Rookied/qwen2-mini
Full fine-tune vs Qwen/Qwen2-1.5B-Instruct. Fine-tune: 100% of the weights changed, across 28 of 28 layers.
Tensors changed: Embeddings 1/1, Norms 34/57, MLP 84/84, Attention 191/196
1 derivatives checked
- mradermacher/qwen2-mini-GGUF: Same weights, converted