mergekit-community/CodeMix-JPID-3B-Llama3.2

Full fine-tune vs AXCXEPT/EZO-Llama-3.2-3B-Instruct-dpoE. Fine-tune: 100% of the weights changed, across 28 of 28 layers.

Tensors changed: Embeddings 1/1, Norms 57/57, MLP 84/84, Attention 112/112

2 derivatives checked

• mradermacher/CodeMix-JPID-3B-Llama3.2-i1-GGUF: Quantization
• mradermacher/CodeMix-JPID-3B-Llama3.2-GGUF: Same weights, converted

On the Hub · Downloads on Model Pulse