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

[Bugfix] Remove cuda related lines and add additional pip mirror (#1252)

Browse Source 
### What this PR does / why we need it?
- For npu environment, we should use `PYTORCH_NPU_ALLOC_CONF ` rather
than `PYTORCH_CUDA_ALLOC_CONF`
- Add `PIP_EXTRA_INDEX_URL` to make nightly_benchmarks happy


---------

Signed-off-by: wangli <wangli858794774@gmail.com>
This commit is contained in:
Li Wang
2025-06-17 21:25:40 +08:00
committed by GitHub
parent 23ca68d0c8
commit f8029945c3
2 changed files with 3 additions and 1 deletions
Download Patch File Download Diff File Expand all files Collapse all files

2
vllm_ascend/device_allocator/camem.py
View File

@@ -144,7 +144,7 @@ class CaMemAllocator:
return CaMemAllocator.instance
def __init__(self):
conf = os.environ.get("PYTORCH_CUDA_ALLOC_CONF", "")
conf = os.environ.get("PYTORCH_NPU_ALLOC_CONF", "")
assert "expandable_segments:True" not in conf, \
("Expandable segments are not compatible with memory pool. "
"Please track https://github.com/pytorch/pytorch/issues/147851 "