Adun/olmOCR-7B-thai-v3.2
LoRA-style vs allenai/olmOCR-7B-0225-preview. Looks like a merged LoRA (or a fine-tune with frozen norms): only the weight matrices of 0, 2, down_proj, gate_proj, k_proj, o_proj, q_proj, up_proj, v_proj changed, in 28 of 32 layers; norms, biases and embeddings are untouched.
Tensors changed: Embeddings 0/2, Norms 0/187, MLP 86/216, Attention 112/324, lm_head 0/1
2 derivatives checked
- mradermacher/olmOCR-7B-thai-v3.2-i1-GGUF: Quantization (fewer layers)
- mradermacher/olmOCR-7B-thai-v3.2-GGUF: Same weights, converted (fewer layers)