Target 10 out of bounds while training on complete dataset

[parthkvv]

I am training my own dataset with 2 classes. Got error : "CUDA error: device-side assert triggered." When I tried to run with CPU, got the error : "IndexError: Target 10 is out of bounds".
Running the model on a much smaller sample of my data works perfectly fine, the error only pops up when I try to train it on my complete data.
(during training, batch size : 4 )
input image shape : torch.Size([4, 3, 513, 513])
target shape : torch.Size([4, 513, 513])
output shape : torch.Size([4, 2, 513, 513])

Tried changing batch size and and the number of workers in arguments, but no use.
@jfzhang95 Can you suggest anything based upon this?

[chiba1sonny]

The same problem.
Have you solved it?