aloobun/Meta-Llama-3-7B-28Layers

Partial fine-tune vs meta-llama/Meta-Llama-3-8B. Partial change: 15% of the weights changed (attention, mlp, norms), in 5 of 28 layers.

Tensors changed: Embeddings 0/1, Norms 10/57, MLP 15/84, Attention 20/112, lm_head 0/1

1 derivatives checked

• mradermacher/Meta-Llama-3-7B-28Layers-GGUF: Same weights, converted

On the Hub · Downloads on Model Pulse