Downloading speed of python packages in jetson board too poor

[0831ww59]

I need to install the scikit-learn package and more packages on my board over time. When I attempted to install scikit-learn, numpy, joblib, and scipy were downloaded at a very slow speed. After scipy was downloaded, it remained stagnant for over 5 minutes. After that, it stated that it would execute setup.py, but even that took an unbearable amount of time. Eventually, the download failed, and through the progress bar, I could tell that the time taken to download packages from the web wasn’t the issue. In other words, I don’t think it’s a network problem. How can I resolve this issue?

[AastaLLL]

Hi,

Usually, this kind of library will build binary directly on the target.
So it takes time.

Could you share the error you met with us so we can share some suggestions for you?
Thanks.

[0831ww59]

Actually there is no sounds like error.
It just takes a lot of time.
I’m worring if some process will show poor performance like PyTorch, scikit-learn, etc…
I’ll attach a log can show where I got stuck while download

jetsonxavier@jetsonxavier:~/ldlidar_ws$ pip3 install numpy
Collecting numpy
Retrying (Retry(total=4, connect=None, read=None, redirect=None, status=None)) after connection broken by ‘ProtocolError(‘Connection aborted.’, ConnectionResetError(104, ‘Connection reset by peer’))’: /pypi/packages/51/60/3f0fe5b7675a461d96b9d6729beecd3532565743278a9c3fe6dd09697fa7/numpy-1.19.5.zip
Downloading https://mirror.kakao.com/pypi/packages/51/60/3f0fe5b7675a461d96b9d6729beecd3532565743278a9c3fe6dd09697fa7/numpy-1.19.5.zip (7.3MB)
100% |████████████████████████████████| 7.3MB 2.0MB/s
Building wheels for collected packages: numpy
Running setup.py bdist_wheel for numpy … |

[dusty_nv]

@0831ww59

Running setup.py bdist_wheel for numpy … |

@0831ww59 what it’s doing is building numpy from source, which can take some minutes (you can watch it’s progress if you run pip with --verbose flag)

If you upgrade pip first, it should be able to make use of the manyarch packages that were introduced to PyPi, and it may not need to build it from source for you. You can try that with python3 -m pip install --upgrade pip first.