Version with permissive datasets / commercial use
#15
by
michaelfeil
- opened
It would be great to have a version which grants commercial use, or use in for-profit organisations.
I understand this is only because of the use of alpaca dataset, hence it would be great to run this without such datasets included.
Actually, all datasets used to train this except for Anthropic's HH dataset are ChatGPT-distilled and hence only suitable for research purposes. Our Instruct model is permissively licensed, if you have a chat dataset and would like to start from there.
sam-mosaic
changed discussion status to
closed