Getting started with Onepanel

Understanding Onepanel

Onepanel is a cloud platform for running machine learning code, the platform handles the environment so you can focus on the actual project. Onepanel bills jobs by the second, which is great because you don’t need to rent a whole machine for an entire month, which might sit idle most of the time, which is way more expensive. Just remember to shutdown a job when you’re done!

Vocabulary

• Projects : Projects are the highest level container and are the first thing you will need to create to get started. Projects can also be shared with other users, see Collaboration with Others for more information.
• Workspace: Workspaces is an interactive environment for developing and training deep learning models on Onepanel. You can run Jupyter notebooks, Python scripts, access the terminal, and much more. Even better, you can seamlessly toggle between CPU and GPU machines while you’re working, right when you’re ready for that extra computing power. All the files and data in a Workspace are preserved across sessions, just like your local computer.
• Jobs: Jobs are instances of your project that is being run in Onepanel. They are containers that come pre-installed with all the dependencies needed for the project, including GPU drivers etc. (if you are running an instance with GPU enabled).
• Datasets: Onepanel separates your project files from your data, because your data is likely to stay the same whereas the machine learning model changes constantly. By uploading datasets separately, you save a lot of time uploading large files, and you could reuse the same dataset in multiple projects.
• Computer Vision Annotation Toolkit: CVAT is interactive video and image annotation tool for computer vision. Onepanel has added CVAT as an environment into their platform and a running demo of CVAT can be accessed at CVAT Public Demo.

More about Jobs and Workspace

When an instance of a job is created, imagine as if the project is being run in a new virtual machine (even though it’s not). In this “virtual machine”, there are three important directories you need to know about:

• /onepanel/output: When you run a Workspace or a Job on Onepanel, the output files of the Workspace / Job are stored in this directory.
• /onepanel/input: Whenever you mount a dataset, it will always go here. For example, if you run $ onepanel download user/datasets/dataset_name/1, it would be available at/onepanel/input/datasets/dataset_name
• /onepanel/code/: The Scripts for each Workspace / Job is stored in this directory.

Create a new project

To create a Project, click on the “Create Project” button:

Creating a Onepanel project from GitHub:

You can easily clone a project from GitHub by copying the GitHub URL from the GitHub repository:

Then pasting it into the GitHub URL

Creating Projects with CLI

To initialize an existing directory as a project:

onepanel projects init

To create a new project and its corresponding folder:

onepanel projects create <project-name>

Creating a Workspace is very easy, once you are in a project, click “Workspaces” and then the “Create” button.

Once a workspace is launched, you can easily switch between machine types by clicking the edit button next to the machine type:

Switching between applications in workspaces

You can easily toggle between different applications within a workspace using the application switcher drop-down:

Deleting workspaces

You can terminate workspaces by pressing the ‘DELETE’ button on the side panel of each individual workspace container.

NOTE: Once you delete a workspace, everything stored in the workspace will be deleted.