RocketML Data Science Workbench (DSW) is a machine learning (ML) platform to enable Data scientists and ML engineers across your organization to easily build and deploy ML solutions at scale.
RocketML can be accessed via a browser at an URL given out by your IT department. Authentication into RocketML system is through single sign on setup implemented by your IT department.
Once user is logged into RocketML system. They will see a home page as shown below:
We advice users to start by clicking on the Create Project button
Data Scientist's work is done in a project, which is a collection of workspaces and project files.
For every user, workbench creates a folder in your corporate S3 bucket. For every project created by a user, DSW creates a project folder in the user folder. Project folder name can be accessed from Projects Details page. All workspaces created in this project will have read-write access to project files folder. Any data, code, results can be stored in project files folder for future use.
Workspaces are single-node and multi-node CPU/GPU clusters with read-only access to data sources and read-write access to project files.
- c5.large
- c5.xlarge
- c5.2xlarge
- c5.4xlarge
- c5.9xlarge
- c5.12xlarge
- c5.18xlarge
- c5.24xlarge
- p3.2xlarge
- p3.8xlarge
- p3.16xlarge
- g4dn.xlarge
- g4dn.2xlarge
- g4dn.4xlarge
- g4dn.8xlarge
- g4dn.16xlarge
A convenient Data Explorer functionality is offerred on DSW to view available data sources, datasets.
Corporate data sources are populated by Admins of DSW application with IT departments help and consent. These data sources and datasets can be viewed and accessed via DSW application within a Workspace. The steps to access this data for machine learning and data science is described below (later).
Mainly two types of Data Sources are available
- Object Stores (S3 buckets) for files (both structured and unstructured)
- Databases (structured)
Data from these sources can be ingested directly from Workspaces of any project.
RocketML provides a few popular public datasets as seed data for DSW users to get started quickly. These datasets are read only for the users. However admins can disable access to these.
Data Elsewhere
Data elsewhere
Open a terminal and install the package at the terminal
pip install --user <package name>
Open a terminal and install the package at the terminal
sudo /opt/conda/bin/conda install -c <channel> <package name>
Having trouble with Data Science Workbench? or Contact us and we’ll help you sort it out.