
# if check_against_reference() is updated to check gradients
# w.r.t. weights and then only check w.r.t. inputs if any
# inputs require it.
any_requires_grad = any(input.requires_grad for input in f_args_variable)
# Check against Python module as reference
check_against_reference(self, create_script_module, create_nn_module,
lambda x: x, f_args_variable,
no_grad=no_grad or not any_requires_grad)
if 'slowTest' in kwargs:
do_test = slowTest(do_test)
post_add_test(test_name, (), do_test, TestJitGeneratedModule)
def post_add_test(test_name, skipTestIf, do_test, test_class):
assert not hasattr(test_class, test_name), 'Two tests have the same name: ' + test_name
for skip in skipTestIf:
do_test = skip(do_test)
if not (TEST_WITH_UBSAN and test_name in UBSAN_DISABLED_TESTS):
setattr(test_class, test_name, do_test)
def normalize_check_ad(check_ad, name):
# normalized check_ad is 3-element tuple: (bool, List[str], List[str])
if len(check_ad) == 0:
check_ad = [False, ['aten::' + name], []]
elif len(check_ad) == 1:
check_ad = [check_ad[0], ['aten::' + name], []]
elif len(check_ad) == 2:
check_ad = [check_ad[0], check_ad[1], []]
elif len(check_ad) == 3:
check_ad = list(check_ad)
else:
raise Exception('Invalid check_ad, requires (bool, str|List[str], str|List[str])')
check_ad = [[t] if isinstance(t, str) else t for t in check_ad]
return check_ad
class TestProducerVersion(TestCase):
def test_version(self):
# issue gh-32561
self.assertTrue(torch.__version__.startswith(torch.onnx.producer_version))
for test in module_tests + new_module_tests + additional_module_tests:
add_nn_module_test(**test)
for test in criterion_tests:
test['no_grad'] = True
add_nn_module_test(**test)
if __name__ == '__main__':
run_tests()
import jit.test_module_interface
suite = unittest.findTestCases(jit.test_module_interface)
unittest.TextTestRunner().run(suite)