win10/Qwen2.5-2B-Instruct
Partial fine-tune (fewer layers) vs Qwen/Qwen2.5-1.5B-Instruct. Qwen/Qwen2.5-1.5B-Instruct matches part of win10/Qwen2.5-2B-Instruct (338 of its 627 tensors; the rest of win10/Qwen2.5-2B-Instruct has no counterpart). On the shared part: Partial change: 79% of the weights changed (attention, mlp, norms), in 26 of 28 layers.
Tensors changed: Embeddings 0/1, Norms 52/57, MLP 78/84, Attention 182/196
1 derivatives checked
- mradermacher/Qwen2.5-2B-Instruct-GGUF: Same weights, converted