Training recipe
Would you share your training script? What libraries did you use exactly, the script that you used?
This would be super useful for other folks (like me) to experiment with tiny LLMs like the one you're building (and I am building), basically we could learn from each other, if you discover something that doesn't work it saves time for everybody and same for me.
My training code: https://github.com/croqaz/vintage-LLM/tree/main/training
So, would you share your code, it could be in HuggingFace model repo, just a new folder called "training" or whatever and share the code?
Cheers! And good luck with Rose Pro!
I appreciate mate
But sharing the code is not , I m comfortable with at the moment
But I may share codes in further future for sure, right now My own pipeline is bit complicated
Secondly I use to code on jyupter notebook , habbit back from kaggle days.
The Message
I m a self taught guy , I strongly beilve one should learn by experimenting -failing-improving
Its the best part of the journey.
If you have discord we can talk there "godelev"
And I can help you as much as I can
Hope you understand
Thank You !!