search

PangzeCheung / Dual-task-Pose-Transformer-Network

[CVPR 2022] Exploring Dual-task Correlation for Pose Guided Person Image Generation
Other
95 stars 11 forks source link

About the training epochs? #3

Closed Blackkinggg closed 2 years ago

Blackkinggg commented 2 years ago

How many epochs should I train to get the results you shown in the paper. It seems need 140 epochs, which will cost 1 week. I use one Tesla V100-32g in training time, and when using multiple gpus, there will be serious load imbalance.

PangzeCheung commented 2 years ago

How many epochs should I train to get the results you shown in the paper. It seems need 140 epochs, which will cost 1 week. I use one Tesla V100-32g in training time, and when using multiple gpus, there will be serious load imbalance.

Thanks for your attention. We did not optimize the load imbalance caused by multi GPU training. For Deepfashion, DPTN does take around 200 epochs to train (over one week on NVIDIA A100 40GB PCIe). Results may converge after 170 epochs (see Figure 6 in our paper).