rokotmidi/rokot-omr-2b

Full fine-tune vs Qwen/Qwen3-VL-2B-Instruct. Fine-tune: 100% of the weights changed, across 28 of 28 layers.

Tensors changed: Embeddings 4/4, Norms 137/217, MLP 181/196, Attention 208/208

1 derivatives checked

• mradermacher/rokot-omr-2b-GGUF: Full fine-tune (fewer layers)

On the Hub · Downloads on Model Pulse