PaddleOCR-VL-1.5 fine-tuning not happening as expectation

#24
by Kaith-jeet123 - opened

By fine-tuning using the provided scripts of PaddleOCR-VL, the fine-tuned model for PaddleOCR-VL-1.5 is hallucinating a lot and it is performing worse than base model.

Is this because of the latest version packages updates? Has anyone the solution of it?

Sign up or log in to comment