complexa-sweep

Run cartesian-product parameter sweeps over Proteina-Complexa design pipelines. Pick or author a sweeper YAML in configs/sweeps/, expand it to N inference configs with script_utils/generate_inference_configs.py, loop complexa design over those configs, then aggregate per-config success metrics into a ranked summary CSV plus a manifest.

Important: The complexa design CLI does NOT accept --sweeper directly. Sweeps are driven by script_utils/generate_inference_configs.py, which writes one configs/inference_configs/inf_{idx}_{run_name}.yaml per sweep combination. You then loop complexa design over those generated configs.

What this skill enables

• Pick an existing sweeper YAML from configs/sweeps/ (beam_width, bb_ca_temperature, search_replicas, example).
• Author a new sweeper YAML with arbitrary dot-notation axes (cartesian product).
• Generate N inference + evaluation config pairs with script_utils/generate_inference_configs.py.
• Loop complexa design over the generated configs (one at a time on a single GPU, or in parallel on a multi-GPU host by sharding the config list across CUDA_VISIBLE_DEVICES).
• Walk per-config output directories and parse the analyze-step CSV from each.
• Emit sweep_summary.csv (one row per config: axis values + success rate + mean iPAE + diversity) and sweep_manifest.json.
• Identify the best config by success rate and the Pareto frontier (wall-clock vs success).

Step 1: Pre-flight