đź’ˇ
The emergence of the Free Willy (Orca) models, particularly Free Willy 2, released by Stability AI, presents a promising advancement in open-source language models that outperform the Llama 2 models in some benchmarks. These models, fine-tuned using the Orca approach proposed in the original Orca paper, demonstrate exceptional performance due to their carefully curated data sets. However, despite their open-source nature, they cannot be used for commercial purposes. The comparison between Free Willy 2 and Free Willy 1 shows that data set quality plays a vital role in model performance, highlighting the need for attention to training data when fine-tuning
