Add detailed error message for iOS test (#79140)

As title.
Tested locally.

Pull Request resolved: https://github.com/pytorch/pytorch/pull/79140
Approved by: https://github.com/kit1980
diff --git a/ios/TestApp/TestAppTests/TestLiteInterpreter.mm b/ios/TestApp/TestAppTests/TestLiteInterpreter.mm
index 37c8692..1afe4ec 100644
--- a/ios/TestApp/TestAppTests/TestLiteInterpreter.mm
+++ b/ios/TestApp/TestAppTests/TestLiteInterpreter.mm
@@ -21,17 +21,21 @@
   XCTAssertTrue(outputTensor.numel() == 1000);
 }
 
-- (void)testModel:(NSString*)filename {
-  // model generated using the current pytorch revision
-  [self runModel:[NSString stringWithFormat:@"%@_temp", filename]];
-  // model generated using older pyotrch revision
-  [self runModel:filename];
+- (void)testModel:(NSString*)modelName {
+  NSString* modelPath = [[NSBundle bundleForClass:[self class]] pathForResource:modelName
+                                                                         ofType:@"ptl"];
+  XCTAssertNotNil(modelPath, @"Model not found. See https://github.com/pytorch/pytorch/tree/master/test/mobile/model_test#diagnose-failed-test.");
+  [self runModel:modelPath];
+
+  // model generated on the fly
+  NSString* onTheFlyModelName = [NSString stringWithFormat:@"%@", modelName];
+  NSString* onTheFlyModelPath = [[NSBundle bundleForClass:[self class]] pathForResource:onTheFlyModelName
+                                                                         ofType:@"ptl"];
+  XCTAssertNotNil(onTheFlyModelPath, @"On-the-fly model not found. Follow https://github.com/pytorch/pytorch/tree/master/test/mobile/model_test#diagnose-failed-test to generate them and run the setup.rb script again.");
+  [self runModel:onTheFlyModelPath];
 }
 
-- (void)runModel:(NSString*)filename {
-  NSString* modelPath = [[NSBundle bundleForClass:[self class]] pathForResource:filename
-                                                                         ofType:@"ptl"];
-  XCTAssertNotNil(modelPath);
+- (void)runModel:(NSString*)modelPath {
   c10::InferenceMode mode;
   auto module = torch::jit::_load_for_mobile(modelPath.UTF8String);
   auto has_bundled_input = module.find_method("get_all_bundled_inputs");