FireRedTeam/FireRed-OCR
Partial fine-tune vs Qwen/Qwen3-VL-2B-Instruct. Partial change: 86% of the weights changed (attention, embeddings, mlp, norms), in 28 of 28 layers.
Tensors changed: Embeddings 1/4, Norms 116/217, MLP 100/196, Attention 112/208
2 derivatives checked
- mradermacher/FireRed-OCR-GGUF: Same weights, converted (fewer layers)
- prithivMLmods/FireRed-OCR-GGUF: Same weights, converted (fewer layers)
1 other repos hold byte-for-byte identical weight files.