COMPSs/PyCOMPSs and disLib are designed to ease the development of ML/AI applications targeting the Cloud-Edge-IoT Continuum. COMPSs is a task-based programming model that orchestrates the execution of such tasks in a serverless manner on top of any distributed platform. PyCOMPSs is an enhanced version of the programming model exploiting the benefits of the Python programming language. DisLib provides application developers with a set of built-in AI algorithms that leverages on PyCOMPSs to distribute computation.
Open Source, Apache 2.0
PyCOMPSs is part of the design tools layer.
The application developer provides a sequential Python script whose functions are annotated through decorators; these annotations are used by the runtime to run those parts of code as asynchronous parallel tasks code. When executed, the user code (the annotated part) is intercepted and analysed by the runtime, generating an execution graph.