Skip to main content

Conda Quick Reference

Creating new Conda environment​

To create a new Conda environment, use the conda create command. Here's the basic syntax:

conda create --name <env_name> python=<version>

Example:​

conda create --name myenv python=3.10

This will:

  • Create a new environment named myenv
  • Install Python 3.10 in it

After creating the environment:​

Activate it with:

conda activate myenv

To deactivate it and return to the base environment:

conda deactivate

Adding packages during environment creation​

conda create --name myenv python=3.10 numpy pandas

Installing dependency in an active conda environment​

To install a dependency (package) in an active Conda environment, use the following command:

Method 1: Install a Package in the Active Environment​

conda install <package_name>

Example:

conda install numpy

Method 2: Install from a Specific Conda Channel​

conda install -c <channel_name> <package_name>

Example (install tensorflow from conda-forge):

conda install -c conda-forge tensorflow

Method 3: Install Using Pip (if the package isn't available via Conda)​

pip install <package_name>

Example:

pip install openai

Best Practices​

  1. Check if the environment is activated:

    • Run conda activate <env_name> (if not already active).
    • Verify using: 
      conda env list
      (The active environment has an * next to it.)

  2. Specify a version (recommended for reproducibility):

    conda install numpy=1.23.5

  3. Install multiple packages at once:

    conda install pandas scikit-learn matplotlib

  4. Export environment for sharing (after installing dependencies):

    conda env export > environment.yml

Troubleshooting​

  • If Conda doesn’t find the package, try:
    • Searching first: conda search <package_name>
    • Using conda-forge: conda install -c conda-forge <package_name>
    • Falling back to pip (but prefer Conda when possible to avoid conflicts).

Check Installed Dependency Versions in Conda​

To determine the versions of Python dependencies installed in a running Conda environment, you have several options. Here are the most common methods:

1. Using conda list​

Run this command from within your active Conda environment:

conda list

This will output a table of all installed packages and their versions, like:

# packages in environment at /path/to/env:
#
# Name                    Version                   Build  Channel
numpy                    1.24.3           py310h5f9c477_0
pandas                   1.5.3            py310h1c4a608_0

2. Using pip list (for pip-installed packages)​

If some packages were installed via pip (even inside the Conda environment), you can also run:

pip list

This will show you all packages installed via pip, which may not appear in conda list.

3. Using environment YAML file​

You can export the environment to a YAML file to view all packages and versions:

conda env export > environment.yml

Then open environment.yml to see a list like:

dependencies:
  - python=3.10.4
  - numpy=1.24.3
  - pandas=1.5.3
  - pip:
    - some-pip-package==0.1.0

This is especially useful for sharing or recreating environments.

Pro Tip: Show only name and version​

To list just the package names and versions using conda list, you can combine with awk or Python parsing:

conda list | awk '{print $1, $2}'

Removing unnecessary environments​

Conda virtual environments can consume a significant amount of disk space, especially if you've created many of them over time. Each environment typically has its own copy of Python, libraries, and dependencies, which adds up quickly.

Here's how you can free up space:​

1. List Conda Environments​

First, see which environments exist:

conda env list

or

conda info --envs

2. Remove Unused Environments​

If you spot environments you no longer need, remove them with:

conda remove --name myenv --all

Replace myenv with the name of the environment you want to delete.

3. Clear Conda Package Cache​

Conda keeps a cache of downloaded packages. This can grow large over time. You can clear it using:

conda clean --all

You’ll be prompted for confirmation. This removes:

  • Unused packages
  • Tarballs
  • Index cache
  • Unused caches

To do it without prompt:

conda clean --all --yes

4. Manually Check the envs Directory​

Sometimes orphaned environments or large installations live in:

~/miniconda3/envs

or

~/anaconda3/envs

You can manually inspect and delete any large folders corresponding to unused environments only if you're sure they're not needed.

Quick Tip​

Here's a one-liner command that lists all Conda environments and shows their disk usage, sorted by size:

du -sh $(conda env list | awk '{print $NF}' | grep -E '^/|^/Users|^/opt') 2>/dev/null | sort -hr

What this does:​

  • conda env list: Lists all environments.
  • awk '{print $NF}': Extracts the path from each line.
  • grep -E '^/|^/Users|^/opt': Filters only actual environment paths (skips the header).
  • du -sh: Shows disk usage in human-readable format.
  • sort -hr: Sorts output from largest to smallest.

Example Output:​

2.3G    /opt/anaconda3/envs/ai_projects
1.9G    /opt/anaconda3
1.4G    /Users/abc/anaconda3/envs/mlflow-d758da66ae6b5dac8a2b69972cb371638d4cc6f3
...

This helps you identify which environments are consuming the most space so you can clean up the largest, unused ones.

Delete complete Anaconda and reinstall it​

You can delete and reinstall Anaconda to reclaim space and start fresh. It's a clean way to remove all accumulated bloat (including unused environments, cached packages, etc.).

⚠️ Before you proceed:​

  • You will lose all Conda environments, unless you back them up.
  • If you've installed any custom packages in the base environment, they will be gone.
  • You may want to export environments you care about (see below).

Step 1: (Optional) Backup environments​

For any environment you want to keep, export it:

conda activate myenv
conda env export > myenv.yml

You can later recreate it with:

conda env create -f myenv.yml

Step 2: Remove Anaconda completely​

Run the following commands:

rm -rf /opt/anaconda3

Then, if you added Conda to your shell config (e.g., .bashrc, .zshrc, etc.), remove or comment out the lines like:

# >>> conda initialize >>>
# ... bunch of lines ...
# <<< conda initialize <<<

You can check and clean it with:

nano ~/.zshrc   # or ~/.bashrc, depending on your shell

Then restart your shell:

exec $SHELL

Step 3: Install a fresh version​

Download from: https://docs.conda.io/en/latest/miniconda.html Then run:

bash Miniconda3-latest-MacOSX-x86_64.sh

Option B: Anaconda (if you need everything pre-installed)​

Download from: https://www.anaconda.com/products/distribution Then run:

bash Anaconda3-latest-MacOSX-x86_64.sh

After install: verify with​

conda --version
conda info --envs