RuntimeError: CUDA out of memory when train about mprnet

[adityac8]

Hi

We use 4 Tesla V100 GPUs to train our model.
You can use either of the options to train your model:

1. Use a smaller patch size by setting the variable TRAIN_PS
2. Use a smaller batch size by setting the variable BATCH_SIZE
3. Use a lighter model as explained here

Thanks

[adityac8]

Hi @Zz1041328406

Could you give details on what task are you trying to run and what dataset are you using.

Thanks

[Zz1041328406]

I tried to train the task : Deblurring, use the Dataset:Gopro(about 14G) , when batch_size chooses 16.

@adityac8