xinyun/vllm
1
0
Fork 0
mirror of https://git.datalinker.icu/vllm-project/vllm.git synced 2025-12-14 02:54:57 +08:00
Code Issues Packages Projects Releases Wiki Activity

[Bugfix] [Performance] DeepEPHighThroughput + DeepSeek : Quant before Dispatch (#21837)

Browse Source 
Signed-off-by: Varun Sundar Rabindranath <vsundarr@redhat.com>
Co-authored-by: Varun Sundar Rabindranath <vsundarr@redhat.com>
This commit is contained in:
Varun Sundar Rabindranath 2025-08-01 22:44:38 +05:30 committed by GitHub
parent d6664664b4
commit ac45c44d98
No known key found for this signature in database
GPG Key ID: B5690EEEBB952194
1 changed files with 8 additions and 5 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

13
vllm/model_executor/layers/fused_moe/deepep_ht_prepare_finalize.py
View File

@ -144,12 +144,13 @@ class DeepEPHTPrepareAndFinalize(mk.FusedMoEPrepareAndFinalize):
"apply_router_weight_on_input is only implemented for topk=1")
a1 = a1 * topk_weights.to(a1.dtype)
if quant_config.per_act_token_quant:
if quant_config.is_block_quantized:
# Quant and Dispatch
a1q, a1q_scale = moe_kernel_quantize_input(
a1,
a1_scale,
quant_dtype=quant_config.quant_dtype,
per_act_token_quant=True,
per_act_token_quant=quant_config.per_act_token_quant,
block_shape=quant_config.block_shape,
)
if a1q_scale is not None and a1q_scale.numel() == 1:
@ -162,8 +163,10 @@ class DeepEPHTPrepareAndFinalize(mk.FusedMoEPrepareAndFinalize):
rank_topk_weights=topk_weights,
num_experts=num_experts)
else:
# DeepEP kernels only support dispatching per-token-quant
# quantization. dispatch in bfloat16.
# Dispatch and Quant
# DeepEP kernels only support dispatching block-quantized
# activation scales.
# Dispatch in bfloat16
(expert_x, _, expert_tokens_meta, expert_topk_ids,
expert_topk_weights) = self._do_dispatch(
tokens=a1,
@ -171,7 +174,7 @@ class DeepEPHTPrepareAndFinalize(mk.FusedMoEPrepareAndFinalize):
rank_topk_ids=topk_ids,
rank_topk_weights=topk_weights,
num_experts=num_experts)
# quantize now
# Quantize after dispatch.
expert_x_scale = None
if expert_x.numel() != 0:
expert_x, expert_x_scale = moe_kernel_quantize_input(