lightblue/suzume-llama-3-8B-multilingual
Full fine-tune vs meta-llama/Meta-Llama-3-8B-Instruct. Fine-tune: 93% of the weights changed, across 32 of 32 layers.
Tensors changed: Embeddings 0/1, Norms 22/65, MLP 96/96, Attention 128/128, lm_head 1/1
8 derivatives checked
- lightblue/suzume-llama-3-8B-multilingual-orpo-borda-top25: Full fine-tune
- lightblue/suzume-llama-3-8B-multilingual-orpo-borda-full: Full fine-tune
- lightblue/suzume-llama-3-8B-multilingual-orpo-borda-half: Full fine-tune
- lightblue/suzume-llama-3-8B-multilingual-orpo-borda-top75: Full fine-tune
- QuantFactory/suzume-llama-3-8B-multilingual-GGUF: Quantization
- mradermacher/suzume-llama-3-8B-multilingual-i1-GGUF: Quantization
- mradermacher/suzume-llama-3-8B-multilingual-GGUF: Same weights, converted
- AlekseyElygin/suzume-llama-3-8B-multilingual: Full fine-tune
1 other repos hold byte-for-byte identical weight files.