Dolly Replication

This is a replication of the "dolly" model from Databricks, which is GPT-J-6B fine-tuned on the alpaca instruction following dataset. It was trained in 6 hours on a high-end workstation (2xA6000 GPUs).

Licensing

Because the alpaca dataset was created from text generated by OpenAI's text-davinci-003 model, it is not clear whether this model can be used commercially; OpenAI's ToS does not allow one to "use output from the Services to develop models that compete with OpenAI".

datasets:

tatsu-lab/alpaca