I have been using python venv module for all my virtual environments needs. It is basic and works very well. However since I got my M1 chip MacBook, things changed. A lot of python libraries are not supported yet. Miniforge supports the Rosetta 2 layer – as a result quite a few of the libraries can work again.
To get over the issues, I decided to use community version of miniforge. Miniforge is a community driven version for miniconda and all libraries are downloaded from conda-forge. Since all of these are new for me, I decided to put the basic commands in one place so that I can refer back to them whenever needed.
Installation
I will be using the community driven version of conda. You can download miniforge installer from https://github.com/conda-forge/miniforge. Following list shows all builds for miniforge.
For MacBook with M1 chip, you will only need to run Miniforge3-MacOSX-arm64.sh and follow prompts to install Miniforge.
For this task I will install PyTorch in a virtual environment. This should give an example for how to create and use a fresh virtual conda environment.
Create a Virtual Environment
First let’s do some housekeeping stuff.
$ conda update conda -y Collecting package metadata (current_repodata.json): done Solving environment: done All requested packages already installed. $ conda clean --packages Cache location: There are no unused packages to remove $ conda -V conda 4.10.2 $ conda env list conda environments: # base * /opt/homebrew/Caskroom/miniforge/base
We updated all outdated packages in conda install to latest version. Since conda does not cleanup old packages by default, we have also made sure that we clean them in the next command. The commands following this are informational and prints the version and all installed virtual environments.
Now let’s create a new virtual environment and install pytorch in it. We will use the default version of python installed. Conda allows multiple versions of python to be maintained. We can use following command to list all available versions of python.
$ conda search "^python$" Loading channels: done Name Version Build Channel python 3.8.5 h05baefb_8_cpython conda-forge python 3.8.6 h12cc5a1_1_cpython conda-forge python 3.8.6 h12cc5a1_2_cpython conda-forge python 3.8.6 h12cc5a1_3_cpython conda-forge ::::: and many more :::::
Use the following command to create an empty virtual environment with the default version of python installed.
$ conda create -n pytorch Collecting package metadata (current_repodata.json): done Solving environment: done Package Plan environment location: /opt/homebrew/Caskroom/miniforge/base/envs/pytorch Proceed ([y]/n)? y Preparing transaction: done Verifying transaction: done Executing transaction: done # To activate this environment, use # $ conda activate pytorch # To deactivate an active environment, use # $ conda deactivate $ conda activate pytorch
The last command activates pytorch virtual environment. Next step is to install pytorch.
$ conda install pytorch -y :::: lots of output ::: Preparing transaction: done Verifying transaction: done Executing transaction: done $ conda install notebook -y
We installed pytorch and jupyter notebook above. Let’s run some basic test now.
Running a Basic Test
First we start jupyter notebook
$ conda list packages in environment at /opt/homebrew/Caskroom/miniforge/base/envs/pytorch: # $ jupyter notebook
The first command list all installed packages in pytorch environment. Before starting notebook, we make sure that pytorch and jupyter notebook are installed. Try some pytorch commands and see if it works.
Everything seems to be working. As the next step, we will remove this virtual environment. This is not necessary as we would probably continue to work on this environment.
$ conda deactivate $ conda env remove -n pytorch
Pip vs Conda
| Feature | PIP command | Conda command |
|---|---|---|
| Version | pip -V | conda -V |
| Create Environment | python3 -m venv NAME | conda create -n NAME |
| Activate | source NAME/bin/activate | conda activate NAME |
| Deactivate | deactivate | conda deactivate |
| Install package | pip install PCKG | conda install PCKG |
| Uninstall package | pip uninstall PCKG | conda remove PCKG |
| List packages installed | pip list | conda list |
| Remove venv | rm -rf ./NAME | conda env remove -n NAME |
| Export | pip freeze > rqmt.txt | conda list –export > rqmt.txt |
| Import | pip install -r rqmt.txt | conda create -n NAME –file rqmt.txt |
Conclusion
I started my journey with conda with the list of commands above. Hopefully it helps someone. Ciao for now!