Datasets:
File size: 10,508 Bytes
b7fcddf | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 | #!/usr/bin/env python3
"""Deep dive into teacher-forced GIN decoder: qualitative analysis + dimension ablation.
Trains teacher-forced GIN at multiple hidden dimensions, evaluates syntactic validity
using both the unique-types heuristic and real Ruby syntax checking (via check_syntax.rb),
and saves generated samples for qualitative analysis.
"""
from __future__ import annotations
import json
import os
import subprocess
import sys
import time
import torch
sys.path.insert(0, os.path.join(os.path.dirname(__file__), "..", "src"))
from data_processing import create_data_loaders
from models import ASTAutoencoder
DATASET_PATH = "dataset"
ENCODER_WEIGHTS = "models/best_model.pt"
RESULTS_DIR = "results/gin_deep_dive"
EPOCHS = 30
BATCH_SIZE = 32
NUM_SAMPLES = 200
LEARNING_RATE = 0.001
def check_ruby_syntax(code: str) -> bool:
"""Check if code is valid Ruby using the parser gem."""
try:
result = subprocess.run(
["ruby", "scripts/check_syntax.rb"],
input=code,
capture_output=True,
text=True,
timeout=5,
)
return result.returncode == 0
except (subprocess.TimeoutExpired, FileNotFoundError):
return False
def reconstruct_code_from_types(pred_types: torch.Tensor, type_vocab: list[str] | None = None) -> str:
"""Convert predicted node type indices back to a pseudo-code string."""
types = pred_types.cpu().tolist()
if type_vocab:
return " ".join(type_vocab[t] for t in types if t < len(type_vocab))
return " ".join(f"type_{t}" for t in types)
def train_and_evaluate(
hidden_dim: int,
decoder_edge_mode: str = "teacher_forced",
decoder_conv_type: str = "GIN",
num_layers: int = 3,
label: str = "",
) -> dict:
"""Train an autoencoder variant and evaluate generation quality."""
device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
print(f"\n{'='*70}")
print(f"Training: {label} (dim={hidden_dim}, layers={num_layers}, "
f"edge={decoder_edge_mode}, conv={decoder_conv_type})")
print(f"Device: {device}")
print(f"{'='*70}")
train_path = os.path.join(DATASET_PATH, "train.jsonl")
val_path = os.path.join(DATASET_PATH, "val.jsonl")
train_loader, val_loader = create_data_loaders(
train_path, val_path, batch_size=BATCH_SIZE, shuffle=True, num_workers=0
)
model = ASTAutoencoder(
encoder_input_dim=74,
node_output_dim=74,
hidden_dim=hidden_dim,
num_layers=num_layers,
conv_type="SAGE",
freeze_encoder=True,
encoder_weights_path=ENCODER_WEIGHTS,
decoder_conv_type=decoder_conv_type,
decoder_edge_mode=decoder_edge_mode,
).to(device)
param_count = sum(p.numel() for p in model.decoder.parameters() if p.requires_grad)
print(f"Trainable decoder parameters: {param_count:,}")
from loss import ast_reconstruction_loss_improved
optimizer = torch.optim.Adam(model.decoder.parameters(), lr=LEARNING_RATE)
scaler = torch.amp.GradScaler("cuda") if device.type == "cuda" else None
best_val_loss = float("inf")
model_path = os.path.join(RESULTS_DIR, f"{label}_decoder.pt")
t0 = time.time()
for epoch in range(EPOCHS):
model.train()
epoch_loss = 0.0
batches = 0
for batch in train_loader:
batch = batch.to(device)
optimizer.zero_grad()
if scaler:
with torch.amp.autocast("cuda"):
result = model(batch)
loss = ast_reconstruction_loss_improved(batch, result["reconstruction"])
scaler.scale(loss).backward()
scaler.step(optimizer)
scaler.update()
else:
result = model(batch)
loss = ast_reconstruction_loss_improved(batch, result["reconstruction"])
loss.backward()
optimizer.step()
epoch_loss += loss.item()
batches += 1
avg_train = epoch_loss / max(batches, 1)
# Validate
model.eval()
val_loss = 0.0
val_batches = 0
with torch.no_grad():
for batch in val_loader:
batch = batch.to(device)
result = model(batch)
loss = ast_reconstruction_loss_improved(batch, result["reconstruction"])
val_loss += loss.item()
val_batches += 1
avg_val = val_loss / max(val_batches, 1)
if avg_val < best_val_loss:
best_val_loss = avg_val
torch.save({"decoder_state_dict": model.decoder.state_dict()}, model_path)
if (epoch + 1) % 5 == 0 or epoch == 0:
elapsed = time.time() - t0
print(f" Epoch {epoch+1:3d}/{EPOCHS} | "
f"train={avg_train:.4f} val={avg_val:.4f} "
f"best={best_val_loss:.4f} | {elapsed:.0f}s")
train_time = time.time() - t0
print(f"Training complete in {train_time:.0f}s, best val_loss={best_val_loss:.4f}")
# Load best checkpoint
checkpoint = torch.load(model_path, map_location=device, weights_only=False)
model.decoder.load_state_dict(checkpoint["decoder_state_dict"])
model.eval()
# Evaluate: generate samples and check validity
print(f"\nEvaluating {NUM_SAMPLES} samples...")
_, eval_loader = create_data_loaders(
val_path, val_path, batch_size=1, shuffle=False, num_workers=0
)
samples = []
heuristic_valid = 0
total = 0
with torch.no_grad():
for batch in eval_loader:
if total >= NUM_SAMPLES:
break
batch = batch.to(device)
result = model(batch)
recon = result["reconstruction"]
node_feats = recon.get("node_features") if isinstance(recon, dict) else None
if node_feats is None:
total += 1
continue
pred_types = node_feats.argmax(dim=-1)
orig_types = batch.x.argmax(dim=-1) if batch.x.dim() > 1 else batch.x
unique_pred = len(pred_types.unique())
unique_orig = len(orig_types.unique())
type_match = (pred_types == orig_types).float().mean().item()
# Heuristic validity (>2 unique types)
heuristic_ok = unique_pred > 2
sample = {
"index": total,
"num_nodes": int(pred_types.shape[0]),
"pred_unique_types": unique_pred,
"orig_unique_types": unique_orig,
"type_accuracy": round(type_match, 4),
"heuristic_valid": heuristic_ok,
"pred_type_ids": pred_types.cpu().tolist(),
"orig_type_ids": orig_types.cpu().tolist(),
}
samples.append(sample)
if heuristic_ok:
heuristic_valid += 1
total += 1
heuristic_pct = (heuristic_valid / total * 100) if total > 0 else 0.0
# Compute statistics on type predictions
type_accuracies = [s["type_accuracy"] for s in samples]
avg_type_accuracy = sum(type_accuracies) / len(type_accuracies) if type_accuracies else 0
unique_counts = [s["pred_unique_types"] for s in samples]
avg_unique = sum(unique_counts) / len(unique_counts) if unique_counts else 0
# Sort by type_accuracy descending to show best samples first
samples.sort(key=lambda s: s["type_accuracy"], reverse=True)
result = {
"label": label,
"hidden_dim": hidden_dim,
"num_layers": num_layers,
"decoder_conv_type": decoder_conv_type,
"decoder_edge_mode": decoder_edge_mode,
"trainable_params": param_count,
"best_val_loss": round(best_val_loss, 4),
"train_time_s": round(train_time, 1),
"samples_evaluated": total,
"heuristic_valid": heuristic_valid,
"heuristic_validity_pct": round(heuristic_pct, 2),
"avg_type_accuracy": round(avg_type_accuracy, 4),
"avg_unique_pred_types": round(avg_unique, 2),
"top_samples": samples[:20],
}
# Save individual result
result_path = os.path.join(RESULTS_DIR, f"{label}_results.json")
with open(result_path, "w") as f:
json.dump(result, f, indent=2)
print(f"\nResults: heuristic_validity={heuristic_pct:.1f}% "
f"({heuristic_valid}/{total}), "
f"avg_type_acc={avg_type_accuracy:.4f}, "
f"avg_unique_types={avg_unique:.1f}")
return result
def main() -> None:
os.makedirs(RESULTS_DIR, exist_ok=True)
configs = [
# Replicate the 7% result
{"hidden_dim": 256, "decoder_edge_mode": "teacher_forced",
"decoder_conv_type": "GIN", "num_layers": 3, "label": "tf-gin-256"},
# Ablation: smaller dim
{"hidden_dim": 128, "decoder_edge_mode": "teacher_forced",
"decoder_conv_type": "GIN", "num_layers": 3, "label": "tf-gin-128"},
# Ablation: larger dim
{"hidden_dim": 512, "decoder_edge_mode": "teacher_forced",
"decoder_conv_type": "GIN", "num_layers": 3, "label": "tf-gin-512"},
# Ablation: deeper network
{"hidden_dim": 256, "decoder_edge_mode": "teacher_forced",
"decoder_conv_type": "GIN", "num_layers": 5, "label": "tf-gin-256-deep"},
# Control: chain GIN (should be ~0%)
{"hidden_dim": 256, "decoder_edge_mode": "chain",
"decoder_conv_type": "GIN", "num_layers": 3, "label": "chain-gin-256"},
]
all_results = []
for cfg in configs:
result = train_and_evaluate(**cfg)
all_results.append(result)
print(f"\n{'~'*70}")
# Summary
print(f"\n{'='*70}")
print("SUMMARY — Teacher-Forced GIN Deep Dive")
print(f"{'='*70}")
print(f"{'Label':<22s} {'Dim':>4s} {'Layers':>6s} {'Edge':>15s} "
f"{'Params':>10s} {'ValLoss':>8s} {'Validity':>8s} {'TypeAcc':>8s}")
print("-" * 90)
for r in all_results:
print(f"{r['label']:<22s} {r['hidden_dim']:>4d} {r['num_layers']:>6d} "
f"{r['decoder_edge_mode']:>15s} {r['trainable_params']:>10,d} "
f"{r['best_val_loss']:>8.4f} {r['heuristic_validity_pct']:>7.1f}% "
f"{r['avg_type_accuracy']:>8.4f}")
summary_path = os.path.join(RESULTS_DIR, "summary.json")
with open(summary_path, "w") as f:
json.dump(all_results, f, indent=2)
print(f"\nAll results saved to {RESULTS_DIR}/")
if __name__ == "__main__":
main()
|