This is a Flask-based web server for performing Hamiltonian predictions, based on a refactored version of Yang Zhong's HamGNN 2.0.
It's designed to provide fast and efficient predictions by keeping the HamGNN model continuously loaded in memory. This eliminates the "cold start" time typically required for script-based predictions.
Clients can send graph data (or a path to the data) via an HTTP request and receive the Hamiltonian prediction in response.
This platform has been refactored into a microservices architecture to handle large-scale, automated workflows. The system is managed by a central Orchestrator Server that dispatches jobs to three distinct services: an openmxServerAPI for preprocessing, the hot-started HamGNNServerAPI for core predictions, and a postprocessAPI for tasks like band structure calculations. The entire workflow is managed asynchronously using a Celery task queue, enabling robust and efficient processing of large batches of jobs.
support postprocess with pyatb