Provide single magic invocation for cuml.accel and cudf.pandas

[csadorf]

Currently, users need to use two separate magic commands to enable both cuml.accel and cudf.pandas in Jupyter notebooks:

%load_ext cudf.pandas
from cuml.experimental.accel import install
install()

This feature request aims to simplify this process by providing a single magic invocation that will enable both accelerators at once.

Motivation

• Improve user experience by reducing the number of steps needed to enable acceleration
• Make it easier for users to get started with RAPIDS acceleration
• Reduce potential confusion about the correct order of operations
• Align with the zero-code-change philosophy of both projects

Implementation Details

The new magic command should:

1. Load the cudf.pandas extension
2. Install the cuml.accel accelerator
3. Provide appropriate feedback to the user about what has been enabled

Related Components

• cuml.accel: Zero code change acceleration for Scikit-Learn
• cudf.pandas: Zero code change acceleration for Pandas
• Jupyter magic command system

Next Steps

1. Design the new magic command interface
2. Implement the combined magic command
3. Add appropriate documentation
4. Add tests for the new functionality
5. Update the FAQ to reflect the new simplified usage

References

• Current documentation: Zero Code Change Acceleration
• Remove or modify promise about single magic invocation #6688