ianua/OvisOCR2

Full fine-tune vs Qwen/Qwen3.5-0.8B. Fine-tune: 100% of the weights changed, across 24 of 24 layers.

Tensors changed: Attention 213/216, Norms 127/129, Embeddings 4/4, MLP 124/124

1 derivatives checked

• mradermacher/OvisOCR2-GGUF: Same weights, converted

5 other repos hold byte-for-byte identical weight files.

On the Hub · Downloads on Model Pulse