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

[Core] [Bugfix] [Multimodal] Fix multimodal profiling and generation for SFT/PTQed models (#20058)

Browse Source 
Signed-off-by: Kyle Sayers <kylesayrs@gmail.com>
This commit is contained in:
Kyle Sayers
2025-06-30 13:26:49 -04:00
committed by GitHub
parent 551ef1631a
commit d8cf819a9a
41 changed files with 207 additions and 38 deletions
Download Patch File Download Diff File Expand all files Collapse all files

5
vllm/model_executor/models/internvl.py
View File

@@ -758,11 +758,13 @@ class BaseInternVLMultiModalProcessor(BaseMultiModalProcessor[_I]):
prompt: str,
mm_data: Mapping[str, object],
mm_kwargs: Mapping[str, object],
tok_kwargs: Mapping[str, object],
) -> Mapping[str, NestedTensors]:
processed_outputs = super()._call_hf_processor(
prompt=prompt,
mm_data=mm_data,
mm_kwargs=mm_kwargs,
tok_kwargs=tok_kwargs,
)
hf_processor = self.info.get_hf_processor(**mm_kwargs)
@@ -941,9 +943,10 @@ class InternVLMultiModalProcessor(
prompt: str,
mm_data: Mapping[str, object],
mm_kwargs: Mapping[str, object],
tok_kwargs: Mapping[str, object],
) -> Mapping[str, NestedTensors]:
processed_outputs = super()._call_hf_processor(prompt, mm_data,
mm_kwargs)
mm_kwargs, tok_kwargs)
hf_processor = self.info.get_hf_processor(**mm_kwargs)
if self.info.supports_video and (