kanshan/TensorRT-LLMs
1
0
Fork 0
mirror of https://github.com/NVIDIA/TensorRT-LLM.git synced 2026-02-16 07:53:55 +08:00
Code Issues Actions 1 Packages Projects Releases Wiki Activity

[None][chore] Pass without_comm to cutlass and deepgemm (#11229)

Browse Source 
Signed-off-by: xxi <xxi@nvidia.com>
This commit is contained in:
xxi 2026-02-05 15:07:59 +08:00 committed by GitHub
parent 36cb5f8c93
commit 4c1d9d0c10
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
tensorrt_llm/_torch/modules/fused_moe/create_moe.py
View File

@ -196,6 +196,7 @@ def create_moe_backend(
swiglu_beta=swiglu_beta,
swiglu_limit=swiglu_limit,
init_load_balancer=init_load_balancer,
without_comm=without_comm,
activation_type=activation_type,
)
elif moe_cls == WideEPMoE:
@ -256,6 +257,7 @@ def create_moe_backend(
weight_loading_mode=weight_loading_mode,
apply_router_weight_on_input=apply_router_weight_on_input,
layer_idx=layer_idx,
without_comm=without_comm,
)
elif moe_cls == TritonFusedMoE:
assert not apply_router_weight_on_input, "apply_router_weight_on_input is not supported in TritonFusedMoE."