xinyun/vllm
1
0
Fork 0
mirror of https://git.datalinker.icu/vllm-project/vllm.git synced 2025-12-10 06:55:01 +08:00
Code Issues Packages Projects Releases Wiki Activity

optimize get_kv_cache_torch_dtype (#18531)

Browse Source 
Signed-off-by: idellzheng <idellzheng@tencent.com>
This commit is contained in:
chunxiaozheng 2025-05-27 21:08:44 +08:00 committed by GitHub
parent aaa4ac1c95
commit 6b6d496114
No known key found for this signature in database
GPG Key ID: B5690EEEBB952194
1 changed files with 3 additions and 4 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

7
vllm/utils.py
View File

@ -759,16 +759,15 @@ def get_kv_cache_torch_dtype(
model_dtype: Optional[Union[str, torch.dtype]] = None) -> torch.dtype:
if isinstance(cache_dtype, str):
if cache_dtype == "auto":
if isinstance(model_dtype, str):
if isinstance(model_dtype,
str) and model_dtype in STR_DTYPE_TO_TORCH_DTYPE:
torch_dtype = STR_DTYPE_TO_TORCH_DTYPE[model_dtype]
elif isinstance(model_dtype, torch.dtype):
torch_dtype = model_dtype
else:
raise ValueError(f"Invalid model dtype: {model_dtype}")
elif cache_dtype in ["half", "bfloat16", "float"]:
elif cache_dtype in STR_DTYPE_TO_TORCH_DTYPE:
torch_dtype = STR_DTYPE_TO_TORCH_DTYPE[cache_dtype]
elif cache_dtype == "fp8":
torch_dtype = torch.uint8
else:
raise ValueError(f"Invalid kv cache dtype: {cache_dtype}")
elif isinstance(cache_dtype, torch.dtype):