The question of training the model.
#2
by
Kotokin
- opened
Hi! Thank you for your work, I just tried version 1.3, it thrills me! I have a question, do you have the opportunity to finetune llama 3.2 1b or 3b? They would be ideal for speculative decoding models, requiring less resources than 8b.
Thanks! That is a good case for training the smaller models. I might go and do that at some point.
Kotokin
changed discussion status to
closed