Memory usage is quite good, but why is always in the last column 1%

CodePudding user response:

Basic is useless, only 1%

CodePudding user response:

I was worse than you, is 0%, baidu won't solve the fundamental, it is shit,

CodePudding user response:

1. Set the batch_size bigger (such as the batch_size=512), as far as possible using the GPU
2. The train_loader pin_memory=True, num_workers=8 or 16, improve the calculation speed

CodePudding user response:

The dynamic, and it should be 1% of that figure