NousResearch/Hermes-3-Llama-3.1-70B
Full fine-tune vs meta-llama/Llama-3.1-70B. Fine-tune: 99% of the weights changed, across 80 of 80 layers.
Tensors changed: Embeddings 0/1, Norms 160/161, MLP 240/240, Attention 320/320, lm_head 1/1
18 derivatives checked
- bartowski/Hermes-3-Llama-3.1-70B-GGUF: Quantization
- legraphista/Hermes-3-Llama-3.1-70B-IMat-GGUF: Same weights, converted
- mradermacher/Hermes-3-Llama-3.1-70B-i1-GGUF: Quantization
- backyardai/Hermes-3-Llama-3.1-70B-GGUF: Quantization
- mradermacher/Hermes-3-Llama-3.1-70B-GGUF: Quantization
- mbley/NousResearch-Hermes-3-Llama-3.1-70B-AWQ: Quantization
- featherless-ai-quants/NousResearch-Hermes-3-Llama-3.1-70B-GGUF: Quantization
- DevQuasar-4/Hermes-2-Theta-Llama-3-70B-GGUF: Full fine-tune
- Guilherme34/Hermes-3-Llama-3.1-70B-Uncensored: LoRA-style
- Saxo/Linkbricks-Horizon-AI-Korean-Advanced-70B: LoRA-style
- gbueno86/Brinebreath-Llama-3.1-70B: Full fine-tune
- mlabonne/Hermes-3-Llama-3.1-70B-lorablated: Partial fine-tune
- olaverse/MIST-1-70B: Full fine-tune
- mergesloppa123123/Hermyale-stack-90B: Partial fine-tune
- Fireheart92/vire-protocol-70b: LoRA-style
- harkov000/Hermes-Euryale-70B: Partial fine-tune
- mergesloppa123123/Hanames-90B-L3.1: Full fine-tune
- Nexesenex/Llama_3.x_70b_SmarTricks_v1.55_MS: Full fine-tune
3 other repos hold byte-for-byte identical weight files.