Model does not produce meaningful output for VQA
#6
by rpeinl - opened
Model always produces "#####################################################" as the only output if you ask questions about images.
The problem is solved with using transformers==4.49.0 instead of the suggested version from the model card
see: https://huggingface.co/Qwen/Qwen2.5-VL-7B-Instruct-AWQ/discussions/2