You firstly need to prepare the dataset. you can follow this step to prepare the dataset. our dataset preparation.
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"
For ckpt, you may load from our pretrained model checkpoints:
MiniGPT-v2 (after stage-2) | MiniGPT-v2 (after stage-3) | MiniGPT-v2 (online developing demo) |
---|---|---|
Download | Download | Download |
torchrun --nproc-per-node NUM_GPU train.py --cfg-path train_configs/minigptv2_finetune.yaml