mergekit-community/Holgerim-Llama-7b
Full fine-tune vs dominguesm/Canarim-7B-Instruct. Fine-tune: 97% of the weights changed, across 31 of 31 layers.
Tensors changed: Embeddings 1/1, Norms 63/63, MLP 90/93, Attention 120/124, lm_head 1/1
2 derivatives checked
- mradermacher/Holgerim-Llama-7b-i1-GGUF: Quantization
- mradermacher/Holgerim-Llama-7b-GGUF: Same weights, converted