Density Estimation on ImageNet 64x64 (val)
3.4Bits/dimVDM
Evaluation results
| Method | Links | |
|---|---|---|
| VDMType=Diff2022.02 | 3.4 | |
| Perceiver ARType=AR2022.02 | 3.4 | |
| CombinerType=AR2022.02 | 3.42 | |
| Routing TransformerType=AR2022.02 | 3.43 | |
| Sparse TransformerNumber of parameters=152M, Attention pattern=strided2019.04 | 3.44 | |
| Sparse TransformerType=AR2022.02 | 3.44 | |
| SPNNumber of parameters=150M2019.04 | 3.52 | |
| PixelCNN2019.04 | 3.57 | |
| PixelCNNType=AR2022.02 | 3.57 | |
| PixelRNN2016.05 | 3.63 | |
| Parallel Multiscale2019.04 | 3.7 | |
| Glow2019.04 | 3.81 | |
| Real NVP2016.05 | 3.98 |