xinyun/vllm
1
0
Fork 0
mirror of https://git.datalinker.icu/vllm-project/vllm.git synced 2026-07-10 03:27:09 +08:00
Code Issues Packages Projects Releases Wiki Activity

Fix unknown attribute of topk_indices_dtype in CompressedTensorsW8A8Fp8MoECutlassMethod (#20507)

Browse Source 
Co-authored-by: Lucia (Lu) Fang <fanglu@meta.com>
This commit is contained in:
Lucia Fang 2025-07-05 14:03:20 +08:00 committed by GitHub
parent 906e05d840
commit 8aeaa910a2
No known key found for this signature in database
GPG Key ID: B5690EEEBB952194
1 changed files with 2 additions and 0 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

2
vllm/model_executor/layers/quantization/compressed_tensors/compressed_tensors_moe.py
View File

@ -368,6 +368,7 @@ class CompressedTensorsW8A8Fp8MoEMethod(CompressedTensorsMoEMethod):
"weights") "weights")
self.input_quant = self.quant_config.target_scheme_map["Linear"].get( self.input_quant = self.quant_config.target_scheme_map["Linear"].get(
"input_activations") "input_activations")
self.topk_indices_dtype = None
per_tensor = (self.weight_quant.strategy == QuantizationStrategy.TENSOR per_tensor = (self.weight_quant.strategy == QuantizationStrategy.TENSOR
and self.input_quant.strategy and self.input_quant.strategy
@ -738,6 +739,7 @@ class CompressedTensorsW8A8Fp8MoECutlassMethod(CompressedTensorsMoEMethod):
from vllm.model_executor.layers.fused_moe.cutlass_moe import ( from vllm.model_executor.layers.fused_moe.cutlass_moe import (
cutlass_moe_fp8) cutlass_moe_fp8)
self.topk_indices_dtype = None
self.fused_experts = cutlass_moe_fp8 # type: ignore self.fused_experts = cutlass_moe_fp8 # type: ignore
self.disable_expert_map = False self.disable_expert_map = False