This is a fork of gpt-llm-trainer.
I have edited the prompt in order to create a dataset to fine-tune an llm like Llama 2 to return arguments to a described function in the prompt in a json readable format, similar to OpenAI's Function Calling . You can find the created dataset in the functioncalling_gpt_dataset.pkl file
You can simply read the pickle file, and use the provided notebook to fine-tune your llm of choice on the dataset.