taresco/KarantaOCR
Full fine-tune vs Qwen/Qwen2.5-VL-3B-Instruct. Fine-tune: 100% of the weights changed, across 36 of 36 layers.
Tensors changed: Embeddings 2/2, Norms 79/138, MLP 283/304, Attention 380/380
2 derivatives checked
- mradermacher/KarantaOCR-i1-GGUF: Quantization (fewer layers)
- mradermacher/KarantaOCR-GGUF: Same weights, converted (fewer layers)