noirchan/Llama-3-8B_Suzume_DARE
Partial fine-tune vs lightblue/suzume-llama-3-8B-japanese. Partial change: 32% of the weights changed (attention, mlp, norms), in 32 of 32 layers.
Tensors changed: Embeddings 0/1, Norms 9/65, MLP 33/96, Attention 53/128, lm_head 0/1
1 derivatives checked
- mradermacher/Llama-3-8B_Suzume_DARE-GGUF: Same weights, converted