[Feature]: Remove Chunking From FusedMoE (#34086)

Signed-off-by: SouthWest7 <am1ao@qq.com>
Signed-off-by: Southwest <1403572259@qq.com>
Signed-off-by: southwest <am1ao@qq.com>
Signed-off-by: Xinan Miao <1403572259@qq.com>
Co-authored-by: SouthWest7 <am1ao@qq.com>
This commit is contained in:
Xinan Miao
2026-03-12 14:24:38 -04:00
committed by GitHub
co-authored by SouthWest7
parent c973ecdead
commit 2cdf92228c
28 changed files with 152 additions and 523 deletions
+2 -4
View File
@@ -190,9 +190,8 @@ class FusedMoEWithLoRA(BaseLayerWithLoRA):
use_int8_w8a16=False,
use_int4_w4a16=False,
)
CHUNK_SIZE = envs.VLLM_FUSED_MOE_CHUNK_SIZE
num_tokens = hidden_states.size(0)
M = min(num_tokens, CHUNK_SIZE)
M = num_tokens
max_lora_rank = self.w13_lora_a_stacked[0].shape[-2]
shrink_config, expand_config = self._get_lora_moe_configs(
op_prefix="w13",
@@ -281,9 +280,8 @@ class FusedMoEWithLoRA(BaseLayerWithLoRA):
use_int8_w8a16=False,
use_int4_w4a16=False,
)
CHUNK_SIZE = envs.VLLM_FUSED_MOE_CHUNK_SIZE
num_tokens = hidden_states.size(0)
M = min(num_tokens, CHUNK_SIZE)
M = num_tokens
max_lora_rank = self.w2_lora_a_stacked[0].shape[-2]
shrink_config, expand_config = self._get_lora_moe_configs(
op_prefix="w2",