Will Consider Continue SFT?

by kevinpro - opened Jul 19

Jul 19

Thank you for your awesome work! And I am surprised about the data and the performance of LLM tuned on these data.

I notice that you guys prefer to train the model from the base model and achieve the comparable performance with the official released model.
Will you consider continuing SFT on the official released Chat Model, and will it bring further promising performance?

kevinpro changed discussion status to closed Jul 19

kevinpro changed discussion status to open Jul 19

hyxmmm

Beijing Academy of Artificial Intelligence org Aug 30

Apologies for my late reply. We will of course continue to post new versions of the SFT data, as well as fine-tuned versions of the new model on Infinity-Instruct. You can keep an eye on our dataset project page (https://huggingface.co/datasets/BAAI/Infinity-Instruct) for our new releases. Thanks

Upload images, audio, and videos by dragging in the text input, pasting, or clicking here.

Tap or paste here to upload images

· Sign up or log in to comment