Multi-job runs

When you run large evaluations across several SLURM jobs, each job writes its own per_example_metrics_job{N}.csv. Once they all finish, concatenate them by passing the metrics directory as the first positional argument.

python -m rfproteina.inference_utils.aggregate_metrics ./inference/my_run/metrics/

That concatenates the per example job CSVs, writes aggregated_metrics.csv and per_task_metrics.csv, and moves the individual job files into individual_job_results/. Pass --output_dir <dir> to write elsewhere. launch_generate.sh runs this for you as a dependent SLURM job.

Pass --cluster_with_foldseek to also run structural diversity clustering, described in FoldSeek diversity clustering.