[MOE Refactor] Remove QuantType in prepare_finalize.py (#6534)
### What this PR does / why we need it? To prevent confusion between different QuantType classes, we remove** QuantType in prepare_finalize.py - vLLM version: v0.15.0 - vLLM main: https://github.com/vllm-project/vllm/commit/v0.15.0 Signed-off-by: shenchuxiaofugui <1311027364@qq.com>
This commit is contained in:
@@ -30,7 +30,6 @@ from vllm_ascend.ops.fused_moe.prepare_finalize import (
|
||||
PrepareAndFinalizeWithAll2All,
|
||||
PrepareAndFinalizeWithAllGather,
|
||||
PrepareAndFinalizeWithMC2,
|
||||
QuantType,
|
||||
)
|
||||
from vllm_ascend.ops.fused_moe.token_dispatcher import (
|
||||
MoETokenDispatcher,
|
||||
@@ -38,6 +37,7 @@ from vllm_ascend.ops.fused_moe.token_dispatcher import (
|
||||
TokenDispatcherWithAllGather,
|
||||
TokenDispatcherWithMC2,
|
||||
)
|
||||
from vllm_ascend.quantization.methods.base import QuantType
|
||||
|
||||
_MoECommMethods: dict[MoECommType | None, MoECommMethod] = {}
|
||||
|
||||
|
||||
Reference in New Issue
Block a user