Interchained/imagine-v10
Full fine-tune vs Interchained/imagine-v9. Fine-tune: 95% of the weights changed, across 24 of 24 layers.
Tensors changed: lm_head 1/1, Embeddings 0/1, Norms 5/49, MLP 72/72, Attention 96/96
Full fine-tune vs Interchained/imagine-v9. Fine-tune: 95% of the weights changed, across 24 of 24 layers.
Tensors changed: lm_head 1/1, Embeddings 0/1, Norms 5/49, MLP 72/72, Attention 96/96