Bittensor Model Fine-tuning

Model Fine-tuning script for Bittensor miners using the dataset generated by validators

Example wandb link generated by one of the opentensor team validators

Example tuned-model uploaded to HuggingFace

Overview

• dataset generated by validators: this will be a link to a wandb run where they can pull the pandas DF
• models are used by miners
• miners train model (any model, like a base llama for instance) on the dataset. They use their own compute (i.e. mining machine with GPU capabilities)
• Once trained, the model can be used as the base of a miner on bittensor.

Models Supported

• GPT4All