complexa-sweep

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

bash .claude/skills/_shared/scripts/preflight.sh

Read ./complexa_setup/preflight.json. A sweep multiplies GPU time by the number of configs. Before launching, confirm the cost with the user:

"This sweep produces N configs × ~M minutes per config ≈ TOTAL GPU-hours. OK to proceed? (y / reduce / cancel)"

If gpu.available=false, stop — sweeps are not feasible on CPU.

Step 2: Pick the pipeline + target

Use the same dialogue as complexa-design — do not duplicate it here. See .claude/skills/complexa-design/SKILL.md Step 2 ("Pick the pipeline") and Step 3 ("Gather parameters"). Capture:

• pipeline_config_name — e.g. search_binder_local_pipeline (default), search_ligand_binder_local_pipeline, search_ame_local_pipeline.
• task_name — e.g. 02_PDL1, 22_DerF21, 39_7V11_LIGAND. Passed as --override generation.task_name=<task>.
• run_name — short tag for output dir naming.

Step 3: Pick or author the sweeper YAML

Sweeper YAMLs live in configs/sweeps/. Each key is a dot-notation Hydra path; each value is a list. The cartesian product becomes N configs.

Canned sweepers

If one matches the user's intent, use it as-is. Otherwise author a new file.

Authoring a new sweeper

Minimal multi-axis example (saved to configs/sweeps/my_sweep.yaml):

# 3 beam widths × 2 nsteps = 6 configs
generation.search.beam_search.beam_width:
  - 2
  - 4
  - 8
 

Rules (from script_utils/generate_inference_configs.py:load_sweeper_file):

• Top-level mapping only. Keys are dot-notation Hydra paths.
• Values must be lists. A scalar is auto-wrapped into a single-element list (which pins a value without adding a dimension).
• Cartesian product: total configs = product of list lengths. Two 4-value axes = 16 configs; budget accordingly.
• If a key appears in both the sweeper file and an --override, the override wins and that axis collapses.

See reference/sweep_axes.md for the full catalogue of swept keys (typical ranges, cost multipliers, what improves/regresses).

Dry-run preview before generating

Always confirm the config count first:

The output lists every axis + value list and prints DRY RUN — would generate N config pair(s).

Step 4: Generate configs + loop complexa design

Once the dry-run looks right, drop --dryrun to materialize inf_{idx}_{run_name}.yaml + eval_{idx}_{run_name}.yaml pairs under configs/inference_configs/ and configs/eval_configs/. Then loop complexa design over the inference configs.

This serialises the sweep on a single GPU — be honest with the user that wall-clock = N × per-run.

Multi-GPU host (optional speed-up)

On a host with K GPUs, shard the config list K ways and launch one loop per GPU in parallel. Each complexa design call uses exactly one GPU at default gen_njobs=1 / eval_njobs=1, so pinning via CUDA_VISIBLE_DEVICES keeps them from colliding:

Drop gen_njobs / eval_njobs overrides into the loop only if you intentionally want each complexa design invocation to consume multiple GPUs (and you have a way to keep them out of each other's way).

Step 5: Collect results

Each swept config writes to its own ./inference/inf_{idx}_{run_name}/ directory; the analyze step writes ./evaluation_results/eval_{idx}_{run_name}/results_*.csv. After the sweep finishes:

For each config, parse the analyze CSV (one row per generated binder). Standard columns used for ranking: i_pae, i_plddt, sc_rmsd, binder_seq, passes_filter (bool). If passes_filter is missing, derive a success flag with the user's chosen thresholds (defaults: i_pae < 10, i_plddt > 0.7, sc_rmsd < 2.0 — confirm with user).

Step 6: Rank configs

Emit sweep_summary.csv to the run directory. One row per config:

Then report:

• Best config = argmax of success_rate. Tie-break on mean_i_pae (lower).
• Pareto frontier over (wall_clock_min, success_rate): a config is on the frontier iff no other config is both faster AND has higher success rate.

Print the best config + the frontier to the terminal. Save the full table to sweep_summary.csv.

Step 7: Emit manifest

Capture the resolved invocation + outputs for replay. The shared helper takes a single --output-dir (it walks for CSVs and pulls the Hydra config from the run's .hydra/), so call it against the best-config's evaluation directory and pass the parent complexa design command for that run:

Alongside the manifest, save the ranked sweep_summary.csv from Step 6 to ./sweep_runs/my_sweep/sweep_summary.csv and surface both paths to the user.

Recommended sweep recipes

Hardware

Total GPU-time for a sweep = N_configs × per_run_GPU_time. A single search_binder_local_pipeline run on one A100 is roughly 30–90 min (binder length + nsteps dependent). A 4-axis × 4-value sweep = 256 configs × ~60 min = ~256 GPU-hours — at that scale, plan on a multi-GPU host with the Step-4 sharding pattern.

Refer to the shared hardware reference for the per-run baseline + VRAM minima.

Troubleshooting

For per-axis reference (typical ranges, cost, what gets better/worse), see reference/sweep_axes.md.

For the user-facing sweep system overview (config generation, output layout), see docs/SWEEP.md.