ArkAiLab-Adl/nexora-ocr-v0.1-2b
LoRA-style vs Qwen/Qwen3.5-2B. Looks like a merged LoRA (or a fine-tune with frozen norms): only the weight matrices of down_proj, gate_proj, in_proj_a, in_proj_b, in_proj_qkv, in_proj_z, k_proj, linear_fc1, linear_fc2, o_proj, out_proj, proj, q_proj, qkv, up_proj, v_proj changed, in 24 of 24 layers; norms, biases and embeddings are untouched.
Tensors changed: Attention 162/268, Norms 0/183, Embeddings 0/5, MLP 120/175, other 0/1
2 derivatives checked
- mradermacher/nexora-ocr-v0.1-2b-i1-GGUF: Quantization
- mradermacher/nexora-ocr-v0.1-2b-GGUF: Same weights, converted