huge gpu memory consume

[forvd]

I tried to train on TITAN(12GB Total), but it always went out of memory. Not even batch_size = 1.Is that normal?

[xy-guo]

For each image, it consumes around 4.2GB memory on average. You may need to set both batch_size and test_batch_size to 1.

[forvd]

sry, i modified some of the code and forget about it, that cause the problem. But I also want to know about memory consumption in inference.

[xy-guo]

The inference memory depends on the input size and the code. I forgot the accurate memory consumption in the inference mode