benchmarks.SpacemiTX60.baseline_dual_model_async.plot_results
Source: benchmarks/SpacemiTX60/baseline_dual_model_async/plot_results.py
benchmarks.SpacemiTX60.baseline_dual_model_async.plot_results
Plot sweep results produced by run_remote_sweep.py.
- Reads the CSV
- Filters successful runs (exit_code==0 and csv_parse_ok==1)
- Aggregates repeats (mean/std) per sweep value
- Produces PNG plots and aggregated CSV tables per sweep_param
- Summarizes min_all/max_all and renders 2D grid heatmaps (if present)
- NEW: Automatically plots ALL aggregated metrics (anything ending in *_mean)
Requires
pip install matplotlib
Usage
python3 plot_sweep_results.py --in_csv results/sweep.csv --out_dir results/plots
Optional
--no_plot_all_metrics Disable auto-plotting of every aggregated metric --verbose Print each generated PNG/CSV path
auto_plot_all_metrics(sp, xs, table_rows, out_dir, xlabel, baseline_vals, baseline_row, already_plotted, verbose)
For this sweep, plot every aggregated metric series that appears as _mean (with _std if present). Returns number of PNGs generated.
baseline_for_metric(mean_key, baseline_vals, baseline_row)
Try to find a sensible baseline horizontal line for an aggregated metric. - If we stored it in baseline_vals, use that. - Else, if it's a raw metric like d_lat_p99_ms_mean -> look up d_lat_p99_ms in baseline row.
effective_seconds(row, fallback_duration=10.0)
Many totals are accumulated only after warmup. Use effective = duration_s - warmup_s (per row).