NiuTrans/GRAM-LLaMA3.2-3B-RewardModel
Full fine-tune vs meta-llama/Llama-3.2-3B-Instruct. Fine-tune: 100% of the weights changed, across 28 of 28 layers.
Tensors changed: Embeddings 1/1, Norms 54/57, MLP 84/84, Attention 112/112
3 derivatives checked
- mradermacher/GRAM-LLaMA3.2-3B-RewardModel-i1-GGUF: Quantization
- prithivMLmods/GRAM-LLaMA3.2-3B-RewardModel-GGUF: Same weights, converted
- mradermacher/GRAM-LLaMA3.2-3B-RewardModel-GGUF: Same weights, converted