EngineX/xc-llm-kunlun
3
0
Fork 0
Code Issues Pull Requests Actions Projects Releases Wiki Activity

提交vllm0.11.0开发分支

Browse Source
This commit is contained in:
chenyili
2025-12-10 17:51:24 +08:00
parent deab7dd0b6
commit 7c22d621fb
175 changed files with 31856 additions and 8683 deletions
Download Patch File Download Diff File Expand all files Collapse all files

14
setup_env.sh Normal file → Executable file
View File

@@ -1,11 +1,13 @@
unset XPU_DUMMY_EVENT
export XPU_VISIBLE_DEVICES=0,1,2,3,4,5,6,7
export XPU_USE_MOE_SORTED_THRES=1
export XFT_USE_FAST_SWIGLU=1
export XFT_USE_FAST_SWIGLU=1 #使用快速swiglu实现
export XPU_USE_FAST_SWIGLU=1 #使用moe算子中快速swiglu实现
export XMLIR_CUDNN_ENABLED=1
export XPU_USE_DEFAULT_CTX=1
export XMLIR_FORCE_USE_XPU_GRAPH=1
export XPU_USE_FAST_SWIGLU=1
export XMLIR_FORCE_USE_XPU_GRAPH=1 # 优化图间sync
export XPU_USE_MOE_SORTED_THRES=128 # Moe sort threshold
export VLLM_HOST_IP=$(hostname -i)
export XMLIR_ENABLE_MOCK_TORCH_COMPILE=false
export FUSED_QK_ROPE_OP=0
export XMLIR_ENABLE_MOCK_TORCH_COMPILE=false
VLLM_USE_V1=1
##默认值为1，设置为0启用QWN3融合大算子
USE_ORI_ROPE=1