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

[Bug] [ROCm] Fix Llama 4 Enablement Bug on ROCm: V0 ROCmFlashAttentionImpl and Triton Fused MoE bugs (#16198)

Browse Source 
Signed-off-by: tjtanaa <tunjian.tan@embeddedllm.com>
Signed-off-by: kliuae <kuanfu.liu@embeddedllm.com>
Co-authored-by: Hongxia Yang <hongxia.yang@amd.com>
Co-authored-by: kliuae <kuanfu.liu@embeddedllm.com>
This commit is contained in:
TJian
2025-04-09 10:12:34 +08:00
committed by GitHub
parent 102bf967f0
commit 2976dc27e9
3 changed files with 15 additions and 9 deletions
Download Patch File Download Diff File Expand all files Collapse all files

2
vllm/model_executor/layers/fused_moe/fused_moe.py
View File

@@ -1002,6 +1002,7 @@ direct_register_custom_op(
op_func=inplace_fused_experts,
mutates_args=["hidden_states"],
fake_impl=inplace_fused_experts_fake,
tags=(torch.Tag.needs_fixed_stride_order, ),
)
@@ -1060,6 +1061,7 @@ direct_register_custom_op(
op_func=outplace_fused_experts,
mutates_args=[],
fake_impl=outplace_fused_experts_fake,
tags=(torch.Tag.needs_fixed_stride_order, ),
)