As you progress in your Picsellia journey, you use your annotated Datasets in your trainings and track your Experiments so all of your artifacts such as checkpoints and config are stored in it, you might want to make your process reproducible so everyone in your organization can leverage previous work seamlessly.
To do this, you will have to "register" your experiments so they become available in your Model Registry, you will then be able to attach Docker images to them, or even Colab notebook links, so they become usable everywhere.
Here is what you'll learn during this tutorial :
- A list of what can be possibly registered in a model
- How to register everything you need in your Models
- How to use a Model from your Registry as a base for an Experiment
- Export an Experiment as a Model
- Package the script of your Experiment into a reusable, GPU enabled, Docker image
Updated about 1 month ago