Kindly not that you are running one notebook at an instance.This is because when you run 2 or more notebooks: each is trying to access one GPU resource which in turn results in memory error.
On Kaggle, select the T4 X2 GPU, and make sure to set the device in the train call to (0, 1). That way, Kaggle trains on both GPUs and it should fit. You can also reduce the batch size.
Kindly not that you are running one notebook at an instance.This is because when you run 2 or more notebooks: each is trying to access one GPU resource which in turn results in memory error.
Run one project at a time
On Kaggle, select the T4 X2 GPU, and make sure to set the device in the train call to (0, 1). That way, Kaggle trains on both GPUs and it should fit. You can also reduce the batch size.