You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
I notice that you mentioned "lora support is only for simple lora with only q_proj and v_proj", and I wonder whether I can apply lora with other modules like k_proj、o_proj、donw_proj and so on.
It seems the implementation is not only designed for “q_proj and v_proj“, so why lora support is only for simple lora with only q_proj and v_proj ?
The text was updated successfully, but these errors were encountered:
You can add more modules in training scripts if you want add something else like k_proj
"lora support is only for simple lora with only q_proj and v_proj" is about the optimization for inference in this repo, but is already very slow now. It is recommended to use this repo for training and exllamav2 (fastest for now) for inference.
Thanks for your awesome repo!
I notice that you mentioned "lora support is only for simple lora with only q_proj and v_proj", and I wonder whether I can apply lora with other modules like k_proj、o_proj、donw_proj and so on.
It seems the implementation is not only designed for “q_proj and v_proj“, so why lora support is only for simple lora with only q_proj and v_proj ?
The text was updated successfully, but these errors were encountered: