Skip to content

fix precision of fp8 moe#737

Open
devalshahamd wants to merge 4 commits into
mainfrom
feat/perf_0614
Open

fix precision of fp8 moe#737
devalshahamd wants to merge 4 commits into
mainfrom
feat/perf_0614

Conversation

@devalshahamd

Copy link
Copy Markdown
Contributor

This pull request makes a targeted change to the get_compute_precision method in moe_perf_model_extensions.py to ensure that when the input data type is either "fp8" or "bf16", the method returns "fp8" instead of mapping the type using torch_dtype_map. This likely addresses a specific requirement for handling these data types in the performance model.

  • Precision handling update:
    • Modified the get_compute_precision method in moe_perf_model_extensions.py to return "fp8" when the input data type is "fp8" or "bf16", otherwise it falls back to torch_dtype_map(dtype)

@devalshahamd devalshahamd marked this pull request as ready for review June 19, 2026 16:17
@devalshahamd devalshahamd requested review from Copilot and tsrikris June 19, 2026 16:17

Copilot AI left a comment

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

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

Pull request overview

This PR updates the MoE performance model’s compute-precision selection logic to special-case certain input dtypes, aiming to treat "fp8" and "bf16" inputs as "fp8" compute precision for the fused blockscale MoE model.

Changes:

  • Adjusted get_compute_precision() in the AITER FP8 block-scale fused MoE extension to return "fp8" for "fp8"/"bf16" input dtypes instead of always using torch_dtype_map().

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

Comment thread TraceLens/PerfModel/extensions/moe_perf_model_extensions.py
@devalshahamd devalshahamd requested a review from gabeweisz June 19, 2026 16:26
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

4 participants