biondizzle/vllm
1
0
Fork 0
Code Issues Pull Requests Actions 2 Packages Projects Releases Wiki Activity

[FEAT] [ROCm]: Add AITER CK 2 Stages MoE support (#17110)

Browse Source 
Signed-off-by: tjtanaa <tunjian.tan@embeddedllm.com>
Co-authored-by: Gregory Shtrasberg <Gregory.Shtrasberg@amd.com>
This commit is contained in:
TJian
2025-05-14 18:03:11 +08:00
committed by GitHub
parent 38fe728d60
commit 612c2edb4f
7 changed files with 201 additions and 112 deletions
Download Patch File Download Diff File Expand all files Collapse all files

3
vllm/model_executor/layers/fused_moe/fused_moe.py
View File

@@ -1100,9 +1100,6 @@ def torch_vllm_outplace_fused_experts(**kwargs) -> torch.Tensor:
def dispatch_fused_experts_func(inplace: bool) -> Callable[..., torch.Tensor]:
if is_rocm_aiter_moe_enabled():
from .rocm_aiter_fused_moe import rocm_aiter_fused_experts
return rocm_aiter_fused_experts
if inplace:
return torch_vllm_inplace_fused_experts
return torch_vllm_outplace_fused_experts