freeai-org/Scalpel-VL-1.8B

Partial fine-tune vs Qwen/Qwen3-VL-2B-Instruct. Partial change: 60% of the weights changed (attention, mlp, norms), in 21 of 24 layers.

Tensors changed: Embeddings 0/4, Norms 56/189, MLP 63/175, Attention 84/180

1 derivatives checked

• mradermacher/Scalpel-VL-1.7B-GGUF: Same weights, converted (fewer layers)

On the Hub · Downloads on Model Pulse