michaelgathara/vit-face-affectnet
Full fine-tune vs trpakov/vit-face-expression. Fine-tune: 100% of the weights changed, across 12 of 12 layers.
Tensors changed: other 50/50, Embeddings 4/4, Attention 96/96, Norms 50/50
Full fine-tune vs trpakov/vit-face-expression. Fine-tune: 100% of the weights changed, across 12 of 12 layers.
Tensors changed: other 50/50, Embeddings 4/4, Attention 96/96, Norms 50/50