Search papers, labs, and topics across Lattice.
1
0
3
37
Train high-resolution vision models on a desktop GPU: AFFMAE slashes FLOPs by 7x and halves memory usage compared to ViT-MAE, without sacrificing performance.
