Loading a vision lora fails with ValueError: Unrecognized model in lora_model. Should have a model_type key in its config.json

[saum7800]

Ran the notebook for Qwen2 VL directly and ran into this issue when trying to load trained and saved lora model

not sure why the bug appears here but not in FastLanguageModel

[KarimMusleh]

Can you link it please, I ran both of the Qwen notebooks and they seem to be working fine

[saum7800]

Did you change the False in the second last cell to True?

https://colab.research.google.com/drive/12pWo14IBaLgtlbxnsp-ocy59d8XujDZw?usp=sharing

interestingly, if I add qwen in the name of saved lora, it works, but that seems hacky

[danielhanchen]

Can repro - will fix it asap - sorry on the issue guys!