[MoE Refactor] Rename FusedMoE.make_expert_params_mapping to fused_moe_make_expert_params_mapping (#40671)

Signed-off-by: Bill Nell <bnell@redhat.com>
This commit is contained in:
bnellnm
2026-04-23 11:22:34 -04:00
committed by GitHub
parent 8824f50f1f
commit 1c2c1eb8b9
53 changed files with 254 additions and 98 deletions
+4 -2
View File
@@ -40,7 +40,9 @@ from vllm.config.multimodal import BaseDummyOptions
from vllm.distributed import get_tensor_model_parallel_world_size
from vllm.inputs import MultiModalDataDict
from vllm.model_executor.layers.attention import MMEncoderAttention
from vllm.model_executor.layers.fused_moe import FusedMoE
from vllm.model_executor.layers.fused_moe import (
fused_moe_make_expert_params_mapping,
)
from vllm.model_executor.layers.linear import (
ColumnParallelLinear,
QKVParallelLinear,
@@ -1072,7 +1074,7 @@ class Llama4ForConditionalGeneration(
def get_expert_mapping(self) -> list[tuple[str, str, int, str]]:
# Params for weights, fp8 weight scales, fp8 activation scales
# (param_name, weight_name, expert_id, shard_id)
return FusedMoE.make_expert_params_mapping(
return fused_moe_make_expert_params_mapping(
self,
ckpt_gate_proj_name="gate_proj",
ckpt_down_proj_name="down_proj",