-
Notifications
You must be signed in to change notification settings - Fork 364
fix: Add support for fake tensors #1955
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Merged
gs-olive
merged 2 commits into
pytorch:main
from
gs-olive:remove_fake_tensor_unsupported
Jun 28, 2023
Merged
Changes from all commits
Commits
Show all changes
2 commits
Select commit
Hold shift + click to select a range
File filter
Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
There are no files selected for viewing
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
114 changes: 114 additions & 0 deletions
114
py/torch_tensorrt/dynamo/backend/test/test_specialized_models.py
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
| Original file line number | Diff line number | Diff line change |
|---|---|---|
| @@ -0,0 +1,114 @@ | ||
| from utils import lower_graph_testing | ||
| from torch.testing._internal.common_utils import run_tests, TestCase | ||
| import torch | ||
| from torch_tensorrt.dynamo import compile | ||
|
|
||
|
|
||
| class TestFakeTensors(TestCase): | ||
| def test_lowering_mul_int(self): | ||
| class MulInt(torch.nn.Module): | ||
| def forward(self, x): | ||
| return x * 7 | ||
|
|
||
| # Operations expected to be included in the traced graph after decompositions | ||
| expected_ops = { | ||
| torch.ops.aten.mul.Tensor, | ||
| } | ||
|
|
||
| inputs = [ | ||
| torch.rand( | ||
| 3, | ||
| 5, | ||
| 7, | ||
| ).cuda(), | ||
| ] | ||
|
|
||
| fx_graph = torch.fx.symbolic_trace(MulInt()) | ||
| _, expected_ops_unseen = lower_graph_testing( | ||
| fx_graph, | ||
| inputs, | ||
| expected_ops=expected_ops, | ||
| min_block_size=1, | ||
| ) | ||
|
|
||
| self.assertEquals( | ||
| len(expected_ops_unseen), | ||
| 0, | ||
| f"The following expected ops were not encountered: {expected_ops_unseen}", | ||
| ) | ||
|
|
||
| torch._dynamo.reset() | ||
|
|
||
| # Validate that the results between Torch and Torch-TRT are similar | ||
| optimized_model = compile( | ||
| fx_graph, inputs, min_block_size=1, pass_through_build_failures=True | ||
| ) | ||
| optimized_model_results = optimized_model(*inputs).detach().cpu() | ||
| torch_model_results = fx_graph(*inputs).detach().cpu() | ||
|
|
||
| max_diff = float( | ||
| torch.max(torch.abs(optimized_model_results - torch_model_results)) | ||
| ) | ||
| self.assertAlmostEqual( | ||
| max_diff, | ||
| 0, | ||
| msg=f"MulInt TRT outputs don't match with the original model.", | ||
| ) | ||
| torch._dynamo.reset() | ||
|
|
||
| def test_lowering_add_float(self): | ||
| class AddFloat(torch.nn.Module): | ||
| def forward(self, x): | ||
| return x + 84.0 | ||
|
|
||
| # Operations expected to be included in the traced graph after decompositions | ||
| expected_ops = { | ||
| torch.ops.aten.add.Tensor, | ||
| } | ||
|
|
||
| inputs = [ | ||
| torch.rand( | ||
| 1, | ||
| 5, | ||
| 7, | ||
| 9, | ||
| ).cuda(), | ||
| ] | ||
|
|
||
| fx_graph = torch.fx.symbolic_trace(AddFloat()) | ||
| _, expected_ops_unseen = lower_graph_testing( | ||
| fx_graph, | ||
| inputs, | ||
| expected_ops=expected_ops, | ||
| min_block_size=1, | ||
| ) | ||
|
|
||
| self.assertEquals( | ||
| len(expected_ops_unseen), | ||
| 0, | ||
| f"The following expected ops were not encountered: {expected_ops_unseen}", | ||
| ) | ||
|
|
||
| torch._dynamo.reset() | ||
|
|
||
| # Validate that the results between Torch and Torch-TRT are similar | ||
| optimized_model = compile( | ||
| fx_graph, inputs, min_block_size=1, pass_through_build_failures=True | ||
| ) | ||
| optimized_model_results = optimized_model(*inputs).detach().cpu() | ||
| torch_model_results = fx_graph(*inputs).detach().cpu() | ||
|
|
||
| max_diff = float( | ||
| torch.max(torch.abs(optimized_model_results - torch_model_results)) | ||
| ) | ||
| self.assertAlmostEqual( | ||
| max_diff, | ||
| 0, | ||
| msg=f"AddFloat TRT outputs don't match with the original model.", | ||
| ) | ||
|
|
||
| torch._dynamo.reset() | ||
|
|
||
|
|
||
| if __name__ == "__main__": | ||
| run_tests() | ||
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Oops, something went wrong.
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Are we enable the faketensor in this test? I do not quite understand the purpose of this test file?
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Based on the changes in
backends.py, which remove all calls to@fake_tensor_unsupported, fake tensors will be enabled by default via Dynamo/AOT. The purpose of this test is to verify that utilities likecreate_constantdo not instantiate Torch tensors when provided scalar inputs. For example, in the testtest_lowering_mul_intbelow, the only op in the graph will be something like:Without the changes in this PR, the above will fail at runtime because
create_constantwill make atorch.Tensorfor the scalar 7, and this tensor will be fake (hold no values), so when TRT goes to extract the value to make a constant tensor, the script fails.