search

MVIG-SJTU / AlphaPose

Real-Time and Accurate Full-Body Multi-Person Pose Estimation&Tracking System
http://mvig.org/research/alphapose.html
Other
7.95k stars 1.97k forks source link

Jetson TX2 #233

Open ghost opened 5 years ago

ghost commented 5 years ago

Hello, I have installed jetpack 3.3 on my jetson and also was able to install Tensorflow which directly available for Jetson TX2 on Nvidia website. But I am unable to install pytorch on the Jetson TX2. Is there any workaround for this?

JaviLaplaza commented 5 years ago

My experience tells me that the best way to go with Jetson TX2 and pytorch is:

  1. Create a virtual environment (mind the python version that you want to use in said virtual environment). For python 3 just do> python3 -m venv

  2. Inside this virtual environment, install all the required pytorch dependencies: pip install numpy scipy matplotlib scikit-image scikit-learn ipython protobuf jupyter tqdm ipykernel

  3. Install pytorch from source since I believe that the pip wheels don't support Jetson TX2 processor. To do so, get PyTorch source with: git clone --recursive https://github.com/pytorch/pytorch cd pytorch python setup.py build_deps --user python setup.py install --user

  4. You will probably want to install torchvision as well: pip --no-deps install torchvision

I don't know if this is the proper way to do it, but it worked for me with a Jetson flashed with Jetpack 3.3. If you want to install Pytorch 0.4 instead of the newest version, I guess you can get the source code switching the github branch with "git checkout ".