search

simonmeister / UnFlow

UnFlow: Unsupervised Learning of Optical Flow with a Bidirectional Census Loss
MIT License
294 stars 57 forks source link

ResourceExhaustedError - conv6 #32

Closed lenazherdeva closed 6 years ago

lenazherdeva commented 6 years ago

Hello! Thank you for work!

But I have a question. I'm trying to run run.py with kitti dataset. But I got an error ResourceExhaustedError (see above for traceback): OOM when allocating tensor of shape [1024] and type float [[Node: flownet_c/conv6/biases/Adam/Initializer/zeros = Const[_class=["loc:@flownet_c/conv6/biases"], dtype=DT_FLOAT, value=Tensor<type: float shape: [1024] values: 0 0 0...>, _device="/job:localhost/replica:0/task:0/gpu:0"]()]].

Actually, I used 2 gpu GeForce GTX 108. I played with a batch_size, but nothing helped...

What can cause such a problem?

simonmeister commented 6 years ago

Hi! In general this means that you are out of GPU memory. Do you mean you are using the GTX 1080? How much GPU memory do you have and which batch size did you try? You could run nvidia-smi on the command line to check if something occupies your GPU memory first.

lenazherdeva commented 6 years ago

Yes, you're right! I've checked with nvidia-smi and some process occupied my memory. Now, it works! Thanks a lot!

simonmeister commented 6 years ago

Great! No problem :)