nvidia/vgg-ttt
Partial fine-tune vs facebook/VGGT-1B. Partial change: 8% of the weights changed (attention), in 0 of 24 layers.
Tensors changed: other 0/229, Norms 0/312, Attention 96/208, MLP 0/212, Embeddings 0/345
Partial fine-tune vs facebook/VGGT-1B. Partial change: 8% of the weights changed (attention), in 0 of 24 layers.
Tensors changed: other 0/229, Norms 0/312, Attention 96/208, MLP 0/212, Embeddings 0/345