Issue #998 Fix

[DevyRuxpin]

Possible solution to issue #998

[pytorch-bot]

🔗 Helpful Links

🧪 See artifacts and rendered test results at hud.pytorch.org/pr/pytorch/ao/1074

• 📄 Preview Python docs built from this PR

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

❌ 4 New Failures, 1 Unrelated Failure

As of commit 19b014c with merge base e7b33bc ():

NEW FAILURES - The following jobs have failed:

• Run Float8 Tests / test (SM-89, linux.g6.4xlarge.experimental.nvidia.gpu, --pre torch --index-url https://download.p... / linux-job (gh)
  RuntimeError: Command docker exec -t 683bbd9a80b2d4043cb392c90d30d4b77842d73501a3f76fbbb33e09379df6e7 /exec failed with exit code 1
• Run Regression Tests / test (CUDA 2.2.2, linux.g5.12xlarge.nvidia.gpu, torch==2.2.2 "numpy<2" , cuda, 12.1) / linux-job (gh)
  RuntimeError: Command docker exec -t 7dba2628ec1ea50279b617477fea4a36a2ca23ef1ca6c6b4cdb34e2315190f3d /exec failed with exit code 1
• Run Regression Tests / test (CUDA 2.3, linux.g5.12xlarge.nvidia.gpu, torch==2.3.0, cuda, 12.1) / linux-job (gh)
  RuntimeError: Command docker exec -t 6196b1f0382e93d5304d6b111913fb973cbd5a190f80a69489009db836a585ac /exec failed with exit code 1
• Run Regression Tests / test (CUDA 2.4, linux.g5.12xlarge.nvidia.gpu, torch==2.4.0, cuda, 12.1) / linux-job (gh)
  RuntimeError: Command docker exec -t 52018d1451d0d3b44ff25a2f6541b3443d9655afed5c454fff3d8bcb676ba12b /exec failed with exit code 1

BROKEN TRUNK - The following job failed but were present on the merge base:

👉 Rebase onto the `viable/strict` branch to avoid these failures

• Run Regression Tests / test (CUDA Nightly, linux.g5.12xlarge.nvidia.gpu, --pre torch --index-url https://download.pytorc... / linux-job (gh) (trunk failure)

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

[facebook-github-bot]

Hi @DevyRuxpin!

Thank you for your pull request and welcome to our community.

Action Required

In order to merge any pull request (code, docs, etc.), we require contributors to sign our Contributor License Agreement, and we don't seem to have one on file for you.

Process

In order for us to review and merge your suggested changes, please sign at https://code.facebook.com/cla. If you are contributing on behalf of someone else (eg your employer), the individual CLA may not be sufficient and your employer may need to sign the corporate CLA.

Once the CLA is signed, our tooling will perform checks and validations. Afterwards, the pull request will be tagged with CLA signed. The tagging process may take up to 1 hour after signing. Please give it that time before contacting us about it.

If you have received this in error or have any questions, please contact us at [email protected]. Thanks!

[facebook-github-bot]

Thank you for signing our Contributor License Agreement. We can now accept your code for this (and any) Meta Open Source project. Thanks!

[msaroufim]

Might need to also add some tests so CI signal is helpful and just to be clear are we only supporting bf16 moving forward or both fp16 and bf16?

[gau-nernst]

I think we shouldn't remove FP16 support, since it is the original dtype that Quant-LLM support. Not sure how we can structure it nicely (i.e. support both FP16 and BF16), but you can come up with a way first. Ideally it shouldn't have intrusive changes to the original code, so if there are upstream updates, it's easier for us to pull the new updates in. Maybe some kind of templates or just have a separate folder for BF16?

[DevyRuxpin]

I will work on this further. Thank you for feedback, i will draw things back some on the changes and incorporate your suggestions. i will work on this tonight/tomorrow.

[tobiasvanderwerff]

@gau-nernst @msaroufim I have already been working on a templated solution that supports both FP16 and BF16, and it is almost finished (see the diff here). I think this would be a better solution than removing FP16 all together.

I'm just smoothing out some final bugs and it should be as good as ready. I didn't make a PR yet because I wanted to make sure that everything worked properly first. In hindsight, it might have been better to already create a PR to avoid the duplicate work done here (although I did say I was working on this feature in the original issue).