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

Fix WOQ int8 failures #884

Open
wants to merge 5 commits into
base: main
Choose a base branch
from
Open

Fix WOQ int8 failures #884

wants to merge 5 commits into from
+3 −11

Conversation

leslie-fang-intel
Copy link
Collaborator

Fix WOQ int8 failures as discussed in #843 (comment)

@pytorch-bot PyTorch Bot
Copy link

pytorch-bot bot commented Sep 13, 2024 •
edited
Loading

🔗 Helpful Links

🧪 See artifacts and rendered test results at hud.pytorch.org/pr/pytorch/ao/884

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

❌ 1 New Failure

As of commit 26e9c44 with merge base 8236a87 (image):

NEW FAILURE - The following job has failed:

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

@facebook-github-bot facebook-github-bot added the CLA Signed This label is managed by the Facebook bot. Authors need to sign the CLA before a PR can be reviewed. label Sep 13, 2024
@leslie-fang-intel
Copy link
Collaborator Author

cc @jerryzh168 @bdhirsh

@leslie-fang-intel leslie-fang-intel mentioned this pull request Sep 13, 2024
Merged
jerryzh168
jerryzh168 reviewed Sep 13, 2024
View reviewed changes
test/integration/test_integration.py Outdated
if not TORCH_VERSION_AT_LEAST_2_5:
if (
not TORCH_VERSION_AT_LEAST_2_5
) or torch._inductor.config.freezing:
Copy link
Contributor

Choose a reason for hiding this comment

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

when is freezing set? how do user know that they need to call unwrap_tensor_subclass in that case?

Copy link
Collaborator Author

Choose a reason for hiding this comment

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

Hi Jerry, freezing is setting in

ao/test/integration/test_integration.py

Line 829 in 8236a87

@torch._inductor.config.patch({"freezing": True})
when running the UT. User can also set it as torch._inductor.config.freezing=True in script or using the environment variable as TORCHINDUCTOR_FREEZING=1

Copy link
Collaborator Author

@leslie-fang-intel leslie-fang-intel Sep 13, 2024 •
edited
Loading

Choose a reason for hiding this comment

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

how do user know that they need to call unwrap_tensor_subclass in that case?

That's a good question. Did you mean we should write document somewhere? And the root-cause for why freezing needs unwrap_tensor_subclass is in pytorch/pytorch#123522.

Copy link
Contributor

Choose a reason for hiding this comment

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

I see, thanks for the context! yeah I think it would be good to add this in the README: https://github.com/pytorch/ao/tree/main/torchao/quantization#workaround-with-unwrap_tensor_subclass-for-export-aoti-and-torchcompile-pytorch-24-and-before-only and link the issue as well

Copy link
Collaborator Author

Choose a reason for hiding this comment

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

Thanks and adding the readme with issue link in this PR. please help to take a look again.

This was referenced Sep 13, 2024
Closed
Closed
jerryzh168
jerryzh168 approved these changes Sep 14, 2024
View reviewed changes
Copy link
Contributor

@jerryzh168 jerryzh168 left a comment

Choose a reason for hiding this comment

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

LGTM

@jerryzh168
Copy link
Contributor

please fix CI before landing

@leslie-fang-intel
Copy link
Collaborator Author

leslie-fang-intel commented Sep 16, 2024 •
edited
Loading

please fix CI before landing

Thanks @jerryzh168, we should land the PR in PyTorch at first.
Could you help to take a review pytorch/pytorch#135928?

@leslie-fang-intel leslie-fang-intel force-pushed the leslie/fix_woq_int8_failure branch 2 times, most recently from 5fa2288 to 828571c Compare September 20, 2024 00:33
@leslie-fang-intel
Copy link
Collaborator Author

Hi @jerryzh168, since pytorch/pytorch#135928 has been merged into PyTorch, could you approve the CI run for this PR to check if it passes all tests?

@leslie-fang-intel
Copy link
Collaborator Author

It seems test_int8_weight_only_quant_subclass_api with float16 inputs still failed which is probably due to numerical issue. Will take a further look.

@leslie-fang-intel leslie-fang-intel mentioned this pull request Sep 20, 2024
Closed
@leslie-fang-intel
Copy link
Collaborator Author

leslie-fang-intel commented Sep 20, 2024 •
edited
Loading

Creating another PR: pytorch/pytorch#136353 to fix the float16 input failure.

@yanbing-j yanbing-j mentioned this pull request Sep 23, 2024
Closed
pytorchmergebot pushed a commit to pytorch/pytorch that referenced this pull request Sep 25, 2024
@leslie-fang-intel @pytorchmergebot
[Inductor][CPP] Fix int8 cvt half (#136353)
3c7edf1 
Fix the correctness issue of pytorch/ao#884. The current implementation for converting between `Half/BFloat16` and `int8/uint8` incorrectly assumes that 1/4 of the int8/uint8 vector lane maps to 1/2 of the Half/BFloat16 vector lane. This assumption leads to accuracy issues after the full bit-width vectorization of the Half data type was introduced. When converting between int8 weights and the half data type, the generated code is as the following:
```
#include "/tmp/torchinductor_leslie/xw/cxww3s7wxrujoyxna7mlcjktid2uu6nntixqwm542xfkd756gl3x.h"
extern "C"  void kernel(const int8_t* in_ptr0,
                       half* out_ptr0)
{
    {
        for(int64_t x0=static_cast<int64_t>(0L); x0<static_cast<int64_t>(2048L); x0+=static_cast<int64_t>(32L))
        {
            auto tmp0 = at::vec::Vectorized<int8_t>::loadu(in_ptr0 + static_cast<int64_t>(x0), static_cast<int64_t>(32));
            auto tmp1 = at::vec::convert<half>(tmp0);
            tmp1.store(out_ptr0 + static_cast<int64_t>(x0), static_cast<int64_t>(32));
        }
    }
}
```

In this PR, we address the issue by changing the implementation to convert 1/2 of the int8/uint8 vector lane into a full vector lane of Half/BFloat16.

**TestPlan**
* AO: `python test/integration/test_integration.py -k test_int8_weight_only_quant_subclass_api`
* `python -u -m pytest -s -v test/inductor/test_cpu_repro.py -k test_convert_int8_to_half_vec`
* Due to the CPP backend legalization pass, we are unable to create a unit test to simulate the conversion from `Half` to `int8`. Instead, we rely on a C++ test case.
  * `./build/bin/vec_test_all_types_AVX512 --gtest_filter="VecConvertTestsReducedFloat/*.ConvertReduced"`
  * `./build/bin/vec_test_all_types_AVX2 --gtest_filter="VecConvertTestsReducedFloat/*.ConvertReduced"`

Pull Request resolved: #136353
Approved by: https://github.com/jgong5, https://github.com/jerryzh168
BoyuanFeng pushed a commit to BoyuanFeng/pytorch that referenced this pull request Sep 25, 2024
@leslie-fang-intel @BoyuanFeng
[Inductor][CPP] Fix int8 cvt half (pytorch#136353)
d6abebe 
Fix the correctness issue of pytorch/ao#884. The current implementation for converting between `Half/BFloat16` and `int8/uint8` incorrectly assumes that 1/4 of the int8/uint8 vector lane maps to 1/2 of the Half/BFloat16 vector lane. This assumption leads to accuracy issues after the full bit-width vectorization of the Half data type was introduced. When converting between int8 weights and the half data type, the generated code is as the following:
```
#include "/tmp/torchinductor_leslie/xw/cxww3s7wxrujoyxna7mlcjktid2uu6nntixqwm542xfkd756gl3x.h"
extern "C"  void kernel(const int8_t* in_ptr0,
                       half* out_ptr0)
{
    {
        for(int64_t x0=static_cast<int64_t>(0L); x0<static_cast<int64_t>(2048L); x0+=static_cast<int64_t>(32L))
        {
            auto tmp0 = at::vec::Vectorized<int8_t>::loadu(in_ptr0 + static_cast<int64_t>(x0), static_cast<int64_t>(32));
            auto tmp1 = at::vec::convert<half>(tmp0);
            tmp1.store(out_ptr0 + static_cast<int64_t>(x0), static_cast<int64_t>(32));
        }
    }
}
```

In this PR, we address the issue by changing the implementation to convert 1/2 of the int8/uint8 vector lane into a full vector lane of Half/BFloat16.

**TestPlan**
* AO: `python test/integration/test_integration.py -k test_int8_weight_only_quant_subclass_api`
* `python -u -m pytest -s -v test/inductor/test_cpu_repro.py -k test_convert_int8_to_half_vec`
* Due to the CPP backend legalization pass, we are unable to create a unit test to simulate the conversion from `Half` to `int8`. Instead, we rely on a C++ test case.
  * `./build/bin/vec_test_all_types_AVX512 --gtest_filter="VecConvertTestsReducedFloat/*.ConvertReduced"`
  * `./build/bin/vec_test_all_types_AVX2 --gtest_filter="VecConvertTestsReducedFloat/*.ConvertReduced"`

Pull Request resolved: pytorch#136353
Approved by: https://github.com/jgong5, https://github.com/jerryzh168
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Reviewers

@jerryzh168 jerryzh168 jerryzh168 approved these changes

Assignees
No one assigned
Labels
CLA Signed This label is managed by the Facebook bot. Authors need to sign the CLA before a PR can be reviewed.
Projects
None yet
Milestone
No milestone
Development

Successfully merging this pull request may close these issues.
3 participants
@leslie-fang-intel @jerryzh168 @facebook-github-bot