Ctrl+K

RAPIDS on Google Colab#

Overview#

This guide is broken into two sections:

  1. RAPIDS Quick Install - applicable for most users

  2. RAPIDS Custom Setup Instructions - step by step set up instructions covering the must haves for when a user needs to adapt instance to their workflows

In both sections, will be installing RAPIDS on colab using pip or conda. Here are the differences between the two installation methods

  • Pip installation allows users to install cuDF, cuML, cuGraph, and cuSpatial stable versions in a few minutes (1/5 ease of install)

  • Conda installation installs the complete, customized RAPIDS library package (such as installing stable or nightly) however, it can take around 15 minutes to install and has a couple of break points requiring the user to manually continue the installation (2/5 ease of install)

RAPIDS install on Colab strives to be an “always working” solution, and sometimes will pin RAPIDS versions to ensure compatiblity.

Section 1: RAPIDS Quick Install#

Section 2: User Customizable RAPIDS Install Instructions#

1. Launch notebook#

To get started in Google Colab, click File at the top toolbar to Create new or Upload existing notebook

2. Set the Runtime#

Click the Runtime dropdown and select Change Runtime Type

Screenshot of create runtime and runtime type

Choose GPU for Hardware Accelerator

Screenshot of gpu for hardware accelerator

3. Check GPU type#

Check the output of !nvidia-smi to make sure you’ve been allocated a Rapids Compatible GPU, i.e [Tesla T4, P4, or P100].

Screenshot of nvidia-smi

4. Install RAPIDS on Colab#

You can install RAPIDS using

  1. pip

  2. conda

4.1. Pip#

Checks GPU compatibility with RAPIDS, then installs the latest stable versions of RAPIDSAI’s core libraries (cuDF, cuML, cuGraph, and xgboost) using pip.

# Colab warns and provides remediation steps if the GPUs is not compatible with RAPIDS.

!git clone https://github.com/rapidsai/rapidsai-csp-utils.git
!python rapidsai-csp-utils/colab/pip-install.py

4.2. Conda#

If you need to install any RAPIDS Extended libraries or the nightly version, you can use the RAPIDS Conda Colab Template notebook and install via conda.

  1. Create and run a cell with the code below to update Colab’s gcc. Ignore all Colab alerts.

!bash rapidsai-csp-utils/colab/update_gcc.sh
import os
os._exit(00)

  1. Create and run a cell with the code below to install conda on Colab. Ignore all Colab alerts.

import condacolab
condacolab.install()

[Optional] Run the test and conda install test cell.

import condacolab
condacolab.check()

  1. Before running the RAPIDS install cell, you can change the installation type between stable and nightly. All disclaimers around nightly installs apply.

  2. Run the rest of the cells to complete the installation of RAPIDS on Colab.

!python rapidsai-csp-utils/colab/install_rapids.py stable # example runs stable
import os
os.environ['NUMBAPRO_NVVM'] = '/usr/local/cuda/nvvm/lib64/libnvvm.so'
os.environ['NUMBAPRO_LIBDEVICE'] = '/usr/local/cuda/nvvm/libdevice/'
os.environ['CONDA_PREFIX'] = '/usr/local'

5. Test Rapids#

import cudf

gdf = cudf.DataFrame({"a":[1,2,3],"b":[4,5,6]})
gdf
    a   b
0   1   4
1   2   5
2   3   6

6. Next steps#

Check out this guide for an overview of how to access and work with your own datasets in Colab.

For more RAPIDS examples, check out our RAPIDS notebooks and notebooks-contrib repos

On this page