2 ms·
"Fine-tuning even Lora on the open source models is nearly always better than these other approaches" Can you expand on that? I've not seen evidence of that my
by simonw 3y ago
"Fine-tuning even Lora on the open source models is nearly always better than these other approaches"
Can you expand on that? I've not seen evidence of that myself yet, but maybe I haven't looked in the right places.
- a_ml 3y agoFrom Aug 2022, but still worth a look "Few-Shot Parameter-Efficient Fine-Tuning is Better and Cheaper than In-Context Learning" https://arxiv.org/abs/2205.05638 https://arxiv.org/abs/2205.05638