Runtime error: Cuda error:out of memory

VITA-Group / FasterSeg

[ICLR 2020] "FasterSeg: Searching for Faster Real-time Semantic Segmentation" by Wuyang Chen, Xinyu Gong, Xianming Liu, Qian Zhang, Yuan Li, Zhangyang Wang

MIT License

526 stars 107 forks source link

Runtime error: Cuda error:out of memory #9

Closed riazspace closed 4 years ago

riazspace commented 4 years ago

Hi Thanks for sharing your nice work. i have found an issue while using your code for training on my Nvidia GTX 1080 ti card. Can you please help me about this issue?

The issue is

when i run train_search.py it generate the Runtime Error. The error pic is attached Screenshot from 2020-01-16 14-22-52

chenwydj commented 4 years ago

As it says, the GPU card you are using is running out of its graphic memory.

Check if you have multiple jobs on one card. If not, you could turn down the batch size or crop size to reduce the GPU memory usage.

TamuseDomain commented 3 years ago

Hi, @riazspace I would like to know if you were able to run. I'm trying with GeForce GTX 1650 but I get a similar error as your screenshot:

I already tried turning down the batch size, crop size and even the number of workers. Will the GeForce be able to run this work, or do I need a more capable GPU?