This repository contains standarized Croissant Task examples for three different use cases: a data processing task (DICOM to NIfTI conversion for medical images), a Benchmark, loosely based on the MLCommons AILuminate Benchmark and another benchmark loosely based on the MLCommons MLPerf Benchmark. All examples are hypothetical and do NOT constitute complete use cases. For all use cases, both a Croissant Task Problemn, a Croissant Task Solution and a Croissant Task Instance example are provided. Someone can define a Croissant Task Problem, leaving in specifications for data, assets and/or implementations that must be provided as part of corresponding Task Solutions. This Task Problem may then be shared with a large community, from which other users can generate their own Croissant Task Problem solutions. Data, assets and/or implementations from the Solutions should conform to the specification of the Croissant Task Problem, even though they will differ in each Solution. Finally, when executing a given solution on real data, a Task Instance is generated as a record of that execution.
NOTE Currently, only the data processing example is up-to-date with the latest developments of the Croissant Task format.