Replies: 2 comments 4 replies
-
The learning ability of QLoRA is limited, increasing lora rank may not work, consider use ReLoRA or init multiple adapters |
Beta Was this translation helpful? Give feedback.
0 replies
-
Dataset is not pretraining but QA-sft and Is QLoRA + ReLoRA possible? if not, propably OOM 😅 |
Beta Was this translation helpful? Give feedback.
4 replies
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
-
In QLoRA Training with large datasets~300GB, and given dataset i use
--lora_rank 256 --lora_alpha 512
The log
loss
ranges from1.0
to1.5
. Is that reasonable LR + arg for the datasets?Beta Was this translation helpful? Give feedback.
All reactions