I’m trying to install python packages on an Xavier to test my system on it, and pip is being extremely slow. Installing packages like numpy, scipy, or cvxpy take hours to install, usually stalling on stages like downloading, or "Running setup.py install for numpy … ". Additionally, when I try to update pip with “pip3 install -U pip”, I end up getting pip module not found for “main”, and have to reinstall pip entirely. I suppose it might be possible there are issues with architecture, not sure how python works for ARM vs x86, but updating pip shouldn’t be caused by that at all.
Maybe maximize the device performance can improve the installation faster.
To fix the pip error, please try the following error:
Edit file ‘/usr/bin/pip’
diff --git a/pip b/pip
index 56bbb2b..62f26b9 100755
@@ -6,6 +6,6 @@ import sys
# Run the main entry point, similarly to how setuptools does it, but because
# we didn't install the actual entry point from setup.py, don't use the
# pkg_resources API.
-from pip import main
+from pip import __main__
if __name__ == '__main__':
Thank you for the script fix.
I’m experiencing exactly the same thing as kurtzco. Downloading, installing packages and building wheels (h5py, grpcio, pyzmq, etc.) which usually takes 2-5 minutes on a generic PC, takes 30+ minutes on the Xavier.
sudo ./jetson_clocks did not fix this.
I’m seeing this as well. Even with jetson_clocks, installing basic packages like numpy is agonizingly slow on the Xavier AGX. Is there anything else that we can try to speed this up?
Sorry that these python package usually requires some compiling, which make the installation slow.
However, this is an one-time job and won’t affect the performance when using these libraries.
Plase modify the scripts in order to use at least 5 or 6 cores! It is compiling on a single core.
Whenever I use make, I put the -j 6 option and it runs obviously a lot faster.