Skip to content
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.

Sign up for GitHub

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

Jump to bottom

[ONNX] Support quantized::linear_relu #109755

Closed
wants to merge 4 commits into from
Closed

[ONNX] Support quantized::linear_relu #109755

wants to merge 4 commits into from

Conversation

gustavla
Copy link
Contributor

@gustavla gustavla commented Sep 21, 2023 •
edited

  • Adds support for quantized::linear_relu
    • Adds weight unpacking pattern matcher
    • Adds to export for opset 10 and 13.
  • Adds QAT test modeled after conv2d+relu fusion test

@gustavla
torch->onnx export support: quantized::linear_relu
0ca3414 
- Adds support for quantized::linear_relu
  - Adds weight packing pattern matcher
  - Adds to export for opset 10 and 13.
- Adds QAT test modeled after conv2d+relu fusion
@pytorch-bot
Copy link

pytorch-bot bot commented Sep 21, 2023 •
edited

🔗 Helpful Links

🧪 See artifacts and rendered test results at hud.pytorch.org/pr/109755

Note: Links to docs will display an error until the docs builds have been completed.

✅ No Failures

As of commit 641fe7c with merge base 4e3b032 (image):
💚 Looks good so far! There are no failures yet. 💚

This comment was automatically generated by Dr. CI and updates every 15 minutes.

@pytorch-bot pytorch-bot bot added the release notes: onnx torch.onnx related changes that should show up in the release notes label Sep 21, 2023
@gustavla
Make bias float32.
e0a6b63
@gustavla
Fix bias again, and correct conv tests.
990711c 
This has me confused, because the bias and shapes didn't match the
in_feature/out_feature of the Conv2d that I was basing the Linear code
on. This is because it was defining the ops incorrectly and then
re-writing the weights swapping the inputs/outputs channels. It never
complained that the original definition had a mismatch.

I have fixed this for Linear now, but also fixed it for three cases of
Conv2d.
@gustavla
Copy link
Contributor Author

Note, I also spotted some typos in the Conv2d definitions and have also pushed a fix for them. This didn't cause any issues, because the weights were replaced anyway, thus changing the definition of the op. However, it is confusing and should be corrected.

Note that the in/out channels are swapped between how a Conv2d is constructed and the actual weights, but currently in the code they use the same order:

In [21]: torch.nn.Conv2d(2, 4, 3).weight.shape
Out[21]: torch.Size([4, 2, 3, 3])

gustavla
gustavla commented Sep 21, 2023
View reviewed changes
test/onnx/test_pytorch_onnx_onnxruntime.py
@@ -13091,7 +13091,7 @@ class M(torch.nn.Module):
def __init__(self):
super().__init__()
self.quant = torch.ao.quantization.QuantStub()
self.conv = torch.nn.Conv2d(2, 4, 3, stride=2)
self.conv = torch.nn.Conv2d(4, 2, 3, stride=2)
Copy link
Contributor Author

@gustavla gustavla Sep 21, 2023 •
edited

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

See my comment about these typos. If you expand the context below, the weights are replaced as follows:

        model.conv.weight = torch.nn.Parameter(
            _construct_tensor_for_quantization_test((2, 4, 3, 3), max_val=2)
        )

But this is not consistent with Conv2d(2, 4, 3), it is consistent with Conv2d(4, 2, 3), which is what I'm changing it to.

@ezyang ezyang added the triaged This issue has been looked at a team member, and triaged and prioritized into an appropriate module label Sep 21, 2023
BowenBao
BowenBao approved these changes Sep 21, 2023
View reviewed changes
Copy link
Collaborator

@BowenBao BowenBao left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

LGTM, thanks!

@BowenBao BowenBao added the topic: improvements topic category label Sep 21, 2023
@BowenBao
Copy link
Collaborator

@pytorchbot merge

@pytorch-bot pytorch-bot bot added the ciflow/trunk Trigger trunk jobs on your pull request label Sep 21, 2023
@pytorchmergebot
Copy link
Collaborator

Merge started

Your change will be merged once all checks pass (ETA 0-4 Hours).

Learn more about merging in the wiki.

Questions? Feedback? Please reach out to the PyTorch DevX Team

Advanced Debugging
Check the merge workflow status
here

@thiagocrepaldi thiagocrepaldi added module: onnx Related to torch.onnx onnx-triaged triaged by ONNX team labels Sep 21, 2023
@pytorchmergebot
Copy link
Collaborator

Merge failed

Reason: 1 jobs have failed, first few of them are: trunk / macos-12-py3-arm64 / test (default, 1, 3, macos-m1-12)

Details for Dev Infra team Raised by workflow job

@justinchuby
Copy link
Collaborator

@pytorchbot merge -i

@pytorchmergebot
Copy link
Collaborator

Merge started

Your change will be merged while ignoring the following 0 checks:

Learn more about merging in the wiki.

Questions? Feedback? Please reach out to the PyTorch DevX Team

Advanced Debugging
Check the merge workflow status
here

@justinchuby justinchuby changed the title torch->onnx export support: quantized::linear_relu [ONNX] Support quantized::linear_relu Sep 21, 2023
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Reviewers

@thiagocrepaldi thiagocrepaldi thiagocrepaldi approved these changes

@BowenBao BowenBao BowenBao approved these changes

@abock abock Awaiting requested review from abock

@wschin wschin Awaiting requested review from wschin wschin is a code owner

Assignees
No one assigned
Labels
ciflow/trunk Trigger trunk jobs on your pull request Merged module: onnx Related to torch.onnx onnx-triaged triaged by ONNX team open source release notes: onnx torch.onnx related changes that should show up in the release notes topic: improvements topic category triaged This issue has been looked at a team member, and triaged and prioritized into an appropriate module
Projects
ONNX
Status: Done
Milestone
No milestone
Development

Successfully merging this pull request may close these issues.

None yet
7 participants
@gustavla @BowenBao @pytorchmergebot @justinchuby @thiagocrepaldi @ezyang @pytorchbot