OpenGVLab/VisualPRM-8B-v1_1

Full fine-tune vs OpenGVLab/InternVL2_5-8B. Fine-tune: 100% of the weights changed, across 24 of 24 layers.

Tensors changed: Embeddings 4/4, Attention 96/96, other 48/48, MLP 98/98, Norms 96/96

2 derivatives checked

• mradermacher/VisualPRM-8B-v1_1-i1-GGUF: Quantization (fewer layers)
• mradermacher/VisualPRM-8B-v1_1-GGUF: Same weights, converted (fewer layers)

On the Hub · Downloads on Model Pulse