elinas/Llama-3-15B-Instruct-ft-v2
Full fine-tune vs elinas/Llama-3-15B-Instruct-zeroed. Fine-tune: 94% of the weights changed, across 64 of 64 layers.
Tensors changed: lm_head 1/1, Embeddings 0/1, Norms 0/129, MLP 188/192, Attention 249/256
3 derivatives checked
- mradermacher/Llama-3-15B-Instruct-ft-v2-i1-GGUF: Quantization
- featherless-ai-quants/elinas-Llama-3-15B-Instruct-ft-v2-GGUF: Quantization
- mradermacher/Llama-3-15B-Instruct-ft-v2-GGUF: Quantization