A batch too large: finding the batch size that fits on GPUs

A batch too large: Finding the batch size that fits on GPUs

A simple function to identify the batch size for your PyTorch model that can fill the GPU memory

I am sure many of you had the following painful experience: you start multiple ML experiments on your GPUs to train overnight and when you came back to check 10 hours later, you realize that the progress bar has barely moved due to hardware underutilization, or worse, all the experiments …

batch-size deep learning gpus programming pytorch