forked from vllm-project/vllm
-
Notifications
You must be signed in to change notification settings - Fork 29
Commit
This commit does not belong to any branch on this repository, and may belong to a fork outside of the repository.
Fix correctness regression (from PR#258) in Llama-3.2-90B-Vision-Inst…
…ruct-FP8-KV test (#294) * Fix correctness regression in Llama-3.2-90B-Vision-Instruct-FP8-KV test * Fixed platform substitution * Re-remove vectorization on Navi * Typo * Using the same signatures in both paths * Using 3 versions of RMS norm kernel: 1. Optimal vectorization using _f16Vec for supported types; 2. Fallback for types that don't support conversion; 3. Fallback for shapes that can't be vectorized * clang-format * Thinking of it, we don't really need the alternative vectorized kernel * clang-format * Using 64 bit types for indices --------- Co-authored-by: wunhuang <[email protected]> Co-authored-by: Gregory Shtrasberg <[email protected]>
- Loading branch information
1 parent
529cefe
commit 6cf8eb4
Showing
1 changed file
with
16 additions
and
24 deletions.
There are no files selected for viewing
This file contains 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