mrinaldi86/llama_tooling

Dataset Card for Dataset Name

This dataset was generated starting from marclove/llama_functions

I have replaced all functions call with the tooling prompt which is used in crew ai

This dataset is still in test. I'll update with more results when I'll use it for fine tuning and do some tests

This dataset use the LLAMA format [INST] [/INST]

As the author stated:

the original dataset is a synthetic dataset generated from a mix of manual curation of OpenAPI endpoints and prompting of OpenAI models. It is further mixed with chat completions from the Guanaco subset of the OASST1 chat dialogue dataset. It is a total of 18,000 rows, 9,000 rows from the synthetic dataset of function calls and 9,000 rows from the Guanaco dataset.

The dataset is mixed with Guanaco in order to maintain accuracy and helpfulness when calling a function is not the appropriate response. I plan to remove the Guanaco portion of the dataset and instead provide fine-tuning recommendations, guidelines for use, more detailed information regarding limitations, and eval stats of 7B, 13B, and 70B models.