ZejunLi/AdaVaR-3B

Full fine-tune vs Qwen/Qwen2.5-VL-3B-Instruct. Fine-tune: 100% of the weights changed, across 36 of 36 layers.

Tensors changed: Embeddings 2/2, Norms 75/138, MLP 275/304, Attention 380/380

2 derivatives checked

• mradermacher/AdaVaR-3B-GGUF: Same weights, converted (fewer layers)
• mradermacher/AdaVaR-3B-i1-GGUF: Quantization (fewer layers)

On the Hub · Downloads on Model Pulse