auto training colab for llama

#6
by legolasyiu - opened

Hello team,

I'm excited to share that I've been working on a very fun project for future automation AI. An AI will be trained automatically after receiving some prompts, then upload the model, and then returns to chat mode.

I'm currently using Colab because it utilizes the T4 Nvidia. I've developed an auto-training process utilizing Unsloth for any Llama models, which involves:
Loading Huggingface models
Entering chat mode, auto-training after receiving some prompts, and then uploading the models

What are Meta's plans for utilizing this auto-training process? Where in the Meta stack does Meta plan to utilize this? Please let me know if you have any further questions? Thanks

Sincerely,
Thomas

any update? interest? I am not joking. I can provide colab to show it!

Sign up or log in to comment