Creating a Custom Conda Environment
Create a custom conda environments with a conda compatible environment file (environment.yaml) using the odsc conda create command.
By default, the create option also installs additional libraries to ensure that the conda environment is compatible with JupyterLab and the OCI services. The most important libraries that are added are oci, ipykernel, jupyterlab, nb_conda_kernels, and cx_oracle. You can overwrite this default behavior by adding the optional argument --empty, -e then no extra libraries are added to your conda environment.
We strongly recommend that you don't use the --empty argument. Because the conda environment that you create might not be accessible as a kernel in JupyterLab.
You can create a conda environment from a conda compatible environment file with:
odsc conda create -n <name> -v <version> -f <environment.yaml> If you want to create the environment from an empty base, use the
--empty optional argument:
odsc conda create -n <name> -v <version> -f <environment.yaml> --empty Using the Anaconda Technology on OCI
To start using Anaconda in OCI Data Science, build or customize your own conda environment.
Following the partnership announcement between Oracle and Anaconda , means that while running workloads in OCI, you can use Anaconda while running workloads in OCI. You can use the Anaconda repository of packages without buying a separate license from Anaconda. Anaconda is the standard distribution channel for open source software in machine learning and AI services.
You can use the Anaconda repository of packages by adding anaconda or main as the first channel listed in a conda compatible environment file (environment.yaml).
This sample environment.yaml file prioritizes anaconda over the community driven conda-forge channel:
channels:
- anaconda
- conda-forge
dependencies:
- keras
- tensorflow
After you've created the conda environment, you can inspect the list of packages that were installed in the conda environment by running this command in a terminal window or in a notebook running inside the conda environment kernel:
conda list Following is a sample output of the conda list command:
Name Version Build Channel
absl-py 0.15.0 pyhd3eb1b0_0 anaconda
aiohttp 3.8.1 py38h7f8727e_1 anaconda
aiosignal 1.2.0 pyhd3eb1b0_0 anaconda
argon2-cffi 21.3.0 pyhd3eb1b0_0 anaconda
argon2-cffi-bindings 21.2.0 py38h7f8727e_0 anaconda
arrow 1.2.3 pypi_0 pypi
astor 0.8.1 py38h06a4308_0 anaconda
The channel column in the response lists the source channel of the Python library that was installed in the environment. In this example, you can see that most packages were installed from anaconda.
For more details about the Anaconda Repository and why Anaconda is the recommended option to download open source packages, watch how to seamlessly leverage Anaconda on OCI presented by the Oracle Developers.
Oracle is licensed to include packages from Anaconda and make these packages available to OCI customers. Preinstalled packages that are embedded in OCI products and services that you license from Oracle might be used under the terms of the applicable Oracle OCI license agreement or terms of service.
You can use Oracle's cloud-hosted products and services with a preinstalled copy of Conda to access additional packages from Anaconda's repository. This access is under the Anaconda's Terms of Service, except that Oracle OCI customers can use the Anaconda packages for commercial purposes on the OCI platform without obtaining a separate paid license from Anaconda. The packages are only for use as part of our services and don't entitle you to download them to your own infrastructure or to use Anaconda's trademarks. Packages might have their own licenses provided by the package authors.
Create an account on Anaconda Nucleus and to get started using Anaconda.