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

fix multiproc executor determine kv cache memory & update Dockerfile

Browse Source 
Signed-off-by: Jing Wang <jingwang96@qq.com>
This commit is contained in:
Jing Wang
2026-04-24 08:31:54 +00:00
parent 6c097beaa5
commit d627a45881
10 changed files with 218 additions and 153 deletions
Download Patch File Download Diff File Expand all files Collapse all files

5
vllm_ascend/patch/platform/patch_executor.py
View File

@@ -42,11 +42,6 @@ def reload_vram(self) -> bool:
time.sleep(0.001)
def determine_available_memory_vnpu_offload_mode(self) -> int:
return self.collective_rpc("determine_available_memory_vnpu_offload_mode")
Executor.is_offloaded = is_offloaded
Executor.offload_vram = offload_vram
Executor.reload_vram = reload_vram
Executor.determine_available_memory_vnpu_offload_mode = determine_available_memory_vnpu_offload_mode