Use the feature of resume_from_checkpoint when using python code to run finetuning.

[Jianshu-Hu]

Dear authors,
Thank you so much for this valuable repository!

I have followed the tuturial from https://github.com/modelscope/ms-swift/blob/main/examples/notebook/qwen2vl-ocr/ocr-sft.ipynb to write a python code for finetuning with my own dataset. I am wondering how should I modify the current code for resuming from a saved checkpoint. I have tried to include this argument resume_from_checkpoint in the training_args by setting resume_from_checkpoint to the path of the saved checkpoint. However, the model is not loaded and the training just starts from scratch.

Thank you in advance for helping with this.

Best,
Jianshu

[Jintao-Huang]

It is recommended to use shell for training, which will be easier to get started.