lightblue/suzume-llama-3-8B-japanese
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 15/65, MLP 96/96, Attention 128/128, lm_head 1/1
10 derivatives checked
- mradermacher/suzume-llama-3-8B-japanese-i1-GGUF: Quantization
- mradermacher/suzume-llama-3-8B-japanese-GGUF: Same weights, converted
- noirchan/Llama-3-8B_Suzume_TIES: Partial fine-tune
- noirchan/Llama-3-8B_Suzume_Task_Arithmetic: Partial fine-tune
- noirchan/Llama-3-8B-suzume_liner0.5: Full fine-tune
- noirchan/Llama-3-8B_Suzume_DARE_TIES0.5: Full fine-tune
- noirchan/Llama-3-8B-suzume_liner0.8: Full fine-tune
- noirchan/Llama-3-8B_Suzume_DARE: Partial fine-tune
- noirchan/Llama-3-8B_Suzume_DARE_TIES0.5_v2: Full fine-tune
- noirchan/Llama-3-8B_Suzume_TIES0.5: Partial fine-tune