zijundeng/pytorch-semantic-segmentation

Configuration issues

frogM opened this issue · 0 comments

frogM commented

We have some issues when training FCN8s:

  1. The error "Torch.floattensor: object does not support indexing" occurs when the code runs to "writer.add_image(snapshot_name, val_visual)" in line 190, file "train.py". Our configuration is python 2.7, pytorch 0.2.0.
  2. According to your setup instructions, we should install tensorboardX, however, your source code uses tensorboard. Which one should I use?
  3. We want to know if the code can run on GPU.
    Thank you very much!