Finetuning script for Gemma3

[2U1]

https://github.com/2U1/Gemma3-Finetune

I made a code for who wants to use the huggingface version to finetune, and having difficult using some other frameworks like me.

This code only uses huggingface for fine-tuning the 4B, 12B and 27B.

Also, you can set different learning_rate for vision_tower and language_model. ( Also for the projector)

Feedback and issues are welcome!

[Gopi-Uppari]

Hi @2U1,

Thanks for sharing this. It's really helpful for those using Hugging Face. You experimented with different types of data and fine-tuned the model. We'll try out your code for various use cases.

Thank you.

[2U1]

@Gopi-Uppari Thanks for the interest in the project! Your expirience would be very valuable for developing the project.

Feedbacks and issues are always welcome!