Guanaco vs. Orca

Overview

Guanaco is an LLM based off the QLoRA 4-bit finetuning method developed by Tim Dettmers et. al. in the UW NLP group. Guanaco achieves 99% ChatGPT performance on the Vicuna benchmark.

Guanaco is an LLM that uses a finetuning method called LoRA that was developed by Tim Dettmers et. al. in the UW NLP group. With QLoRA, it becomes possible to finetune up to a 65B parameter model on a 48GB GPU without loss of performance relative to a 16-bit model. The Guanaco model family outperforms all previously released models on the Vicuna benchmark. However, given the models are based off of the LLaMA model family, commercial use is not permitted.

Initial release: 2023-05-23

Reference

https://github.com/artidoro/qlora

Further Reading

• https://arxiv.org/abs/2305.14314
• https://twitter.com/Tim_Dettmers/status/1661379354507476994

Overview

Orca is a descendant of LLaMA developed by Microsoft with finetuning on explanation traces obtained from GPT-4.

Orca-13B is a LLM developed by Microsoft. It is based on LLaMA with finetuning on complex explanation traces obtained from GPT-4. By using rich signals, Orca surpasses the performance of models such as Vicuna-13B on complex tasks. However, given its model backbone and the data used for its finetuning, Orca is under noncommercial use.

Initial release: 2023-06-05

Reference

https://aka.ms/orca-lm

Further Reading

• https://arxiv.org/pdf/2306.02707.pdf