Creating 'donkey' environment from ubuntu.yml takes forever

[nabanita-]

I am trying to setup the donkey environment in my desktop (Ubuntu18.04.6 LTS) following the instructions here. But solving environment is taking forever.

I also tried by setting channel_priority as strict (conda config --set channel_priority strict) but it resulted in a huge number of conflicts, so I ditched that route.

Is there something that I am missing? I am using the yaml file in the main branch of this github repository.

Here is the yaml file content for easy reference:

name: donkey

channels:
  - defaults
  - conda-forge
  - pytorch
  
dependencies:
  - python=3.7
  - h5py
  - pillow
  - opencv
  - matplotlib
  - tornado
  - docopt
  - pandas
  - pylint
  - pytest
  - pytest-cov
  - codecov
  - pip
  - imgaug
  - progress
  - moviepy
  - paho-mqtt
  - PrettyTable
  - pyfiglet
  - mypy
  - pytorch=1.7.1
  - torchvision
  - torchaudio
  - pytorch-lightning
  - numpy
  - psutil
  - kivy=2.0.0
  - plotly
  - pyyaml
  - tensorflow=2.2.0
  - fastai
  - pip:
    - git+https://github.com/autorope/keras-vis.git
    - simple-pid
    - opencv-python-headless

[c1505]

@nabanita- I don't know why the conda install takes so long, but you aren't alone in that experience of it taking a very long time or hanging. I think mamba should be the recommended install . There is a PR for docs changes for that here that I reviewed autorope/donkeydocs#37 . I don't have write access to merge it though.

Just noticing that the docs you are using are on a separate domain and use a separate github repository than what I have used https://docs.donkeycar.com/ . Both the code and docs have not been updated in about 9 months for the robocarstore org. I am not sure who has access to that repo and how much they are in the donkeycar discord. I would suggest using this repository for code and the previously mentioned link for docs.

[Ezward]

@nabanita- I have an experimental branch that has a bash script for installing on unbuntu that only uses pip, so conda is not required.. Once you clone/checkout the branch, the run ./install.ubuntu.pip.sh to activate the environments prior to running donkey commands source env/bin/activate https://github.com/autorope/donkeycar/tree/990-ubuntu-install-script

[victorjnpires]

@nabanita- I've also had issues with conda taking forever on conflict resolution, and mamba was the solution. This is my simplified yml:

name: donkey

channels:
  - conda-forge
  - defaults

dependencies:
  - python=3.7
  - h5py
  - pillow
  - opencv
  - matplotlib
  - tornado
  - docopt
  - pandas
  - pylint
  - pytest
  - pytest-cov
  - codecov
  - pip
  - imgaug
  - progress
  - moviepy
  - paho-mqtt
  - PrettyTable
  - pyfiglet
  - mypy
  - numpy
  - psutil
  - kivy=2
  - plotly
  - pyyaml
  - tensorflow-gpu=2.7
  - pip:
    - git+https://github.com/autorope/keras-vis.git
    - simple-pid
    - opencv-python-headless

[nabanita-]

@nabanita- I don't know why the conda install takes so long, but you aren't alone in that experience of it taking a very long time or hanging. I think mamba should be the recommended install . There is a PR for docs changes for that here that I reviewed autorope/donkeydocs#37 . I don't have write access to merge it though.

Just noticing that the docs you are using are on a separate domain and use a separate github repository than what I have used https://docs.donkeycar.com/ . Both the code and docs have not been updated in about 9 months for the robocarstore org. I am not sure who has access to that repo and how much they are in the donkeycar discord. I would suggest using this repository for code and the previously mentioned link for docs.

I hadn't noticed the difference in domains! Thank you, I will keep that in mind. Solved my problem ^.^

[DocGarbanzo]

The PR autorope/donkeydocs#37 got merged, so closing this now.