mergekit-community/L3.1-Boshima-b
Partial fine-tune vs OwenArli/Llama-3.1-8B-ArliAI-Formax-v1.0. Partial change: 80% of the weights changed (attention, embeddings, lm_head, mlp, norms), in 32 of 32 layers.
Tensors changed: Embeddings 1/1, Norms 65/65, MLP 72/96, Attention 112/128, lm_head 1/1
2 derivatives checked
- mradermacher/L3.1-Boshima-b-i1-GGUF: Quantization
- mradermacher/L3.1-Boshima-b-GGUF: Same weights, converted