Training VLM

[faithfulnguyen]

Hi thanks for sharing your work, I have a small dataset including images and descriptions, Can I use this code for training on my dataset?

[SumanthRH]

Hi!

Could you highlight which recipe you're trying to extend with image data? We've summarized all the recipes for the different models here:

Currently, the training code is made up of two forks :

• LlamaFactory - Used for Sky-T1-32B-Preview and Sky-T1-32B-Flash. Llamafactory supports using image data.
• VERL - Used for Sky-T1-mini. AFAIK VeRL is text-only, so you might have to customize the code for working with VLMs.

We're also actively working on cleaning up our training code.

Hope that helps!