Hyukkyu/Llama-3.1-8B-RAQUEL-TOFU-M-orig-LoRA-v1
LoRA-style vs meta-llama/Llama-3.1-8B. Looks like a merged LoRA (or a fine-tune with frozen norms): only the weight matrices of down_proj, gate_proj, k_proj, o_proj, q_proj, up_proj, v_proj changed, in 32 of 32 layers; norms, biases and embeddings are untouched.
Tensors changed: Embeddings 0/1, Norms 0/65, MLP 96/96, Attention 128/128, lm_head 0/1
5 derivatives checked
- Hyukkyu/Llama-3.1-8B-RAQUEL-TOFU-Unlearn-IDK-GD-LoRA-v1: LoRA-style
- Hyukkyu/Llama-3.1-8B-RAQUEL-TOFU-Unlearn-GA-GD-LoRA-v1: LoRA-style
- Hyukkyu/Llama-3.1-8B-RAQUEL-TOFU-Unlearn-NPO-GD-LoRA-v1: LoRA-style
- Hyukkyu/Llama-3.1-8B-RAQUEL-TOFU-Unlearn-SAUL-LoRA-v1: LoRA-style
- Hyukkyu/Llama-3.1-8B-RAQUEL-TOFU-Unlearn-RMU-LoRA-v1: LoRA-style