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

[XPU][CI] enhance xpu test support (#20652)

Browse Source 
Signed-off-by: Ma, Liangliang <liangliang.ma@intel.com>
Co-authored-by: zhenwei-intel <zhenweiliu@habana.ai>
This commit is contained in:
Liangliang Ma
2025-07-10 00:53:09 +08:00
committed by GitHub
parent eb58f5953d
commit a3e4e85ece
5 changed files with 18 additions and 12 deletions
Download Patch File Download Diff File Expand all files Collapse all files

5
tests/conftest.py
View File

@@ -759,7 +759,8 @@ class VllmRunner:
- `trust_remote_code`: Set to `True` instead of `False` for convenience.
- `seed`: Set to `0` instead of `None` for test reproducibility.
- `max_model_len`: Set to `1024` instead of `None` to reduce memory usage.
- `block_size`: Set to `16` instead of `None` to reduce memory usage.
- `block_size`: To reduce memory usage, set default to `64` if on XPU
devices, otherwise default to `16`.
- `enable_chunked_prefill`: Set to `False` instead of `None` for
test reproducibility.
- `enforce_eager`: Set to `False` to test CUDA graph.
@@ -777,7 +778,7 @@ class VllmRunner:
dtype: str = "auto",
disable_log_stats: bool = True,
tensor_parallel_size: int = 1,
block_size: int = 16,
block_size: int = 16 if not torch.xpu.is_available() else 64,
enable_chunked_prefill: Optional[bool] = False,
swap_space: int = 4,
enforce_eager: Optional[bool] = False,