xinyun/vllm
1
0
Fork 0
mirror of https://git.datalinker.icu/vllm-project/vllm.git synced 2026-05-24 09:51:19 +08:00
Code Issues Packages Projects Releases Wiki Activity

[Bugfix] Fix isinstance check for tensor types in _load_prompt_embeds to use dtype comparison (#21612)

Browse Source 
Signed-off-by: Alexandre Juan <a.juan@netheos.net>
This commit is contained in:
Alexandre JUAN 2025-07-26 05:11:10 +02:00 committed by GitHub
parent a55c95096b
commit 2f6e6b33fb
No known key found for this signature in database
GPG Key ID: B5690EEEBB952194
1 changed files with 5 additions and 3 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

8
vllm/entrypoints/openai/serving_engine.py
View File

@ -957,9 +957,11 @@ class OpenAIServing:
def _load_and_validate_embed(embed: bytes) -> EmbedsPrompt: def _load_and_validate_embed(embed: bytes) -> EmbedsPrompt:
tensor = torch.load(io.BytesIO(base64.b64decode(embed)), tensor = torch.load(io.BytesIO(base64.b64decode(embed)),
weights_only=True) weights_only=True)
assert isinstance( assert isinstance(tensor, torch.Tensor) and tensor.dtype in (
tensor, torch.float32,
(torch.FloatTensor, torch.BFloat16Tensor, torch.HalfTensor)) torch.bfloat16,
torch.float16,
)
if tensor.dim() > 2: if tensor.dim() > 2:
tensor = tensor.squeeze(0) tensor = tensor.squeeze(0)
assert tensor.dim() == 2 assert tensor.dim() == 2