search

SAIS-FUXI / EvalAlign

Apache License 2.0
13 stars 0 forks source link

How to train #3

Open conquer-pan opened 3 months ago

conquer-pan commented 3 months ago

Is the training code of the project open source? How do I train my model on my own data set? Please tell me

SAIS-FUXI commented 3 months ago

Since our training code is based on the llava framework for SFT and we have not modified any llava code, you can simply refer to the llava finetuning code. By converting your data into the format supported by the llava training scripts, you will be ready to start training.

Fr0zenCrane commented 3 months ago

Hi, @conquer-pan:

EvalAlign is essentially a finetuned LLaVA v1.6, you can just use the finetuning code provided by this repo.

conquer-pan commented 3 months ago

Hi, @conquer-pan:

EvalAlign is essentially a finetuned LLaVA v1.6, you can just use the finetuning code provided by this repo.

thank！