how to finetune with gemma model？

[runningabcd]

how to finetune with gemma model？

[runningabcd]

I had download gemma-7b-it model from hugging face already, but not find the script that can finetune wiht my own data

[runningabcd]

help

[runningabcd]

I had download gemma-7b-it model from hugging face already, but not find the script that can finetune wiht my own data

How to sft with gemma？Can you tell me the sft data format?

[runningabcd]

I had download gemma-7b-it model from hugging face already, but not find the script that can finetune wiht my own data

How to sft with gemma？Can you tell me the sft data format?

@pengchongjin

[pengchongjin]

Hi there, unfortunately, this repo doesn't provide finetuning features.

Here are a few alternatives that might fit your needs:

1. On Gemma model card in Vertex Model Garden, there are a few notebooks which demonstrate how to do finetuning and then deploy to Vertex endpoints.
2. On Gemma model card in Kaggle, there are a few notebooks which uses KerasNLP to do finetuning.
3. HuggingFace demonstrates how to use TRL to do finetuning in this blog post.

Hope it helps.

[r-gheda]

@pengchongjin Is it possible to implement a class for fine tuning the model inside this repo similar to what done with llama-recipes?

[aliasneo1]

Are there any tutorials for fine-tuning 7b-it-quant model ?

[selamw1]

Hi @aliasneo1

There are a few tutorials that demonstrate fine-tuning the gemma-2b model. You can follow similar procedures to fine-tune the Gemma variant gemma-7b-it.

Here are some resources:

• Fine-tuning Gemma using JAX and Flax.
• Fine-Tuning Gemma Models in Hugging Face with PyTorch on GPU and TPU.