Replies: 1 comment
-
For training models, I usually use 4 bit GPTQ models downloaded from TheBloke. I load them with the Transformers model loader, with 'auto-devices' and 'disable_exllama' ticked, and then train them with the Training tab. Then to use the resulting LoRA, I apply them to the same model but loaded with the ExLlama_v2 loader. |
Beta Was this translation helpful? Give feedback.
0 replies
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
-
Can anyone tell me what model and setup they've successfully used to train that model? The combinations I've tried so far have all failed.
Thanks!
Beta Was this translation helpful? Give feedback.
All reactions