Training Scripts and Instructions

[muzairkhattak]

Dear Authors,

Thank you for your great work and contribution to the community!

We are interested in exploring/expanding on this project and have some inquiries as follows:

1. Does the current repository include training code and relevant scripts? As the first step, we would like to reproduce your model from scratch. If you could provide a bit of instruction on how to train both stages of the model, it would be really helpful.

2. We are preparing the RSNA Pneumonia dataset but could not find the training/testing (jpg) images that are present in your provided JSON files on the official website. Is it possible for us to get a link to directly download the preprocessed images?

Looking forward to your kind response!
Thank you!

Regards,
Muhammad Uzair

[Asma-Alkhaldi]

1 - For training please follow these steps:
In the train_configs/minigptv2_finetune.yaml, you need to set up the following paths:
llama_model checkpoint path: "/path/to/llama_checkpoint"
ckpt: "/path/to/pretrained_checkpoint"
ckpt save path: "/path/to/save_checkpoint"

And then type this command to start training:
torchrun --nproc-per-node NUM_GPU train.py --cfg-path train_configs/minigptv2_finetune.yaml

2- link for the RSNA preprocessed data:
https://drive.google.com/drive/folders/1g-HgZPwa2yPtEeN-i-8eQ3IORtwnOVNZ?usp=sharing