I recently updated my Rainbow code to use multiprocessing, and it worked fine on my Macbook:
- OS: macOS High Sierra 10.13.3
- PyTorch version: 0.3.1 py36_cuda0.0_cudnn0.0_2
- How you installed PyTorch (conda, pip, source): conda
- Python version: 3.6
- CUDA/cuDNN version: N/A
- GPU models and configuration: N/A
However, I run into the following error (gdb backtrace) on my Ubuntu machine:
- OS: Ubuntu 14.04.5 LTS
- PyTorch version: 0.3.1 py36_cuda8.0.61_cudnn7.0.5_2
- How you installed PyTorch (conda, pip, source): conda
- Python version: 3.6
- CUDA/cuDNN version: 8.0
- GPU models and configuration: GeForce GTX 980
It fails pretty quickly, whether or not I’m using CUDA (as evidenced by checking nvidia-smi, and also backed up by the backtrace which looks the same either way).
Edit: Should have checked the PyTorch issues more carefully - this was caused by /dev/shm being too small: https://github.com/pytorch/pytorch/issues/5247