xinyun/vllm
1
0
Fork 0
mirror of https://git.datalinker.icu/vllm-project/vllm.git synced 2026-05-25 16:31:18 +08:00
Code Issues Packages Projects Releases Wiki Activity

[Bugfix] Revert MoE Triton Config Default (#12629)

Browse Source 
SUMMARY:
* previous PR for pulling in block configs also changed defaults
(https://github.com/vllm-project/vllm/pull/11589/files) for FP8
* this broke L4 MoE since there was not enough SHM for the default
configuration
* this reverts the non-block example to the default

Signed-off-by: rshaw@neuralmagic.com <rshaw@neuralmagic.com>
This commit is contained in:
Robert Shaw 2025-01-31 18:28:47 -05:00 committed by GitHub
parent 415f19474d
commit 145c2ff648
No known key found for this signature in database
GPG Key ID: B5690EEEBB952194
1 changed files with 11 additions and 30 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

41
vllm/model_executor/layers/fused_moe/fused_moe.py
View File

@ -660,36 +660,17 @@ def get_default_config(
is_marlin: bool, is_marlin: bool,
block_shape: Optional[List[int]] = None, block_shape: Optional[List[int]] = None,
) -> Dict[str, int]: ) -> Dict[str, int]:
if dtype == "fp8_w8a8": if dtype == "fp8_w8a8" and block_shape is not None:
if block_shape is None: # Block-wise quant: BLOCK_SIZE_N must be divisible by block_shape[0]
config = { # BLOCK_SIZE_K must be divisible by block_shape[1]
"BLOCK_SIZE_M": 128, config = {
"BLOCK_SIZE_N": 256, "BLOCK_SIZE_M": 64,
"BLOCK_SIZE_K": 128, "BLOCK_SIZE_N": block_shape[0],
"GROUP_SIZE_M": 32, "BLOCK_SIZE_K": block_shape[1],
"num_warps": 8, "GROUP_SIZE_M": 32,
"num_stages": 4, "num_warps": 4,
} "num_stages": 3,
if M <= E: }
config = {
"BLOCK_SIZE_M": 64,
"BLOCK_SIZE_N": 128,
"BLOCK_SIZE_K": 128,
"GROUP_SIZE_M": 1,
"num_warps": 4,
"num_stages": 4,
}
else:
# Block-wise quant: BLOCK_SIZE_N must be divisible by block_shape[0]
# BLOCK_SIZE_K must be divisible by block_shape[1]
config = {
"BLOCK_SIZE_M": 64,
"BLOCK_SIZE_N": block_shape[0],
"BLOCK_SIZE_K": block_shape[1],
"GROUP_SIZE_M": 32,
"num_warps": 4,
"num_stages": 3,
}
else: else:
config = { config = {
"BLOCK_SIZE_M": 64, "BLOCK_SIZE_M": 64,