Model Incorporation

You can publish a machine learning model to the Ersilia Model Hub by wrapping it into the standardized eos-template format with required metadata, tests, and example artifacts. The process involves cloning, verifying, assigning an ID, generating files, and reporting.

The eos-template format is used to wrap published machine learning models for standardized deployment and discovery within the Ersilia Model Hub. It provides a consistent structure for metadata, tests, and examples to ensure models are easily accessible.

To wrap a published ML model, you provide an accessible repository URL and optional publication details like a paper URL. The workflow then clones the repository, verifies it by running, assigns a model ID, generates template files, and outputs them to a specified directory.

You need an accessible source model repository URL and a configurable output directory to begin wrapping. Optional inputs include a publication paper URL and a pre-assigned model-id to further annotate the wrapped model during the generation phase.

Yes, the model incorporation workflow enforces a multi-phase process that includes verifying the model by running it and performing a final test and report phase. This ensures the wrapped ML model is validated with generated metadata and examples before publishing.

The conversion process requires an accessible repository URL to clone and analyze, meaning inaccessible or private repositories cannot be wrapped without prior access. It also strictly enforces a multi-phase workflow, so skipping the verification or testing phases is not supported.