prithivMLmods/Qwen3-VisionCaption-2B
LoRA-style vs prithivMLmods/Qwen3-VL-2B-Instruct-abliterated-v1. Looks like a merged LoRA (or a fine-tune with frozen norms): only the weight matrices of down_proj, gate_proj, k_proj, o_proj, q_proj, up_proj, v_proj changed, in 28 of 28 layers; norms, biases and embeddings are untouched.
Tensors changed: Attention 112/208, MLP 84/196, Norms 0/217, Embeddings 0/4
3 derivatives checked
- prithivMLmods/Qwen3-VisionCaption-2B-GGUF: Same weights, converted (fewer layers)
- mradermacher/Qwen3-VisionCaption-2B-i1-GGUF: Quantization (fewer layers)
- mradermacher/Qwen3-VisionCaption-2B-GGUF: Same weights, converted (fewer layers)