vllm icon indicating copy to clipboard operation
vllm copied to clipboard

Published 20 hours ago • verified publisher icon vllm-project

[Bugfix][Model] Add base class for vision-language models

Open DarkLight1337 opened this issue 9 months ago • 1 comments

This PR adds a base class VLMBase to avoid importing LlavaForConditionalGeneration in vllm/model_executor/model_loader/loader.py, thus solving #4807.

Along the way, I have also ported the improved error handling logic regarding image_feature_size for LLaVA model.

FIX #4807

DarkLight1337 avatar May 14 '24 10:05 DarkLight1337

@simon-mo The models-test keeps getting interrupted, causing the CI to fail.

DarkLight1337 avatar May 17 '24 02:05 DarkLight1337

@DarkLight1337 @ywang96 @rkooo567 Is this PR ready for merge?

WoosukKwon avatar May 19 '24 06:05 WoosukKwon

@DarkLight1337 @ywang96 @rkooo567 Is this PR ready for merge?

Yes.

DarkLight1337 avatar May 19 '24 06:05 DarkLight1337