almanach/camemberta-chatgptdetect-noisy
Full fine-tune vs almanach/camemberta-base. Fine-tune: 100% of the weights changed, across 12 of 12 layers.
Tensors changed: Embeddings 4/5, Norms 52/52, other 50/50, Attention 96/96
Full fine-tune vs almanach/camemberta-base. Fine-tune: 100% of the weights changed, across 12 of 12 layers.
Tensors changed: Embeddings 4/5, Norms 52/52, other 50/50, Attention 96/96