kelim/deep-ignorance-unlearned-cb
Full fine-tune vs EleutherAI/deep-ignorance-unfiltered. Fine-tune: 91% of the weights changed, across 31 of 32 layers.
Tensors changed: Embeddings 0/2, Attention 124/128, Norms 65/130, MLP 124/128
Full fine-tune vs EleutherAI/deep-ignorance-unfiltered. Fine-tune: 91% of the weights changed, across 31 of 32 layers.
Tensors changed: Embeddings 0/2, Attention 124/128, Norms 65/130, MLP 124/128