Error in running your code.

[sentientmachine]

Here's what I did:

get virtualenv as root: emerge -avNu virtualenv
cd /home/el/code/python/100_liam_wilson_comp_vision_room_analyzer/pytorch.room.layout
virtualenv .env
source .env/bin/activate
#you've left me to guessing and checking versions at this point so lets use most up to date ones.
pip install opencv-python
pip install Pillow
pip install torch
pip install torchvision

We're off to the races, so I tried:

python segment_st.py train -d datasets/sunrgbd/ -c 37 -s 480 --arch drn_d_105 --batch-size 32 --random-scale 1.75 --random-rotate 15 --epochs 100 --lr 0.01 --momentum 0.9 --lr-mode poly

and I'm getting:

THCudaCheck FAIL file=/pytorch/aten/src/THC/THCGeneral.cpp line=74 error=35 : CUDA driver version is insufficient for CUDA runtime version
Traceback (most recent call last):
  File "segment_st.py", line 785, in <module>
    main()
  File "segment_st.py", line 779, in main
    train_seg(args)
  File "segment_st.py", line 391, in train_seg
    model = torch.nn.DataParallel(single_model).cuda()
  File "/home/el/code/python/100_liam_wilson_comp_vision_room_analyzer/pytorch.room.layout/.env/lib/python2.7/site-packages/torch/nn/modules/module.py", line 258, in cuda
    return self._apply(lambda t: t.cuda(device))
  File "/home/el/code/python/100_liam_wilson_comp_vision_room_analyzer/pytorch.room.layout/.env/lib/python2.7/site-packages/torch/nn/modules/module.py", line 185, in _apply

Very likely due to versioning conflicts from my having to guess the layout of your non standard packages, so therefore the solution is for you to show us what versions of all non standard packages you are using, since they are very brittle and the latest packages break your torch specific code. The OpenCV2 seems to indicate you're using python2, yet the parameter iteration you're using seems to be only compatible with python3.

What operating system are you using, what version of python interpreter are you using, and what legacy versions of cv2, pillow, torch and torchvision are you using? Without this info all we can do is get blasted in the face with syntax errors playing guess and check.

Edit: ok lel, you're using Python 3.5 which has syntax incompatible with python 3.4 and python 2.7. Not sure if topkek or NPC wojack. Carry on.

[liamw96]

Hi there,

Actually most codes are borrowed from https://github.com/fyu/drn, which uses python3.

Specifically, I am using

python 3.6.6
torch 0.4.1
gcc 7.2.0
PIL 5.2.0
64-bit ubuntu 16.04.5
cv2 3.1.0
torchvision 0.2.1
linux kernel 4.4.0-103-generic

As for opencv, I am not sure ... but it seems that there is no cv3 according to http://answers.opencv.org/question/63405/why-do-python-bindings-for-v3-still-use-package-name-cv2/