Qwen/Qwen3-VL-Embedding-2B
LoRA-style vs Qwen/Qwen3-VL-2B-Instruct. Looks like a merged LoRA (or a fine-tune with frozen norms): only the weight matrices of down_proj, gate_proj, k_proj, q_proj, up_proj, v_proj changed, in 28 of 28 layers; norms, biases and embeddings are untouched.
Tensors changed: Embeddings 0/4, Norms 0/217, MLP 84/196, Attention 84/208
17 derivatives checked
- shigureui/Qwen3-VL-Embedding-2B-FP8: Quantization
- DevQuasar/Qwen.Qwen3-VL-Embedding-2B-GGUF: Same weights, converted (fewer layers)
- LifetimeMistake/Qwen3-VL-Embedding-2B-AWQ-4bit: Quantization
- mradermacher/Qwen3-VL-Embedding-2B-GGUF: Same weights, converted (fewer layers)
- mradermacher/Qwen3-VL-Embedding-2B-i1-GGUF: Quantization (fewer layers)
- dnotitia/DNA-VL-STEER-2B: Partial fine-tune
- alexliap/Qwen3-VL-Embedding-2B-FP8-DYNAMIC: Quantization
- LifetimeMistake/Qwen3-VL-Embedding-2B-NVFP4: Quantization
- EximiusLabs/fusion-embedding-1-2b-preview: Different architecture
- Rizwan313/Qwen3-VL-Embedding-2B-GGUF: Same weights, converted (fewer layers)
- alpaim/Qwen3-VL-Embedding-2B-GGUF-vecBox: Same weights, converted (fewer layers)
- TitleOS/Qwen3-VL-Embedding-2B-GGUF: Not comparable yet
- whybe-choi/kovre: Partial fine-tune
- EximiusLabs/fusion-embedding-2-2b-preview: Different architecture
- batiai/Qwen3-VL-Embedding-2B-GGUF: Quantization (fewer layers)
- tomaarsen/Qwen3-VL-Embedding-2B-vdr: Partial fine-tune
- jfiekdjdk/Qwen3-VL-Embedding-2B-Q8_0-GGUF: Quantization (fewer layers)
4 other repos hold byte-for-byte identical weight files.