OOM when allocating tensor with shape

[TingZhang08]

Hi, thank you very much for contributing this codes.

However, when I evaluate with ade20k dataset, I got the error:

"OOM when allocating tensor with shape"

I'm using two Tesla K40c with total memory 12G.

Could you help me?

Thank you in advance!

[TingZhang08]

I have solved my problem by changing

config.gpu_options.allow_growth = True

with

config.gpu_options.allow_growth = 0.9

Thanks for your contribution.

[rahul411]

Hey, I get a same issue while training on cityscapes. I tried your solution but it doesn't work, any help would be appreciated.

[ybxbupt]

change BATCH_SIZE = 1 in train.py

[yacineDZ44]

hi, I had the same problem with inference.py I tried changing "config.gpu_options.allow_growth = True" to "config.gpu_options.allow_growth = 0.9" but I get a TypeError how can i change the BATCH_SIZE ?or any other solutions? Thank you.