Build Error: pytorch 1.0.2 for Python 3.6.6 on Mac

Looking for help with this package that is failing to build in my project.

Runtime Details

Failing Packages

  • pytorch 1.0.2

Reasons

  • Build failures: pytorch
  • Test failures: pytorch

Hi @hejibo,

Thanks for letting us know!

It looks like you’ve changed the version of Python that you are using for your build to Python 3.8.2 - and that you are having trouble getting pyqtbuilder to build.

Do you have the full list of dependencies that you need for your build? From the name of the build, it looks like you are making Linux/Mac/Windows runtimes for a Python data science course. Some of those packages have tricky dependencies.

Let us know how we can help.

Cheers!
–zak

Thanks, I teach python twice a year at Tsinghua University on the topics of psychopy, data analysis, graphing, machine learning and python for everything using BeWare. To ease beginners’ worrying and trouble in installing these packages, I hope to use your activestate.com technology to custom build the following packages for my students, so that they can download , use it and do not worry about pip install any pakcages. I have been playing with your website for one week, but I have not been able to complile the following packages for windows 7, 10, and mac. Can you help me? Thanks

About 300 students learn python every year with me.

numpy
scipy

matplotlib
seaborn

pandas
openpyxl
xlrd
xlwt
statsmodels

bs4
nltk
sklearn
wordcloud
jieba

Beeware
Pytorch
Keras
tensorflow

Hey @hejibo,

Thank you for the additional information!

We’ll help out as best we can. One of our technical team will be in touch later today.

Cheers!
–zak

@hejibo,

My name is Gregory Boyce, and I’m the community build engineer here at ActiveState. I’d like to help you get this project up and running.

Since you require Mac OS X support, our best bet for this project is to use Python 3.8.2. I see from your project history that you’ve experienced a number of build failures when you attempted to use that version. I’m going to take a look and determine what I can do in order to fix the packages which are currently not building.

I’ve created a fork of your project in my workspace so that I can make some adjustments:
https://platform.activestate.com/gboyce/hejibo

In the meantime, please don’t hesitate to reach out if there’s any questions I can answer for you.

– Greg

Thanks so much for your quick response. I appreciate. I am just one step away from teaching machine learning EASILY for students now.

Can you help me add the following four packages for the distributions you forked from me? Thanks

sklearn or scikit-learn
Pytorch
Keras
tensorflow

Hi @hejibo,

Sorry for the slow response! We had holidays here. We’ll work on those packages as well.

Cheers!
–zak

@hejibo,

Apologies again for the delay. We’re hitting some snags due to issues with package support on the Mac. Currently we do not support MKL acceleration on the Mac platform, and the lack of those support libraries is causing a problem with a number of these packages.

I’m currently pursing a hybrid option that should allow for better support for you.

1 Like

@hejibo

I have a test project for you to test out:
https://platform.activestate.com/gboyce/hejibo-test

I’ve moved MacOS support to a separate branch so that the project as a whole isn’t held back by the Mac support. The main branch includes more up to date versions of your requested packages, excluding Torch, which we do not yet support. The macos branch strips out the packages which require numpy so that you can add in an OpenBLAS accelerated version via pip.

To set up your environment on Windows or Linux, you would run:

$ state activate gboyce/hejibo-test
$ python3 -m pip install torch

To set up your environment on MacOS you would run:

$ state activate gboyce/hejibo-test --branch macos
$ python3 -m pip install -r requirements.txt

The content of requirements.txt should be:

numpy==1.20.2
tensorflow==2.2.0
tensorboard==2.2.0
scikit-learn==0.24.1
matplotlib==3.3.0
scipy==1.4.1
pandas==1.2.4
keras==2.3.1
statsmodels==0.12.1
chineseize-matplotlib==1.2
matplotlib-doc-zh==2020.9.21.0
matplotlib-inline==0.1.2
openscienceplot-matplotlib==0.2.0
wordcloud==1.8.0
torch

Please let me know what you think of this approach. If it meets your needs, we can merge these changes into your project.

1 Like

Yes. Thanks so much. I appreciate your great help. Pytorch is more popular than tensorflow in academia. If you can make a distribution of pytorch in the future ,please let me know. I will use the distribution you helped me build in my python teaching.

Jibo He

1 Like

Glad to hear this works for you.

If you haven’t already done so, please fork this project into your account. This way you can rename or manage the project, and there’s no risk of me accidentally deleting it.

Thanks for the heads up on pytorch. I’ll be sure that is taken into account when prioritizing new packages.