experiments: freeze oral A confirmation gates

1 files changed, 64 insertions, 0 deletions

diff --git a/experiments/oral_a_full_development.py b/experiments/oral_a_full_development.py
new file mode 100755
index 0000000..0d8b8eb
--- /dev/null
+++ b/experiments/oral_a_full_development.py
@@ -0,0 +1,64 @@
+#!/usr/bin/env python3
+"""Run a shard of the frozen A3 full ResNet-20 local-method development gate."""
+import argparse
+import json
+import os
+import subprocess
+import sys
+
+
+def main():
+    parser = argparse.ArgumentParser()
+    parser.add_argument("--short_selection", default="results/oral_a_short_selection.json")
+    parser.add_argument("--bp_selection", default="results/oral_a_bp_selection.json")
+    parser.add_argument("--device", default="cuda")
+    parser.add_argument("--shard_index", type=int, default=0)
+    parser.add_argument("--num_shards", type=int, default=1)
+    parser.add_argument("--dry_run", action="store_true")
+    args = parser.parse_args()
+    if not 0 <= args.shard_index < args.num_shards:
+        raise ValueError("invalid shard index")
+    with open(args.short_selection) as handle:
+        short = json.load(handle)
+    with open(args.bp_selection) as handle:
+        bp = json.load(handle)
+    if short["status"] != "selected":
+        raise ValueError("A2b advancement gate did not pass")
+    if not bp["status"].startswith("passed_"):
+        raise ValueError("A1 BP reference gate did not pass")
+    dfa = short["selected_dfa"]
+    sdil = short["selected_sdil"]
+    common = [
+        sys.executable, "experiments/conv_run.py", "--device", args.device,
+        "--depth", "20", "--width", "16", "--seed", "0", "--loader_seed", "0",
+        "--batch_size", "128", "--epochs", "200", "--val_examples", "5000",
+        "--split_seed", "2027", "--eval_split", "validation", "--eval_every", "20",
+        "--augment_train", "1", "--lr_schedule", "step", "--lr_milestones", "100,150",
+        "--lr_gamma", "0.1", "--warmup_epochs", "0", "--momentum", "0.9",
+        "--weight_decay", "1e-4", "--normalization", "batchnorm",
+        "--output_lr", "0.1", "--alignment_probe", "32",
+    ]
+    jobs = [
+        ("dfa", common + [
+            "--mode", "dfa", "--lr", str(dfa["lr"]), "--a_scale", "1",
+            "--vectorizer_mode", "spatial_template",
+            "--out", "results/oral_a_dev/dfa_full_r20_s0.json"]),
+        ("sdil", common + [
+            "--mode", "sdil", "--lr", str(sdil["lr"]),
+            "--vectorizer_mode", sdil["vectorizer_mode"],
+            "--a_scale", str(sdil["a_scale"]), "--eta_A", str(sdil["eta_A"]),
+            "--a_warmup_steps", "100", "--pert_sigma", "0.01",
+            "--pert_directions", "1", "--pert_every", "4",
+            "--out", "results/oral_a_dev/sdil_full_r20_s0.json"]),
+    ]
+    os.makedirs("results/oral_a_dev", exist_ok=True)
+    for index, (tag, command) in enumerate(jobs):
+        if index % args.num_shards != args.shard_index:
+            continue
+        print(tag, " ".join(command), flush=True)
+        if not args.dry_run:
+            subprocess.run(command, check=True)
+
+
+if __name__ == "__main__":
+    main()