Windows
Download the wheel for openEXR from:
Python Extension Packages for Windows – Christoph Gohlke (uci.edu)
|
1 2 |
python -m pip install OpenEXR-1.3.0-cp36-cp36m-win_amd64.whl |
Ubuntu:
|
1 2 3 |
sudo apt-get install zlib1g-dev sudo apt-get install libopenexr-dev sudo apt-get install openexr |
|
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 |
grads = {} def save_grad(name): def hook(grad): grads[name] = grad return hook x = Variable(torch.randn(1,1), requires_grad=True) y = 3*x z = y**2 # In here, save_grad('y') returns a hook (a function) that keeps 'y' as name y.register_hook(save_grad('y')) z.register_hook(save_grad('z')) z.backward() print(grads['y']) print(grads['z']) |
Thanks to Adam Paszke’s post in Pytorch Discussion
I struggled with a problem today: My parameter “b” is not updating in the following code:
|
1 2 |
b = nn.Parameter(torch.ones(batch_size, 1) a = torch.norm(b, dim=1) |
There’s nothing wrong with the gradient of “a”. So what’s the problem?
The problem is: I used the wrong initialization of “b”. I init “b” with all zeros. and the gradient of the norm of an all-zero vector is always zero.
Today I met an error:
a is a float32 pytorch tensor; and b is a float64 pytorch tensor with all elements equals 1. When I perform the following multiplication:
|
1 |
c = a * b |
The result of “c” is 0.
The error will be fixed when b is converted to float32.
zWhen trying to install SoftRas & neural renderer, I got error:
unable to execute ‘:/usr/local/cuda/bin/nvcc’: No such file or directory
error: command ‘:/usr/local/cuda/bin/nvcc’ failed with exit status 1
Solution:
|
1 2 |
-export CUDA_HOME=$CUDA_HOME:/usr/local/cuda +export CUDA_HOME=/usr/local/cuda |
Config:
System: Ubuntu 16.04
CUDA: 10.0
Graphic Crad: RTX 2080
——————————————————————————————————————————
|
1 |
git clone https://github.com/CMU-Perceptual-Computing-Lab/openpose.git --recursive |
|
1 2 3 |
cmake .. make all make install |
|
1 2 3 4 |
CMake Error: The following variables are used in this project, but they are set to NOTFOUND. Please set them or make sure they are set and tested correctly in the CMake files: CUDA_curand_LIBRARY (ADVANCED) linked by target "caffe" in directory /home/xiaoxu/Documents/rgb2mesh/Borrowed/openpose/3rdparty/caffe/src/caffe |
Solve: open “openpose/build/caffe/src/openpose_lib-build/CMakeCache.txt” with cmake-gui
Change cuda-9.0 to cuda-10.0
|
1 2 3 4 5 6 7 8 9 10 |
(pytorch_py3) xiaoxu@chuan:~/Documents/rgb2mesh/Borrowed/openpose/build/examples/tutorial_api_python$ python 01_body_from_image.py Error: OpenPose library could not be found. Did you enable `BUILD_PYTHON` in CMake and have this Python script in the right folder? Traceback (most recent call last): File "01_body_from_image.py", line 26, in raise e File "01_body_from_image.py", line 23, in from openpose import pyopenpose as op File "../../python/openpose/__init__.py", line 1, in from . import pyopenpose as pyopenpose ImportError: cannot import name 'pyopenpose' |
|
1 2 3 |
(pytorch_py3) xiaoxu@chuan:~/Documents/rgb2mesh/Borrowed/openpose/build/python/openpose$ make [ 97%] Built target openpose [100%] Built target pyopenpose |
Configuration:
Ubuntu 16.04
CUDA 10.0
GTX 2080
The reference is the step-by-step tutorial.
I met one error:
To solve it, two steps are necessary:
You know, the ubuntu system in my Great Alienware is not healthy. Many functions are not working well, including the wifi connection. It is impossible to connect to a new wifi using the graphical interface. I will get error like:
I solved the connection issue using
|
1 |
nmcil |
|
1 2 3 4 5 |
(base) xiaoxu@chuan:~$ nmcli d DEVICE TYPE STATE CONNECTION wlp4s0 wifi disconnected -- enp5s0 ethernet unavailable -- lo loopback unmanaged -- |
|
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 |
(base) xiaoxu@chuan:~$ nmcli d wifi list * SSID MODE CHAN RATE SIGNAL BARS SECURITY xfinitywifi Infra 6 54 Mbit/s 84 ▂▄▆█ HOME-CBE2 Infra 6 54 Mbit/s 82 ▂▄▆█ WPA1 WPA2 Neptune_EXT Infra 1 54 Mbit/s 77 ▂▄▆_ WPA1 WPA2 NETGEAR91 Infra 11 54 Mbit/s 60 ▂▄▆_ WPA2 Fios-TVB5U Infra 11 54 Mbit/s 54 ▂▄__ WPA2 2NXG7 Infra 6 54 Mbit/s 45 ▂▄__ WPA2 Zeus_EXT Infra 153 54 Mbit/s 45 ▂▄__ WPA1 WPA2 FiOS-A5O2K Infra 6 54 Mbit/s 40 ▂▄__ WPA2 nevetica Infra 6 54 Mbit/s 40 ▂▄__ WPA1 WPA2 -- Infra 6 54 Mbit/s 39 ▂▄__ WPA1 WPA2 Fios-TVB5U-5G Infra 161 54 Mbit/s 35 ▂▄__ WPA2 -- Infra 161 54 Mbit/s 35 ▂▄__ WPA2 -- Infra 1 54 Mbit/s 30 ▂___ WPA1 WPA2 DIRECT-PG-FireTV_3568 Infra 153 54 Mbit/s 30 ▂___ WPA2 -- Infra 1 54 Mbit/s 29 ▂___ WPA1 WPA2 802.1X -- Infra 1 54 Mbit/s 27 ▂___ OutOfService Infra 11 54 Mbit/s 27 ▂___ WPA2 802.1X -- Infra 161 54 Mbit/s 27 ▂___ WPA2 Axify Infra 11 54 Mbit/s 25 ▂___ WPA2 sergek5 Infra 36 54 Mbit/s 25 ▂___ WPA1 WPA2 xfinitywifi Infra 36 54 Mbit/s 24 ▂___ XFINITY Infra 36 54 Mbit/s 24 ▂___ WPA2 802.1X -- Infra 36 54 Mbit/s 24 ▂___ WPA1 WPA2 FiOS-A5O2K-5G Infra 161 54 Mbit/s 24 ▂___ WPA2 -- Infra 36 54 Mbit/s 22 ▂___ WPA1 WPA2 802.1X |
|
1 |
(base) xiaoxu@chuan:~$ nmcli d wifi connect ** password ** |
|
1 2 3 4 5 |
(base) xiaoxu@chuan:~$ nmcli d DEVICE TYPE STATE CONNECTION wlp4s0 wifi connected NETGEAR91 1 enp5s0 ethernet unavailable -- lo loopback unmanaged -- |
|
1 2 3 4 |
value = 0.0205862828046875 print("%.16f" % value) value = value.astype(np.float32) print("%.16f" % value) |
|
1 2 3 |
#output 0.0205862828046875 0.0205862820148468 |
Precision changed!
Another sad experience with cuda.
|
1 |
sudo update-alternatives --install /usr/bin/g++ g++ /usr/bin/g++-5.3 20 |
|
1 |
sudo update-alternatives --install /usr/bin/g++ g++ /usr/bin/g++-7.3 60 |
|
1 2 3 4 5 6 7 8 9 10 11 |
sudo update-alternatives --config gcc // You will see the following: There are 2 choices for the alternative gcc (providing /usr/bin/gcc). Selection Path Priority Status ------------------------------<wbr />------------------------------ * 0 /usr/bin/gcc-7 60 auto mode 1 /usr/bin/gcc-5 60 manual mode 2 /usr/bin/gcc-7 60 manual mode Press <enter> to keep the current choice[*], or type selection number: |