Closed
Conversation
[ghstack-poisoned]
[ghstack-poisoned]
[ghstack-poisoned]
[ghstack-poisoned]
Follow up to #165098 - adding bf16 support for the backward pass. To avoid BC breaking changes/losing precision, we upcast the parameters to fp32 after the op gets called, and downcast the gradients to bf16 before returning. For testing, we upcast to fp32 before calling the reference function. [ghstack-poisoned]
Follow up to #165098 - adding bf16 support for the backward pass. To avoid BC breaking changes/losing precision, we upcast the parameters to fp32 after the op gets called, and downcast the gradients to bf16 before returning. For testing, we upcast to fp32 before calling the reference function. [ghstack-poisoned]
Follow up to #165098 - adding bf16 support for the backward pass. To avoid BC breaking changes/losing precision, we upcast the parameters to fp32 after the op gets called, and downcast the gradients to bf16 before returning. For testing, we upcast to fp32 before calling the reference function. [ghstack-poisoned]
Follow up to #165098 - adding bf16 support for the backward pass. To avoid BC breaking changes/losing precision, we upcast the parameters to fp32 after the op gets called, and downcast the gradients to bf16 before returning. For testing, we upcast to fp32 before calling the reference function. [ghstack-poisoned]
Follow up to #165098 - adding bf16 support for the backward pass. To avoid BC breaking changes/losing precision, we upcast the parameters to fp32 after the op gets called, and downcast the gradients to bf16 before returning. For testing, we upcast to fp32 before calling the reference function. [ghstack-poisoned]
[ghstack-poisoned]
Follow up to #165098 - adding bf16 support for the backward pass. To avoid BC breaking changes/losing precision, we upcast the parameters to fp32 after the op gets called, and downcast the gradients to bf16 before returning. For testing, we upcast to fp32 before calling the reference function. We increase the tolerance to 1e-2 for bf16 inputs because of a difference in casting calculations between python's `x.to(torch.bfloat16)` and cpp's `x.to(at::kBFloat16)` (after comparing intermediate tensors, we found that the numerics diverge after the final casting). We don't explicitly cast in the CPP op but rather let autograd/optimizer handle it. [ghstack-poisoned]
🔗 Helpful Links🧪 See artifacts and rendered test results at hud.pytorch.org/pr/165362
Note: Links to docs will display an error until the docs builds have been completed. ✅ You can merge normally! (1 Unrelated Failure)As of commit 6d643da with merge base fbe0d20 ( FLAKY - The following job failed but was likely due to flakiness present on trunk:
This comment was automatically generated by Dr. CI and updates every 15 minutes. |
[ghstack-poisoned]
[ghstack-poisoned]
[ghstack-poisoned]
[ghstack-poisoned]
[ghstack-poisoned]
[ghstack-poisoned]
[ghstack-poisoned]
andrewor14
approved these changes
Oct 14, 2025
[ghstack-poisoned]
[ghstack-poisoned]
Skylion007
reviewed
Oct 14, 2025
| */ | ||
| float scale_val = scale[0].item<float>(); | ||
|
|
||
| bool is_bfloat16 = (X.scalar_type() == at::kBFloat16); |
Collaborator
There was a problem hiding this comment.
So we shouldn't cast fp16
Contributor
Author
There was a problem hiding this comment.
we enabled bf16 support for per_tensor alongside per_channel in this pr #165325, so if we want to enable fp16 we can do it in a separate pr for both these ops?
| auto dScale = dScale_vec.sum().unsqueeze(0).to(scale_.device()); | ||
| auto dZeroPoint = dZeroPoint_vec.sum().unsqueeze(0).to(zero_point_.device()); | ||
|
|
||
| return std::make_tuple(dX, dScale, dZeroPoint); |
Collaborator
There was a problem hiding this comment.
All these should be std::move into make tuple btw
Contributor
Author
There was a problem hiding this comment.
should we make this change in a follow up PR since this PR doesn't actually touch this code?
|
@liangel-02 has imported this pull request. If you are a Meta employee, you can view this in D84639869. |
Contributor
Author
|
@pytorchbot merge |
Collaborator
Merge startedYour 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 |
Chao1Han
pushed a commit
to Chao1Han/pytorch
that referenced
this pull request
Oct 21, 2025
Adding bf16 for the backward pass of `torch._fake_quantize_learnable_per_tensor_affine()`. Note that for testing, we modified the seed to avoid increasing tolerance due to cases where difference in Python vs CPP downcasting causes tensor mismatches. (e.g. 27.87704 vs 27.8408 before downcasting, 27.7500 vs 27.8750 after downcasting for Python vs CPP op) Pull Request resolved: pytorch#165362 Approved by: https://github.com/andrewor14
zhudada0120
pushed a commit
to zhudada0120/pytorch
that referenced
this pull request
Oct 22, 2025
Adding bf16 for the backward pass of `torch._fake_quantize_learnable_per_tensor_affine()`. Note that for testing, we modified the seed to avoid increasing tolerance due to cases where difference in Python vs CPP downcasting causes tensor mismatches. (e.g. 27.87704 vs 27.8408 before downcasting, 27.7500 vs 27.8750 after downcasting for Python vs CPP op) Pull Request resolved: pytorch#165362 Approved by: https://github.com/andrewor14
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
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
4 participants
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.
Adding bf16 for the backward pass of
torch._fake_quantize_learnable_per_tensor_affine().Note that for testing, we modified the seed to avoid increasing tolerance due to cases where difference in Python vs CPP downcasting causes tensor mismatches. (e.g. 27.87704 vs 27.8408 before downcasting, 27.7500 vs 27.8750 after downcasting for Python vs CPP op)
Stack from ghstack (oldest at bottom):