yyyllll/VecLang-4B
LoRA-style vs Qwen/Qwen3-VL-4B-Instruct. Looks like a merged LoRA (or a fine-tune with frozen norms): only the weight matrices of down_proj, gate_proj, k_proj, linear_fc1, linear_fc2, o_proj, proj, q_proj, qkv, up_proj, v_proj changed, in 36 of 36 layers; norms, biases and embeddings are untouched.
Tensors changed: Embeddings 0/4, Norms 0/249, MLP 164/220, Attention 192/240
1 derivatives checked
- mradermacher/VecLang-4B-GGUF: Same weights, converted (fewer layers)