Alibaba-NLP/gme-Qwen2-VL-2B-Instruct
Partial fine-tune vs Qwen/Qwen2-VL-2B-Instruct. Partial change: 71% of the weights changed (attention, embeddings, mlp, norms), in 28 of 32 layers.
Tensors changed: Embeddings 1/2, Norms 56/187, MLP 88/216, Attention 196/324
5 derivatives checked
- mradermacher/gme-Qwen2-VL-2B-Instruct-i1-GGUF: Quantization (fewer layers)
- sinequa/gme-Qwen2-VL-2B-Instruct-GGUF: Same weights, converted (fewer layers)
- mradermacher/gme-Qwen2-VL-2B-Instruct-GGUF: Same weights, converted (fewer layers)
- DevQuasar/Alibaba-NLP.gme-Qwen2-VL-2B-Instruct-GGUF: Same weights, converted (fewer layers)
- RKNNAI/RK182X-EMBEDDING-gme-Qwen2-VL-2B-Instruct: Quantization
1 other repos hold byte-for-byte identical weight files.