Tips for Fine-Tuning Gemini Models on Custom Data

Hello all,
I’m exploring fine-tuning Gemini models with some domain-specific datasets, but I’m new to the process and want to avoid common pitfalls.

Could anyone share:

  • Best practices for preparing and formatting training data?
  • Recommendations on hyperparameter settings for stable training?
  • Any tools or scripts that have helped streamline fine-tuning in 2025?

Thanks in advance for your advice!

@XyzApkHUB ,

welcome to the community,

Fine tuning gemini is supports only on vertex ai at this point. there is a detailed documentation on how to do this, please find the below documentation for further references

thankyou