search

MIC-DKFZ / medicaldetectiontoolkit

The Medical Detection Toolkit contains 2D + 3D implementations of prevalent object detectors such as Mask R-CNN, Retina Net, Retina U-Net, as well as a training and inference framework focused on dealing with medical images.
Apache License 2.0
1.31k stars 297 forks source link

GPU Memory limitation #85

Closed awp4211 closed 4 years ago

awp4211 commented 5 years ago

When I run the code exec.py on the toy dataset generated by ./experiments/toy_exp, error occurs as: RuntimeError: CUDA error: out of memory

My computer is installed with a NVIDIA TITAN V with 12GB memory.

What is the hardware requirement of the code?

Gregor1337 commented 4 years ago

Hi awp4211, if you create and run the toy experiment with the given default settings, the VRAM requirements should be satisfied with 12GB (it works on our machines with less than 12GB in that case). Have you tried enabling torch.backends.cudnn.benchmark = True at the beginning of training to optimize GPU-resource utilization?