grimjim/Llama-3-Instruct-8B-SimPO-SPPO-Iter3-merge
Full fine-tune vs UCLA-AGI/Llama-3-Instruct-8B-SPPO-Iter3. Fine-tune: 91% of the weights changed, across 32 of 32 layers.
Tensors changed: Embeddings 0/1, Norms 15/65, MLP 93/96, Attention 124/128, lm_head 1/1
3 derivatives checked
- featherless-ai-quants/grimjim-Llama-3-Instruct-8B-SimPO-SPPO-Iter3-merge-GGUF: Quantization
- mradermacher/Llama-3-Instruct-8B-SimPO-SPPO-Iter3-merge-i1-GGUF: Quantization
- mradermacher/Llama-3-Instruct-8B-SimPO-SPPO-Iter3-merge-GGUF: Same weights, converted