qingy2024/Formatter-1.5B
Full fine-tune vs unsloth/Qwen2.5-1.5B. Fine-tune: 100% of the weights changed, across 28 of 28 layers.
Tensors changed: Norms 54/57, MLP 84/84, Attention 196/196
1 derivatives checked
- mradermacher/Formatter-1.5B-GGUF: Same weights, converted