Update docs in control_flow_ops.py for consistent Markdown rendering
diff --git a/tensorflow/python/ops/parallel_for/control_flow_ops.py b/tensorflow/python/ops/parallel_for/control_flow_ops.py
index a764977..d145a7b 100644
--- a/tensorflow/python/ops/parallel_for/control_flow_ops.py
+++ b/tensorflow/python/ops/parallel_for/control_flow_ops.py
@@ -51,8 +51,8 @@
loop_fn: A function that takes an int32 scalar tf.Tensor object representing
the iteration number, and returns a possibly nested structure of tensor
objects. The shape of these outputs should not depend on the input.
- loop_fn_dtypes: dtypes for the outputs of loop_fn.
- iters: Number of iterations for which to run loop_fn.
+ loop_fn_dtypes: dtypes for the outputs of `loop_fn`.
+ iters: Number of iterations for which to run `loop_fn`.
parallel_iterations: The number of iterations that can be dispatched in
parallel. This knob can be used to control the total memory usage.
@@ -137,7 +137,7 @@
`pfor` has functionality similar to `for_loop`, i.e. running `loop_fn` `iters`
times, with input from 0 to `iters - 1`, and stacking corresponding output of
- each iteration. However the implementation does not use a tf.while_loop.
+ each iteration. However the implementation does not use a `tf.while_loop`.
Instead it adds new operations to the graph that collectively compute the same
value as what running `loop_fn` in a loop would compute.
@@ -152,7 +152,7 @@
reads, etc).
- Conversion works only on a limited set of kernels for which a converter
has been registered.
- - loop_fn has limited support for control flow operations. tf.cond in
+ - `loop_fn` has limited support for control flow operations. `tf.cond` in
particular is not supported.
- `loop_fn` should return nested structure of Tensors or Operations. However
if an Operation is returned, it should have zero outputs.
@@ -166,9 +166,9 @@
or Operation objects. Note that if setting `parallel_iterations` argument
to something other than None, `loop_fn` may be called more than once
during graph construction. So it may need to avoid mutating global state.
- iters: Number of iterations for which to run loop_fn.
+ iters: Number of iterations for which to run `loop_fn`.
fallback_to_while_loop: If true, on failing to vectorize an operation, pfor
- fallbacks to using a tf.while_loop to dispatch the iterations.
+ fallbacks to using a `tf.while_loop` to dispatch the iterations.
parallel_iterations: A knob to control how many iterations are vectorized
and dispatched in parallel. The default value of None corresponds to
vectorizing all the iterations. If `parallel_iterations` is smaller than
@@ -337,7 +337,7 @@
"""Parallel map on the list of tensors unpacked from `elems` on dimension 0.
- This method works similar to tf.map_fn but is optimized to run much faster,
+ This method works similar to `tf.map_fn` but is optimized to run much faster,
possibly with a much larger memory footprint. The speedups are obtained by
vectorization (see https://arxiv.org/pdf/1903.04243.pdf). The idea behind
vectorization is to semantically launch all the invocations of `fn` in