WWTCyberLab/abliterated-llama-8b
LoRA-style vs meta-llama/Llama-3.1-8B-Instruct. Looks like a merged LoRA (or a fine-tune with frozen norms): only the weight matrices of down_proj, gate_proj, o_proj, up_proj changed, in 17 of 32 layers; norms, biases and embeddings are untouched.
Tensors changed: Embeddings 0/1, Norms 0/65, MLP 51/96, Attention 17/128, lm_head 0/1
2 derivatives checked
- mradermacher/abliterated-llama-8b-i1-GGUF: Quantization
- mradermacher/abliterated-llama-8b-GGUF: Same weights, converted