Skip to content

benchmarks.SpacemiTX60.baseline_dual_model_async.run_benchmark_remote

Source: benchmarks/SpacemiTX60/baseline_dual_model_async/run_benchmark_remote.py

benchmarks.SpacemiTX60.baseline_dual_model_async.run_benchmark_remote

Remote OFAT sweep runner over SSH (host-side).

Key features: - OFAT sweeps: vary ONE factor at a time vs baseline - optional min_all / max_all corner runs - optional small pairwise 2D grids for interactions - supports up to 8-core masks easily - parses CSV_HEADER/CSV_ROW from the benchmark binary output - stores metadata + parsed metrics into one results CSV - writes full logs per run

Example

python3 run_remote_sweep.py --ssh_host 10.44.86.251 --ssh_user root --remote_dir /home/baseline --bin_name benchmark-baseline-dual-model-async-run --dronet_vmfb dronet.q.int8.vmfb --mlp_vmfb mlp.q.int8.vmfb --out_csv results/sweep.csv --logs_dir results/logs --repeats 3 --mlp_hz 10,20,30 --duration_s 10 --warmup_s 2 --report_hz 0 --dronet_sensor_hz 30,60,120 --dronet_inflight 4,8,16 --mlp_inflight 1,2 --auto_core_masks_1to8 --include_minmax 1 --pairwise_grids core_mask:dronet_inflight,mlp_hz:mlp_inflight

add_pairwise_grids(plan, base, pairwise_specs, values_by_param)

Adds small 2D grids keeping all other params at baseline. pairwise_specs: ["core_mask:dronet_inflight", "mlp_hz:mlp_inflight", ...] values_by_param keys must include both params.