Skip to content

【cherry-pick】upgrade triton moe config. #7979

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Open
xuanyuanminzheng wants to merge 1 commit into
base: release/2.6
Choose a base branch
from
+314 −45
Open

【cherry-pick】upgrade triton moe config. #7979

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
173 changes: 159 additions & 14 deletions fastdeploy/model_executor/layers/moe/fused_moe_triton_backend.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -1879,17 +1879,169 @@ def process_loaded_weights(self, layer: nn.Layer, state_dict):

def _get_default_config(self, M: int, E: int) -> dict:
"""
Heuristic tile config for BF16 MoE, ported verbatim from vLLM's
`get_default_config` (bf16/fp16 non-block_shape branch).
See vllm/model_executor/layers/fused_moe/fused_moe.py:1273-1319.
GPU-aware heuristic tile config for BF16 MoE.

M: number of tokens (A.size(0) in vLLM), i.e. pre-expansion token count.
SM100 (B200): nearest-key lookup from SGLang tuned config
(triton_3_5_1/E=64,N=1856,device_name=NVIDIA_B200.json).
Others: original vLLM-ported heuristic.

M: number of tokens (pre-expansion token count).
E: number of (local) experts.
"""
from fastdeploy.model_executor.utils import get_sm_version

This comment was marked as outdated.

Sign in to view


if get_sm_version() >= 100:
# SM100 (B200): use SGLang tuned lookup, nearest key by abs diff
_SM100_CONFIGS = {
1: {
"BLOCK_SIZE_M": 16,
"BLOCK_SIZE_N": 64,
"BLOCK_SIZE_K": 64,
"GROUP_SIZE_M": 32,
"num_warps": 4,
"num_stages": 5,
},
2: {
"BLOCK_SIZE_M": 16,
"BLOCK_SIZE_N": 64,
"BLOCK_SIZE_K": 128,
"GROUP_SIZE_M": 32,
"num_warps": 4,
"num_stages": 3,
},
4: {
"BLOCK_SIZE_M": 16,
"BLOCK_SIZE_N": 64,
"BLOCK_SIZE_K": 128,
"GROUP_SIZE_M": 64,
"num_warps": 4,
"num_stages": 4,
},
8: {
"BLOCK_SIZE_M": 16,
"BLOCK_SIZE_N": 64,
"BLOCK_SIZE_K": 128,
"GROUP_SIZE_M": 32,
"num_warps": 4,
"num_stages": 3,
},
16: {
"BLOCK_SIZE_M": 16,
"BLOCK_SIZE_N": 64,
"BLOCK_SIZE_K": 128,
"GROUP_SIZE_M": 1,
"num_warps": 4,
"num_stages": 3,
},
24: {
"BLOCK_SIZE_M": 16,
"BLOCK_SIZE_N": 128,
"BLOCK_SIZE_K": 128,
"GROUP_SIZE_M": 16,
"num_warps": 4,
"num_stages": 4,
},
32: {
"BLOCK_SIZE_M": 16,
"BLOCK_SIZE_N": 64,
"BLOCK_SIZE_K": 128,
"GROUP_SIZE_M": 16,
"num_warps": 4,
"num_stages": 4,
},
48: {
"BLOCK_SIZE_M": 16,
"BLOCK_SIZE_N": 64,
"BLOCK_SIZE_K": 128,
"GROUP_SIZE_M": 1,
"num_warps": 4,
"num_stages": 4,
},
64: {
"BLOCK_SIZE_M": 16,
"BLOCK_SIZE_N": 64,
"BLOCK_SIZE_K": 128,
"GROUP_SIZE_M": 1,
"num_warps": 4,
"num_stages": 4,
},
96: {
"BLOCK_SIZE_M": 16,
"BLOCK_SIZE_N": 64,
"BLOCK_SIZE_K": 128,
"GROUP_SIZE_M": 1,
"num_warps": 4,
"num_stages": 3,
},
128: {
"BLOCK_SIZE_M": 16,
"BLOCK_SIZE_N": 64,
"BLOCK_SIZE_K": 128,
"GROUP_SIZE_M": 1,
"num_warps": 4,
"num_stages": 3,
},
256: {
"BLOCK_SIZE_M": 32,
"BLOCK_SIZE_N": 128,
"BLOCK_SIZE_K": 64,
"GROUP_SIZE_M": 1,
"num_warps": 4,
"num_stages": 5,
},
512: {
"BLOCK_SIZE_M": 64,
"BLOCK_SIZE_N": 256,
"BLOCK_SIZE_K": 64,
"GROUP_SIZE_M": 1,
"num_warps": 8,
"num_stages": 5,
},
1024: {
"BLOCK_SIZE_M": 128,
"BLOCK_SIZE_N": 256,
"BLOCK_SIZE_K": 64,
"GROUP_SIZE_M": 1,
"num_warps": 8,
"num_stages": 4,
},
1536: {
"BLOCK_SIZE_M": 256,
"BLOCK_SIZE_N": 256,
"BLOCK_SIZE_K": 64,
"GROUP_SIZE_M": 1,
"num_warps": 8,
"num_stages": 3,
},
2048: {
"BLOCK_SIZE_M": 256,
"BLOCK_SIZE_N": 256,
"BLOCK_SIZE_K": 64,
"GROUP_SIZE_M": 1,
"num_warps": 8,
"num_stages": 3,
},
3072: {
"BLOCK_SIZE_M": 128,
"BLOCK_SIZE_N": 256,
"BLOCK_SIZE_K": 64,
"GROUP_SIZE_M": 1,
"num_warps": 8,
"num_stages": 4,
},
4096: {
"BLOCK_SIZE_M": 256,
"BLOCK_SIZE_N": 256,
"BLOCK_SIZE_K": 64,
"GROUP_SIZE_M": 1,
"num_warps": 8,
"num_stages": 3,
},
}
best_key = min(_SM100_CONFIGS.keys(), key=lambda x: abs(x - M))
return _SM100_CONFIGS[best_key]

# Tile sizes scale with batch: small batches are memory-bound
# (favor tall-K tiles), large batches are compute-bound (favor
# large M/N tiles with more warps).
# Default heuristic for all other GPUs (ported from vLLM)
if M <= 32:
block_m = 16
elif M <= 96:
Expand All @@ -1900,19 +2052,12 @@ def _get_default_config(self, M: int, E: int) -> dict:
block_m = 128

block_n = 64 if M <= 64 else 128

block_k = 64

# Grouping adjacent M-blocks lets them share weight tiles in L2.
# Only helps when there are enough M-blocks per expert to group;
# with many experts each one sees few tokens so grouping is useless.
tokens_per_expert = M // max(E, 1)
group_m = 16 if tokens_per_expert > 128 else 1

# Large batches have enough blocks to saturate the GPU, so we
# use more warps per block to increase arithmetic intensity.
num_warps = 4 if M <= 128 else 8

num_stages = 4 if M <= 32 else 3

return {
Expand Down
Loading
Loading