Conda
- Giang Rudderham
- Cody Johnson
On this page:
Conda is a tool commonly used to manage dependencies and projects. You can create a conda environment for each project and install the packages that you need in that conda environment. If you make changes to one of your conda environments, your other environments are not affected. For more information, please refer to the Conda User Guide.
The following examples illustrate common conda tasks using JupyterLab in IDAS.
Anaconda updated its Terms of Service in 2024.
We encourage all IDAS users to read the following FAQs regarding Anaconda’s updated Terms of Service.
For more details, please see:
Example: Creating a conda environment with a Python kernel
1. First, log in to the IDAS research instance with Python. If you are a student in a class that uses Python in IDAS, follow the instructions here to access your class instance.
2a. In JupyterLab, click the Terminal tile under Other to start a Terminal session.
2b. If this is your first time using conda in the Terminal in IDAS, initialize your shell by running the following command in the Terminal.
This command only needs to be run once. It will create a .bashrc file in your IDAS home directory if a .bashrc doesn’t already exist.
conda init bash
2c. Close the Terminal by clicking the x in the upper right corner of the Terminal window.
2d. Click the Terminal tile to open the Terminal again.
2e. Run the following command in the Terminal. This will source the .bashrc file that was just created.
source ~/.bashrcThe command prompt in your Terminal will change to indicate the active environment, which is the base environment by default.
(base) HawkID@idas-research-HawkID:~$
Every time you open a new Terminal session, you’ll have to run source ~/.bashrc before using a conda environment.
3. In Terminal, the conda create command can be used to create a new conda environment.
In this example, we will create a conda environment with Python 3.8 and name the environment py38.
conda create -n py38 python=3.8When prompted Proceed ([y]/n)?, press y to proceed.
Additional notes:
The conda create command can be modified to fit your needs. Below are a few additional examples:
Additional example 1: Install packages when creating an environment:
conda create -n another-env python=3.8 numpy requestsAdditional example 2: Specify the versions of the packages:
conda create -n another-env python=3.7 numpy=1.16.1 requests=2.19.1Additional example 3: Create a conda environment from an environment.yml file:
conda env create -f environment.yml
For more examples of creating conda environments, please see the section "Managing environments" in the Conda User Guide.
4. Once this conda environment has been created, we can activate it.
conda activate py38
5. The command prompt in your Terminal will change to indicate the active environment.
(py38) hawkid@idas-research-hawkid:~$
6. Next, we create a kernel in order to use Jupyter Notebook with this conda environment. Install the IPython kernel:
conda install ipykernelWhen prompted Proceed ([y]/n)?, press y to proceed.
7. Now install a kernel in this environment:
python -m ipykernel install --user --name py38 --display-name "Python Project Vis"The value for --name is used by Jupyter internally. Any existing kernel with the same --name value will be overwritten. The --display-name will be displayed in the Notebook menu in the JupyterLab Launcher page.
8. Go back to the JupyterLab Launcher page by pressing Ctrl+Shift+L (Windows) or Cmd+Shift+L (Mac). Under Notebook, a new option for your kernel will now be available. In this example screenshot, the "Python Project Vis" kernel was just installed and now became available to use.
Click on that new option to start a notebook. In that notebook, you can use the packages that you installed in the conda environment.
9. To install Python packages to this conda environment using a Jupyter notebook:
a. Start a new Jupyter notebook by selecting the Python kernel you just installed. In this example, that is the "Python Project Vis" kernel in the previous step.
b. In a cell in this new notebook, use the %conda magic that is built into IPython to install packages within the current kernel. For example, install the "pandas" package:
c. Then we can load and use the package as usual:
10. We can also install Python packages to this conda environment using the Terminal:
conda install --name py38 pandas
For more information on searching for and installing packages, see the section “Managing packages” in the Conda User Guide.
Example: Creating a conda environment with an R kernel
Installing conda
If you are using the R instances on IDAS, you will need to install conda before using it.
If you are using the Python or Julia instances on IDAS, you do not need to install conda.
In this example, we will install conda using the Miniconda installer. Users can choose to install a different installer, such as Miniforge. For more details, please see Installing conda in the Conda User Guide.
1. First, log in to the IDAS research instance and select an R instance. If you are a student in a class that uses IDAS, follow the instructions here to access your class instance.
2. In JupyterLab, click the Terminal tile under Other to start a Terminal session.
3. From here, we can follow steps for Linux, under “Quickstart install instructions” in the Miniconda documentation.
Create a new directory, miniconda3, in your home directory:
mkdir -p ~/miniconda3
4. Download the Miniconda installation script and save it as miniconda.sh in the miconda3 directory:
wget https://repo.anaconda.com/miniconda/Miniconda3-latest-Linux-x86_64.sh -O ~/miniconda3/miniconda.sh
5. Run the miniconda.sh installation script in silent mode using bash:
bash ~/miniconda3/miniconda.sh -b -u -p ~/miniconda3
6. Remove the miniconda.sh installation script after the installation is complete:
rm ~/miniconda3/miniconda.sh
7. If this is your first time using conda in the Terminal in IDAS, initialize conda on the bash shell:
This command only needs to be run once. It will create a .bashrc file in your IDAS home directory if a .bashrc doesn’t already exist.
conda init bash
8. Close the Terminal by clicking the x in the upper right corner of the Terminal window.
9. In JupyterLab, click the Terminal tile to open the Terminal again.
10. Run the following to activate miniconda:
source ~/miniconda3/bin/activate
The command prompt in your Terminal will change to indicate the active conda environment, which is the base environment by default.
(base) HawkID@idas-research-HawkID:~$
Every time you open a new Terminal session in an R instance, you’ll have to run source ~/miniconda3/bin/activatesource ~/.bashrc before using a conda environment.
11. Test the install. This will list all conda environments installed in your IDAS account:
conda env listCreating a conda environment with an R kernel
1. In Terminal, the conda create command can be used to create a new conda environment.
In this example, we will create a conda environment with R and name the environment r-env.
conda create -n r-env r-essentials r-baseWhen prompted Proceed ([y]/n)?, press y to proceed.
Additional notes:
You can create an R conda environment with just the
r-basepackage, like so:conda create -n r-env r-base
However, the
r-essentialspackage includes approximately 80 of the most popular packages for R, so it is convenient to install ther-essentialsbundle rather than installing each individual package later.Many R packages are available to install with conda, but not all. The latest index of R packages built by Anaconda, Inc. can be found on Anaconda Cloud or at http://repo.anaconda.com/pkgs/r/
For more information about using R with conda, please see Using R language with Anaconda and R language packages for Anaconda in the Anaconda documentation.
For more examples of creating conda environments, please see "Managing environments" in the Conda User Guide.
2. Once this conda environment is created, we can activate it.
conda activate r-env
3. The command prompt in your Terminal will change to indicate the active environment.
(r-env) hawkid@idas-research-hawkid:~$
4. Next, we create a kernel in order to use Jupyter Notebook with this conda environment. We will install a kernel through IRkernel. The r-irkernel package was already installed with the r-essentials bundle earlier. You can check to make sure you have r-irkernel in your conda environment:
(r-env) hawkid@idas-research-hawkid:~$ conda list irkernel
# packages in environment at /home/hawkid/.conda/envs/r-env:
#
# Name Version Build Channel
r-irkernel 0.8.15 r36_0 defaults
5. Now install a kernel in R:
# launch R in Terminal
R
> library(IRkernel)
> IRkernel::installspec(name = "ir361", displayname = "R 3.6.1")The value for name is used by Jupyter internally. Any existing kernel with the same name value will be overwritten. The displayname will be displayed in the Notebook menu in JupyterLab.
Note that we specified the R version (3.6.1) in the name and displayname. This helps remind us what R version we are using, especially if we have multiple conda environments and kernels.
6. Go back to the JupyterLab Launcher page by pressing Ctrl+Shift+L (Windows) or Cmd+Shift+L (Mac). Under Notebook, a new option for your kernel will now be available, with the name "R 3.6.1"
Click on that new option to start a notebook. In that notebook, you can use the packages that you installed in the conda environment.
7. To install R packages in this conda environment, in Terminal:
conda install --name r-env r-abind
8. Then you can load and use the package in Jupyter Notebook as usual. If you have an open notebook, you can use the package in the notebook right after you install the package in Terminal. You don't need to refresh the notebook.
9. In general, any packages that are installed in the conda environment will be available to use in the notebook. To view a list of installed packages in the conda environment, type in Terminal:
conda list -n r-env
10. To view a list of installed packages in Jupyter notebook, use the installed.packages() function:
Other useful conda commands
The following commands may be useful to manage conda environments. Type the following commands in Terminal. For more information, please refer to the "Managing environments" and "Managing packages" sections in the Conda User Guide.
List all of your environments. In the output, your current environment will be marked with an asterisk (*).
conda info --envsor
conda env list
To activate a conda environment that was previously created, type (without the angle brackets):
conda activate <environment-name>
Deactivate a conda environment once you are finished working with it:
conda deactivate
Or, to return to the base environment, type (with no environment specified):
conda activate
View a list of packages installed in an environment:
If the environment is not activated, type (without the angle brackets):
conda list -n <environment-name>
If the environment was already activated with conda activate earlier:
conda list
To see if a specific package is installed in an environment, type (without the angle brackets):
conda list -n <environment-name> <package-name>
Or, if the environment was already activated with conda activate earlier, type (without the angle brackets):
conda list <package-name>
To install a package in an environment:
conda install --name <environment-name> <package-name>
Remove an environment:
conda remove --name <environment-name> --allor
conda env remove --name <environment-name>
Using pip in a conda environment
Please note that issues may arise when using pip and conda together. The "Using pip in an environment" section in the Conda User Guide outlines best practices for using pip in a conda environment.
Contact
If you have any questions or comments, please contact research-computing@uiowa.edu.