sail/Sailor2-8B
Full fine-tune vs Qwen/Qwen2.5-7B. Fine-tune: 100% of the weights changed, across 28 of 28 layers.
Tensors changed: Norms 57/57, MLP 84/84, Attention 196/196
6 derivatives checked
- sail/Sailor2-8B-Chat: Full fine-tune
- mradermacher/Sailor2-8B-GGUF: Same weights, converted
- sail/Sailor2-3B: Different architecture
- sail/Sailor2-8B-SFT: Full fine-tune
- sail/Sailor2-L-8B: Full fine-tune
- BlossomsAI/BloomVN-8B-instruct: Full fine-tune