EzekielBlaze/Nia-Qwen3-VL-32B-Thinking-eos-fix
LoRA-style vs Qwen/Qwen3-VL-32B-Thinking. Looks like a merged LoRA (or a fine-tune with frozen norms): only the weight matrices of down_proj, gate_proj, k_proj, o_proj, q_proj, up_proj, v_proj changed, in 64 of 64 layers; norms, biases and embeddings are untouched.
Tensors changed: Embeddings 0/4, Norms 0/373, MLP 192/316, Attention 256/364, lm_head 0/1
2 derivatives checked
- mradermacher/Nia-Qwen3-VL-32B-Thinking-eos-fix-i1-GGUF: Quantization (fewer layers)
- mradermacher/Nia-Qwen3-VL-32B-Thinking-eos-fix-GGUF: Quantization (fewer layers)