Nikeytas/videomae-crime-detector-fixed-format
Full fine-tune vs MCG-NJU/videomae-base. Fine-tune: 100% of the weights changed, across 12 of 12 layers.
Tensors changed: Embeddings 2/2, Attention 84/84, other 48/48, Norms 50/50
Full fine-tune vs MCG-NJU/videomae-base. Fine-tune: 100% of the weights changed, across 12 of 12 layers.
Tensors changed: Embeddings 2/2, Attention 84/84, other 48/48, Norms 50/50