How to use multi gpu?

[yifan123]

How to use multi gpu?

The most time-consuming part of the code is the load part of the data set, and I want to use multiple GPUs to speed it up.

Cuda

Cuda will be used by default if it is available. When training on large portions of the dataset, multiple GPUs is favorable.

SIMULATOR_GPU_IDS: [0,1]  # Each GPU runs NUM_ENVIRONMENTS environments
TORCH_GPU_ID: 0
NUM_ENVIRONMENTS: 1

I followed the instructions in the readme. After correcting, it showed repeated training in the log, which seems to be wrong.

Does the code really support multiple GPUs?

2021-06-22 01:49:53,897 [Epoch: 1/15] [Batch: 1/6568] [BatchTime: 118.16s] [EpochTime: 118s] [Loss: 1.7911]
2021-06-22 01:49:54,260 [Epoch: 1/15] [Batch: 1/6568] [BatchTime: 118.52s] [EpochTime: 119s] [Loss: 1.7911]                   
2021-06-22 01:49:55,733 [Epoch: 1/15] [Batch: 2/6568] [BatchTime: 1.83s] [EpochTime: 120s] [Loss: 1.7777] 
2021-06-22 01:49:56,771 [Epoch: 1/15] [Batch: 2/6568] [BatchTime: 2.51s] [EpochTime: 121s] [Loss: 1.7777] 
2021-06-22 01:49:56,820 [Epoch: 1/15] [Batch: 3/6568] [BatchTime: 1.09s] [EpochTime: 121s] [Loss: 1.7504] 
2021-06-22 01:49:57,925 [Epoch: 1/15] [Batch: 3/6568] [BatchTime: 1.15s] [EpochTime: 122s] [Loss: 1.7504] 
2021-06-22 01:50:01,319 [Epoch: 1/15] [Batch: 4/6568] [BatchTime: 4.5s] [EpochTime: 126s] [Loss: 1.6842] 
2021-06-22 01:50:02,464 [Epoch: 1/15] [Batch: 4/6568] [BatchTime: 4.54s] [EpochTime: 127s] [Loss: 1.6842]
2021-06-22 01:50:03,281 [Epoch: 1/15] [Batch: 5/6568] [BatchTime: 1.96s] [EpochTime: 128s] [Loss: 1.5942] 
2021-06-22 01:50:04,933 [Epoch: 1/15] [Batch: 5/6568] [BatchTime: 2.47s] [EpochTime: 129s] [Loss: 1.5942] 
2021-06-22 01:50:08,206 [Epoch: 1/15] [Batch: 6/6568] [BatchTime: 4.92s] [EpochTime: 132s] [Loss: 1.4861] 
2021-06-22 01:50:09,176 [Epoch: 1/15] [Batch: 6/6568] [BatchTime: 4.24s] [EpochTime: 133s] [Loss: 1.4861]
2021-06-22 01:50:12,253 [Epoch: 1/15] [Batch: 7/6568] [BatchTime: 4.05s] [EpochTime: 137s] [Loss: 1.3834]
2021-06-22 01:50:12,607 [Epoch: 1/15] [Batch: 7/6568] [BatchTime: 3.43s] [EpochTime: 137s] [Loss: 1.3834] 
2021-06-22 01:50:13,432 [Epoch: 1/15] [Batch: 1/6568] [BatchTime: 114.51s] [EpochTime: 115s] [Loss: 1.7911] 
2021-06-22 01:50:15,151 [Epoch: 1/15] [Batch: 1/6568] [BatchTime: 116.23s] [EpochTime: 116s] [Loss: 1.7911] 
2021-06-22 01:50:15,406 [Epoch: 1/15] [Batch: 2/6568] [BatchTime: 1.81s] [EpochTime: 116s] [Loss: 1.7777]   
2021-06-22 01:50:15,726 [Epoch: 1/15] [Batch: 1/6568] [BatchTime: 116.8s] [EpochTime: 117s] [Loss: 1.7911] 
2021-06-22 01:50:15,832 [Epoch: 1/15] [Batch: 1/6568] [BatchTime: 116.91s] [EpochTime: 117s] [Loss: 1.7911]
2021-06-22 01:50:17,043 [Epoch: 1/15] [Batch: 3/6568] [BatchTime: 1.64s] [EpochTime: 118s] [Loss: 1.7504] 
2021-06-22 01:50:17,344 [Epoch: 1/15] [Batch: 8/6568] [BatchTime: 5.09s] [EpochTime: 142s] [Loss: 1.2995] 
2021-06-22 01:50:18,344 [Epoch: 1/15] [Batch: 2/6568] [BatchTime: 2.51s] [EpochTime: 119s] [Loss: 1.7777] 
2021-06-22 01:50:18,430 [Epoch: 1/15] [Batch: 8/6568] [BatchTime: 5.81s] [EpochTime: 143s] [Loss: 1.2995] 
2021-06-22 01:50:18,438 [Epoch: 1/15] [Batch: 2/6568] [BatchTime: 3.29s] [EpochTime: 120s] [Loss: 1.7777] 
2021-06-22 01:50:18,953 [Epoch: 1/15] [Batch: 2/6568] [BatchTime: 3.23s] [EpochTime: 120s] [Loss: 1.7777] 
2021-06-22 01:50:19,627 [Epoch: 1/15] [Batch: 3/6568] [BatchTime: 1.28s] [EpochTime: 121s] [Loss: 1.7504]  
2021-06-22 01:50:19,973 [Epoch: 1/15] [Batch: 3/6568] [BatchTime: 1.53s] [EpochTime: 121s] [Loss: 1.7504]  
2021-06-22 01:50:20,715 [Epoch: 1/15] [Batch: 3/6568] [BatchTime: 1.76s] [EpochTime: 122s] [Loss: 1.7504] 
2021-06-22 01:50:21,385 [Epoch: 1/15] [Batch: 4/6568] [BatchTime: 4.34s] [EpochTime: 122s] [Loss: 1.6842]
2021-06-22 01:50:23,499 [Epoch: 1/15] [Batch: 9/6568] [BatchTime: 6.15s] [EpochTime: 148s] [Loss: 1.2126] 
2021-06-22 01:50:23,876 [Epoch: 1/15] [Batch: 4/6568] [BatchTime: 4.25s] [EpochTime: 125s] [Loss: 1.6842] 
2021-06-22 01:50:25,250 [Epoch: 1/15] [Batch: 9/6568] [BatchTime: 6.82s] [EpochTime: 150s] [Loss: 1.2126]   
2021-06-22 01:50:25,257 [Epoch: 1/15] [Batch: 5/6568] [BatchTime: 3.4s] [EpochTime: 126s] [Loss: 1.5942]  
2021-06-22 01:50:26,637 [Epoch: 1/15] [Batch: 5/6568] [BatchTime: 2.76s] [EpochTime: 128s] [Loss: 1.5942] 
2021-06-22 01:50:26,908 [Epoch: 1/15] [Batch: 10/6568] [BatchTime: 3.41s] [EpochTime: 151s] [Loss: 1.1721]  
2021-06-22 01:50:27,124 [Epoch: 1/15] [Batch: 4/6568] [BatchTime: 7.15s] [EpochTime: 128s] [Loss: 1.6842] 
2021-06-22 01:50:27,997 [Epoch: 1/15] [Batch: 4/6568] [BatchTime: 6.14s] [EpochTime: 129s] [Loss: 1.6842] 
2021-06-22 01:50:28,676 [Epoch: 1/15] [Batch: 10/6568] [BatchTime: 3.42s] [EpochTime: 153s] [Loss: 1.1721] 
2021-06-22 01:50:31,479 [Epoch: 1/15] [Batch: 5/6568] [BatchTime: 2.48s] [EpochTime: 133s] [Loss: 1.5942]  
2021-06-22 01:50:31,666 [Epoch: 1/15] [Batch: 5/6568] [BatchTime: 4.54s] [EpochTime: 133s] [Loss: 1.5942] 
2021-06-22 01:50:31,779 [Epoch: 1/15] [Batch: 6/6568] [BatchTime: 6.52s] [EpochTime: 133s] [Loss: 1.4861]