Aronuihyig/IC-WaveDrom-DSL
LoRA-style vs PaddlePaddle/PaddleOCR-VL. Looks like a merged LoRA (or a fine-tune with frozen norms): only the weight matrices of down_proj, fc1, fc2, gate_proj, k_proj, linear_1, linear_2, o_proj, out_proj, q_proj, up_proj, v_proj changed, in 27 of 27 layers; norms, biases and embeddings are untouched.
Tensors changed: lm_head 0/1, MLP 109/170, Norms 0/151, Embeddings 0/5, Attention 180/292, other 0/1
1 derivatives checked
- mradermacher/IC-WaveDrom-DSL-GGUF: LoRA-style (fewer layers)