xinyun/vllm
1
0
Fork 0
mirror of https://git.datalinker.icu/vllm-project/vllm.git synced 2026-04-25 11:07:11 +08:00
Code Issues Packages Projects Releases Wiki Activity

fixes - use-fp8-dispatch

Browse Source 
Signed-off-by: Varun Sundar Rabindranath <vsundarr@redhat.com>
This commit is contained in:
Varun Sundar Rabindranath 2025-06-18 11:15:48 -07:00 committed by Tyler Michael Smith
parent 299f829180
commit 2b5ad9f233
1 changed files with 1 additions and 1 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

2
vllm/model_executor/layers/fused_moe/layer.py
View File

@ -381,7 +381,7 @@ class FusedMoEMethodBase(QuantizeMethodBase):
# Note : We may want to use FP8 dispatch even otherwise just to
# reduce datamovement
use_fp8_dispatch = (quant_dtype == current_platform.fp8_dtype()
and act_quant_block_size
and act_quant_block_size[1]
== DEEPEP_QUANT_BLOCK_SIZE)
# Note (varun): Whether to use FP8 dispatch or not needs some