sotte/pytorch_tutorial

Could not install conda environment on OS X

ellenkoenig opened this issue · 7 comments

ellenkoenig commented

Hi Stefan,

first of all, thanks for providing the install instruction for the tutorial in advance :-)
However, I am struggling to install the dependencies on Mac OS X (e.g. the libgcc-ng package). Does the tutorial only work on Linux?

ellenkoenig commented

After some hacking around, I got it to install and I can at least download the data and run the first jupyter lab notebook. Maybe the following helps some other OS X user.

Here is the environment.yml that I ended up with

name: pydata_pytorch_tutorial
channels:
  - pytorch
  - conda-forge
  - defaults
  - anaconda
dependencies:
  - pip:
    - arrow==0.12.1
    - attrs==17.4.0
    - binaryornot==0.4.4
    - black==18.4a2
    - blessings==1.6.1
    - bpython==0.17.1
    - chardet==3.0.4
    - click==6.7
    - cookiecutter==1.6.0
    - curtsies==0.3.0
    - django==1.11.13
    - django-embed-video==1.1.2
    - easy-thumbnails==2.5
    - ffind==1.2.0
    - flake8==3.5.0
    - future==0.16.0
    - gdata==2.0.18
    - greenlet==0.4.13
    - httpie==0.9.9
    - idna==2.6
    - jinja2-time==0.2.0
    - lxml==4.2.1
    - mccabe==0.6.1
    - pelper==0.1.0
    - pew==1.1.2
    - pipenv==11.10.0
    - poyo==0.4.1
    - pycodestyle==2.3.1
    - pyflakes==1.6.0
    - pyperclip==1.6.0
    - requests==2.18.4
    - torch==0.4.0
    - torchvision==0.2.1
    - typing==3.6.4
    - urllib3==1.22
    - virtualenv==15.1.0
    - virtualenv-clone==0.2.6
    - whichcraft==0.4.1
    - certifi==2018.4.16
    - ipywidgets==7.2.1

  - conda:
    - openblas==0.2.20
    - nodejs
    - pip==9.0.3
    - tqdm==4.23.4
    - backcall==0.1.0
    - blas
    - bleach==2.1.3
    - ca-certificates==2018.03.07
    - cffi==1.11.5
    - cycler==0.10
    - cython==0.28.3
    - dbus==1.13.2
    - decorator==4.3.0
    - entrypoints==0.2.3
    - expat==2.2.5
    - fontconfig==2.12.6
    - freetype
    - glib==2.56.1
    - gmp==6.1.2
    - gst-plugins-base
    - gstreamer
    - html5lib==1.0.1
    - icu==58.2
    - intel-openmp==2018.0.0
    - ipykernel==4.8.2
    - ipython==6.3.1
    - ipython_genutils==0.2.0
    - jedi==0.12.0
    - jinja2==2.10
    - jpeg==9b
    - jsonschema==2.6.0
    - jupyter==1.0.0
    - jupyter_client==5.2.3
    - jupyter_console==5.2.0
    - jupyter_core==4.4.0
    - jupyterlab==0.32.1
    - jupyterlab_launcher==0.10.5
    - kiwisolver==1.0.1
    - libedit==3.1
    - libffi==3.2.1
    - libgcc
    - libgfortran
    - libgfortran-ng
    - libpng==1.6.34
    - libsodium==1.0.16
    - libtiff==4.0.9
    - libxcb==1.13
    - libxml2==2.9.8
    - markupsafe==1.0
    - matplotlib==2.2.2
    - mistune==0.8.3
    - mkl==2018.0.2
    - mkl_fft==1.0.1
    - mkl_random==1.0.1
    - nbconvert==5.3.1
    - nbformat==4.4.0
    - ncurses==6.0
    - ninja==1.8.2
    - notebook==5.4.1
    - numpy==1.14.2
    - olefile==0.45.1
    - openssl==1.0.2o
    - pandoc==1.19.2.1
    - pandocfilters==1.4.2
    - parso==0.2.0
    - pcre==8.42
    - pexpect==4.5.0
    - pickleshare==0.7.4
    - pillow==5.1.0
    - prompt_toolkit==1.0.15
    - ptyprocess==0.5.2
    - pycparser==2.18
    - pygments==2.2.0
    - pyparsing==2.2.0
    - pyqt==5.9.2
    - python==3.6.5
    - python-dateutil==2.7.2
    - pytz==2018.4
    - pyzmq==17.0.0
    - qt==5.9.5
    - qtconsole==4.3.1
    - readline==7.0
    - scikit-learn==0.19.1
    - scipy
    - send2trash==1.5.0
    - setuptools==39.1.0
    - simplegeneric==0.8.1
    - sip==4.19.8
    - six==1.11.0
    - sqlite==3.23.1
    - terminado==0.8.1
    - testpath==0.3.1
    - tk==8.6.7
    - tornado==5.0.2
    - traitlets==4.3.2
    - wcwidth==0.1.7
    - webencodings==0.5.1
    - wheel==0.31.0
    - widgetsnbextension==3.2.1
    - xz==5.2.3
    - zeromq==4.2.5
    - zlib==1.2.11
    - torchvision
    - pytorch

On top of that, I needed to run
conda update freetype matplotlib due to some version conflict when importing matplotlib.

I still get an error in the imports though: ModuleNotFoundError: No module named 'ppt'I can't find a related package by googling either. Where does that import come from?

  • 👍1
sotte commented

Hey Ellen,

thanks a lot for the feedback. I'm trying to trouble shoot what's wrong (I'm actually experiencing some problems with pip on my laptop, maybe unrelated).

You can fix the ppt problem by running:

pip install -e .

ppt is the pydata_pytorch_tutorial. I'll update the README.

sotte commented

I used your environment.yml for mac and added some infos to the readme:
d6f0a6c

Still trying to fix the dependencies. Thanks again!!

sotte commented

I added manual install instructions to the README and verified it on two computers, sadly I don't have a Mac to test it :(

Sorry again, this should not happen one day before the tutorial.

https://github.com/sotte/pytorch_tutorial#manual-install

ellenkoenig commented

Thanks for the additional information, now the imports all worked :)

If you want to make the manual setup steps Mac-friendly as well: The *-cpu packages from pytorch only exist for Linux: conda install -y pytorch-cpu torchvision-cpu ignite -c pytorch

For OSX, we need the regular packages: conda install -y pytorch torchvision ignite -c pytorch

sotte commented

💯 👍

I updated the README. I'll leave the issue open for now in case ppl have the same problem.

sotte commented

I'll close it for now.