Hope to get the training code

[jjkinga]

I really like your work! Is there any plan to open source the training process code?

[vogel-alex]

Thank you — I’m glad you like the work! At the moment, there are no plans to open source the training code as we relied on the existing training scripts provided by the underlying models, with sometimes minor adaptations to enable lora training.

Some examples:

• https://github.com/X-PLUG/mPLUG-DocOwl/blob/main/TinyChart/scripts/train.sh
• https://github.com/X-PLUG/mPLUG-DocOwl/blob/main/DocOwl1.5/scripts/finetune_docowl_lora.sh
  (include the following for successful application of the trained LoRA weights - [Bug Report] DocOwl1.5 unable to load Peft models X-PLUG/mPLUG-DocOwl#123)
• https://github.com/QwenLM/Qwen-VL/blob/master/finetune/finetune_lora_ds.sh