Exposes a fast_fp8_accum option to _scaled_mm #111847
Conversation
🔗 Helpful Links🧪 See artifacts and rendered test results at hud.pytorch.org/pr/111847
Note: Links to docs will display an error until the docs builds have been completed. ✅ No FailuresAs of commit 78c31e8 with merge base 93a9b13 ( This comment was automatically generated by Dr. CI and updates every 15 minutes. |
aten/src/ATen/cuda/CUDABlas.h
Outdated
| ScalarType bias_dtype, | ||
| void* result_ptr, | ||
| const void* result_scale_ptr, | ||
| const void *result_scale_ptr, |
There was a problem hiding this comment.
nit: void* to keep it consistent with lines above and below?
|
@pytorchbot merge |
Merge failedReason: This PR needs a If not, please add the To add a label, you can comment to pytorchbot, for example For more information, see Details for Dev Infra teamRaised by workflow job |
drisspg
added
topic: not user facing
|
@pytorchbot merge |
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 |
Merge failedReason: 1 jobs have failed, first few of them are: linux-binary-manywheel / manywheel-py3_8-cuda11_8-test / test Details for Dev Infra teamRaised by workflow job |
|
@pytorchbot merge -i |
Merge startedYour 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 |
# Summary Adds the option to use fast_accumulation_mode for the fp8 matmul in scaled_mm Information can be found here: https://docs.nvidia.com/cuda/cublas/#cublasltmatmuldescattributes-t defaults to 0 (off) Pull Request resolved: pytorch#111847 Approved by: https://github.com/ipiszy, https://github.com/malfet
# Summary Adds the option to use fast_accumulation_mode for the fp8 matmul in scaled_mm Information can be found here: https://docs.nvidia.com/cuda/cublas/#cublasltmatmuldescattributes-t defaults to 0 (off) Pull Request resolved: pytorch#111847 Approved by: https://github.com/ipiszy, https://github.com/malfet
Summary
Adds the option to use fast_accumulation_mode for the fp8 matmul in scaled_mm
Information can be found here: https://docs.nvidia.com/cuda/cublas/#cublasltmatmuldescattributes-t
defaults to 0 (off)
cc @yanbing-j @vkuzo @albanD @kadeng