prithivMLmods/proxima-ocr-d.markdown-post3.0.l
Full fine-tune vs Qwen/Qwen3-VL-8B-Instruct. Fine-tune: 93% of the weights changed, across 36 of 36 layers.
Tensors changed: Embeddings 3/4, Norms 241/261, MLP 231/232, Attention 252/252, lm_head 1/1
3 derivatives checked
- mradermacher/proxima-ocr-d.markdown-post3.0.l-i1-GGUF: Quantization (fewer layers)
- mradermacher/proxima-ocr-d.markdown-post3.0.l-GGUF: Same weights, converted (fewer layers)
- prithivMLmods/proxima-ocr-d.markdown-post3.0.l-GGUF: Same weights, converted (fewer layers)