Deployment of the official Google Gemma 4 model for this repository's banking function-calling server.
This notebook loads the official source model from Hugging Face and republishes the saved artifacts. It does not train, fine-tune, evaluate, or mutate the model weights, so no banking-specific fine-tuning is applied here.
Intended Use
Use this model with Gemma 4 chat templates and pass the banking tool schemas to processor.apply_chat_template(..., tools=...).
For this repository's local REST server, set:
GEMMA4_MODEL_ID=phattrandeveloper/gemma-4-bank-function-calling uv run uvicorn src.gemma4_server:app --host 127.0.0.1 --port 8001
Source
Base model: google/gemma-4-E2B-it
Source Hugging Face model: google/gemma-4-E2B-it
Source revision: main
Uploaded from snapshot folder: /Users/sonphat.tran/Desktop/Project/fine-tuning-tool-calling/artifacts/huggingface/gemma-4-bank-function-calling/source_model
Treat every generated tool call as a proposal. Application code must validate tool names, arguments, account ownership, limits, recipient identity, and user confirmation before executing any banking action.
