zaddyzaddy/soro
Partial fine-tune vs unsloth/llama-3-8b-bnb-4bit. Partial change: 9% of the weights changed (attention, mlp), in 32 of 32 layers.
Tensors changed: MLP 224/499, Attention 283/661, lm_head 0/1, Embeddings 0/1, Norms 0/65
Partial fine-tune vs unsloth/llama-3-8b-bnb-4bit. Partial change: 9% of the weights changed (attention, mlp), in 32 of 32 layers.
Tensors changed: MLP 224/499, Attention 283/661, lm_head 0/1, Embeddings 0/1, Norms 0/65