phxember/Uni-MuMER-Qwen3-VL-2B
Full fine-tune vs Qwen/Qwen3-VL-2B-Instruct. Fine-tune: 100% of the weights changed, across 28 of 28 layers.
Tensors changed: Embeddings 4/4, Norms 183/217, MLP 196/196, Attention 208/208
3 derivatives checked
- mradermacher/Uni-MuMER-Qwen3-VL-2B-i1-GGUF: Quantization (fewer layers)
- mradermacher/Uni-MuMER-Qwen3-VL-2B-GGUF: Same weights, converted (fewer layers)
- cstr/uni-mumer-qwen3-vl-2b-GGUF: Not comparable yet