Text Generation
Transformers
PyTorch
llama
Inference Endpoints
text-generation-inference

PuddleJumper is one of my favorite models. Plans for 30B version?

#3
by ttkciar - opened

This model's output quality is absolutely amazing for a 13B, and it has become my go-to for general-purpose inference. Fantastic job!

I have written more about it and how it compares to some other models here: https://old.reddit.com/r/LocalLLaMA/comments/16eeyl7/what_models_are_you_running/jzuzy0t/

Do you have any plans for creating a 30B version? The 13B version is already comparable to Vicuna-33B, and I would expect a 30B to blow Vicuna out of the water.

Thanks for bringing us this wonderful model :-)

So this model is based on EverythingLM v2, EverythingLM v3 is in the works. Once I train a 33b v3 model, I can create puddlejumper for that size.

Sign up or log in to comment