nomic-ai/colnomic-embed-multimodal-7b
Adapter vs Qwen/Qwen2.5-VL-7B-Instruct. nomic-ai/colnomic-embed-multimodal-7b is a LoRA adapter, not full weights: rank 32, on custom_text_proj, down_proj, gate_proj, k_proj, o_proj, q_proj, up_proj, v_proj in 28 layers (197 adapted matrices). It modifies Qwen/Qwen2.5-VL-7B-Instruct at load time.
3 other repos hold byte-for-byte identical weight files.