Understanding Gallery Projects

Projects like batch jobs, notebooks or workspaces available in the gallery are meant to be executable with a single click (in case of batch jobs and workspaces). The motivation behind that is to ensure our users, who want to learn and use published projects have to spend zero time in setting up.

It is also important to know that all projects are already containerized and run inside a docker container. For batch jobs, the container is build with all the conda dependencies installed and will execute the main file with any arguments specified in the batch job config (learn more about ‣). This provides an amazing automation. Imagine checking out a model fine tuning code and all you have to do is just to provide a new dataset and decide the compute type - GPUs and type of GPUs. Everything just works!

Users can also find interesting videos in gallery on any topics within AI and Deep Learning. We are still building the curated content, so keep checking out periodically. If you have any ideas or are interested in publishing content, connect with us at [email protected].

Searching for Projects

On the gallery page, you can search by key words or questions based on topics that interest you. Ploutos search is semantic based and it will find any relevant projects based on your question. The search results will include notebooks, batch jobs, workspaces, videos and even papers. Here is a screenshot of the search:

Navigating a Project

Once you click a search result, it will open that project. If a project is of type notebook, batchjob, workspace, it will have tabs for Summary/Readme, Code and Video. On the project card, you will see Ellipsis, that will have actions like Add to Favorite, Copy to my Space.

• Adding to Favorite will add it your favorite list that you can always see on the Gallery page. There is a toggle on the right next to search for toggling between all projects and only favorited projects.
• Copying the project will copy it your space. This means when you launch the Creator’s App, you will see your own copy of the project. You can then edit and run the project the way you like. It is important to note that copied projects are copied with their runtimes which make it running them very easy since they are preconfigured with all the settings necessary. It is important to now that you may choose to use different compute type like GPU and GPU models and count than the original author who published the project. You can change that in the project settings.

Here is a screenshot of an opened project in the Gallery: