suv11235/olmOCR-7B-grpo-v2
Partial fine-tune vs allenai/olmOCR-2-7B-1025. Partial change: 80% of the weights changed (attention, embeddings, mlp, norms), in 32 of 32 layers.
Tensors changed: Embeddings 1/2, Norms 29/122, MLP 238/280, Attention 276/324, lm_head 0/1
2 derivatives checked
- mradermacher/olmOCR-7B-grpo-v2-i1-GGUF: Quantization (fewer layers)
- mradermacher/olmOCR-7B-grpo-v2-GGUF: Same weights, converted (fewer layers)