xinyun/vllm
1
0
Fork 0
mirror of https://git.datalinker.icu/vllm-project/vllm.git synced 2025-12-10 13:44:58 +08:00
Code Issues Packages Projects Releases Wiki Activity

[Bugfix] Fix MoeWNA16Method activation (#14024)

Browse Source
This commit is contained in:
Jee Jee Li 2025-02-28 23:22:42 +08:00 committed by GitHub
parent f58f8b5c96
commit e0734387fb
No known key found for this signature in database
GPG Key ID: B5690EEEBB952194
1 changed files with 2 additions and 1 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

3
vllm/model_executor/layers/quantization/moe_wna16.py
View File

@ -293,9 +293,10 @@ class MoeWNA16Method(FusedMoEMethodBase):
custom_routing_function: Optional[Callable] = None, custom_routing_function: Optional[Callable] = None,
scoring_func: str = "softmax", scoring_func: str = "softmax",
e_score_correction_bias: Optional[torch.Tensor] = None, e_score_correction_bias: Optional[torch.Tensor] = None,
activation: str = "silu",
) -> torch.Tensor: ) -> torch.Tensor:
from vllm.model_executor.layers.fused_moe import fused_experts from vllm.model_executor.layers.fused_moe import fused_experts
assert activation == "silu", "Only SiLU activation is supported."
topk_weights, topk_ids = FusedMoE.select_experts( topk_weights, topk_ids = FusedMoE.select_experts(
hidden_states=x, hidden_states=x,
router_logits=router_logits, router_logits=router_logits,