When we are using pytorch to train a deep learning model, we may get this error: RuntimeError: DataLoader worker (pid xxx) is killed by signal: Killed. In this tutorial, we will introduce you how to fix it.
This error looks like below:
How to fix this RuntimeError?
The reason for his error occurs is there is no enough memory space to store data. In order to fix this error, we can try these methods:
- 1.decrease batch size
- 2.do not save more data in memory, you can load them when need
- 3.increase the memory space
There is another tutorial for fixing memory error:
Fix Tensorflow Allocation Memory: Allocation exceeds 10% of system memory Error