Image Modeling on ImageNet 64x64 (val)
3.4NLL (bits/dim)Perceiver AR
Evaluation results
| Method | Links | |
|---|---|---|
| Perceiver AR#Params=770M2024.06 | 3.4 | |
| MD4#Params=198M2024.06 | 3.4 | |
| Routing Transformer2024.06 | 3.43 | |
| Sparse Transformer#Params=152M2024.06 | 3.44 | |
| Gated PixelCNN2024.06 | 3.57 | |
| PixelRNNResolution=64x64, Data Augmentation=false2016.01 | 3.63 | |
| PixelRNN2024.06 | 3.63 | |
| RESIDUAL FLOWDequantization=uniform2020.02 | 3.76 | |
| GLOWDequantization=uniform2020.02 | 3.81 | |
| FFJORD RNODEDequantization=uniform, TIME=64.12020.02 | 3.83 | |
| REAL NVPDequantization=uniform2020.02 | 3.98 |