Translations:PyTorch/213/en

Of course, batch_size is also an important parameter with respect to a model's performance on a given task (accuracy, error, etc.) and different schools of thought have different views on the impact of using large batches. This page will not go into this subject, but if you have reason to believe that a small (relative to space in GPU memory) batch size is best for your application, skip to Data Parallelism with a single GPU to see how to maximize GPU utilization with small inputs.