gue22/functiongemma-270m-it-mobile-actions

Model Card for functiongemma-mobile-actions

This model is a fine-tuned version of google/functiongemma-270m-it. It has been trained using TRL.

Training was done fully local on a PC with a 32GB Nvidia RTX Pro 4500 GPU (comparable to an RTX 5080) and took roughly 25 mins.
The script was derived from the Google Colab example and is available at ai-bits.org's FunctionGemma repo.

For the time being the litertlm model conversion for edge use (Andoid,..) is available in the functiongemma-mobile-actions-litertlm subdirectory here.

Quick start for the converted-to-litertlm model for Android

Install the Google AI Edge Gallery app from the Play Store. Start Edge Gallery.
In the mobile browser download the .litertlm model version (just one file) from the subdir here.
Click the bottom right plus button in the app to install the litertlm model from Downloads.
Try it in the now populated Mobile Actions widget.

Quick start in the README.md generated at fine-tuning

from transformers import pipeline

question = "If you had a time machine, but could only go to the past or the future once and never return, which would you choose and why?"
generator = pipeline("text-generation", model="None", device="cuda")
output = generator([{"role": "user", "content": question}], max_new_tokens=128, return_full_text=False)[0]
print(output["generated_text"])

Training procedure

This model was trained with SFT.

Framework versions

• TRL: 0.25.1
• Transformers: 4.57.1
• Pytorch: 2.9.1
• Datasets: 4.4.1
• Tokenizers: 0.22.1

Citations

Cite TRL as:

@misc{vonwerra2022trl,
    title        = {{TRL: Transformer Reinforcement Learning}},
    author       = {Leandro von Werra and Younes Belkada and Lewis Tunstall and Edward Beeching and Tristan Thrush and Nathan Lambert and Shengyi Huang and Kashif Rasul and Quentin Gallou{\'e}dec},
    year         = 2020,
    journal      = {GitHub repository},
    publisher    = {GitHub},
    howpublished = {\url{https://github.com/huggingface/trl}}
}