Robertp423/Qwen3-VL-2B-Sanctum
Partial fine-tune vs unsloth/Qwen3-VL-2B-Instruct. Partial change: 81% of the weights changed (attention, embeddings, mlp, norms), in 28 of 28 layers.
Tensors changed: Embeddings 1/4, Norms 91/217, MLP 84/196, Attention 112/208
1 derivatives checked
- mradermacher/Qwen3-VL-2B-Sanctum-GGUF: Same weights, converted (fewer layers)