complexa-sweep

Executing a cartesian-product parameter sweep generates multiple inference configurations from defined hyperparameter axes and orchestrates pipeline execution across GPUs. It automatically aggregates post-run success metrics and ranks design configurations based on Pareto frontier analysis.

Pareto frontier analysis for binder design evaluates generated configurations to find the optimal balance between wall-clock time and success rate. It ranks parameter sets automatically, helping identify the most efficient beam width and nsteps for the pipeline.

Yes, you can run multi-GPU sharding for Proteina-Complexa generative model inference. The skill handles automated orchestration by looping design pipelines over generated configurations and distributing the execution across available single or multiple GPUs.

Yes, you need Hydra-based configuration management to run hyperparameter optimization with this skill. It relies on Hydra to manage the generation of inference configurations and execute the cartesian-product sweeps over the design pipelines.

To aggregate success metrics after a cartesian-product sweep, the skill performs post-run collection and ranks the design configurations. It automatically evaluates the results based on success rate and Pareto frontier analysis to summarize pipeline performance.