Is Instruct 4B-Width also going to be published?
#1
by
Qubitium
- opened
4B is a game changer if the quality is as shown in the papers. Hoping that the instruct model is also released soon. Thank you!
It's not a game changer.
The model may not be but the distillation technique may be if the model evals hold up.
there is their own instruct fine tune https://huggingface.co/nvidia/Llama-3.1-Nemotron-Nano-4B-v1.1 it's trained on different data and has reasoning support