How to load the sft finetuned model based on the diffusers version.

[zyhbili]

Thanks for your great work! The recently released training scripts are quite user-friendly.
I have done the sft finetuning. And i convert the checkpoints using tools/convert_weight_deepspeed2hf.py.
But i fail to load finetuned weight when i specify the new weight path to --model_path in inference/cli_demo.py.
It seems due to the model_index.json is missing under new path.
How to load it in an efficient and straightforward manner?

[zRzRzRzRzRzRzR]

If you are missing model_index.json, you can check the contents that are missing from the original model file, as this file serves as the index for the split safetensors. If the number of .safetensors in your exported model is the same as in the original model, you can directly copy the model_index.json from the original model.