DeepAuto-AI/Explore_Llama-3.1-8B-Inst

Partial fine-tune vs meta-llama/Llama-3.1-8B-Instruct. Partial change: 2% of the weights changed (mlp, norms), in 32 of 32 layers.

Tensors changed: Embeddings 0/1, Norms 65/65, MLP 3/96, Attention 0/128, lm_head 0/1

3 derivatives checked