File size: 10,708 Bytes
d12ab12 |
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 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 |
#!/usr/bin/env python3
"""
Main training script for Architectural Style Classification
Advanced Deep Learning Approach with Hierarchical Multi-Modal Architecture
"""
import os
import sys
import json
import argparse
from typing import Dict, Any
import torch
import pytorch_lightning as pl
# Add src to path
sys.path.append(os.path.join(os.path.dirname(__file__), '..'))
from src.models import HierarchicalArchitecturalClassifier, BaselineModels
from src.training.trainer import ArchitecturalTrainer, ExperimentRunner
from src.training.losses import CombinedLoss
from src.utils.config import load_config, save_config
def create_experiment_configs() -> Dict[str, Dict[str, Any]]:
"""Create different experiment configurations."""
configs = {
# Baseline experiments
'baseline_resnet': {
'experiment_name': 'baseline_resnet',
'model_type': 'resnet',
'num_classes': 25,
'learning_rate': 1e-4,
'max_epochs': 50,
'batch_size': 32,
'use_hierarchical_loss': False,
'use_contrastive_loss': False,
'use_style_relationship_loss': False,
'use_wandb': False
},
'baseline_efficientnet': {
'experiment_name': 'baseline_efficientnet',
'model_type': 'efficientnet',
'num_classes': 25,
'learning_rate': 1e-4,
'max_epochs': 50,
'batch_size': 32,
'use_hierarchical_loss': False,
'use_contrastive_loss': False,
'use_style_relationship_loss': False,
'use_wandb': False
},
'baseline_vit': {
'experiment_name': 'baseline_vit',
'model_type': 'vit',
'num_classes': 25,
'learning_rate': 1e-4,
'max_epochs': 50,
'batch_size': 16, # Smaller batch size for ViT
'use_hierarchical_loss': False,
'use_contrastive_loss': False,
'use_style_relationship_loss': False,
'use_wandb': False
},
# Hierarchical model experiments
'hierarchical_basic': {
'experiment_name': 'hierarchical_basic',
'model_type': 'hierarchical',
'num_classes': 25,
'num_broad_classes': 5,
'num_fine_classes': 25,
'learning_rate': 1e-4,
'max_epochs': 100,
'batch_size': 16,
'use_hierarchical_loss': True,
'use_contrastive_loss': False,
'use_style_relationship_loss': True,
'curriculum_stages': [
{'epochs': 20, 'classes': ['ancient', 'medieval', 'modern']},
{'epochs': 80, 'classes': list(range(25))}
],
'use_wandb': False
},
'hierarchical_contrastive': {
'experiment_name': 'hierarchical_contrastive',
'model_type': 'hierarchical',
'num_classes': 25,
'num_broad_classes': 5,
'num_fine_classes': 25,
'learning_rate': 1e-4,
'max_epochs': 100,
'batch_size': 16,
'use_hierarchical_loss': True,
'use_contrastive_loss': True,
'use_style_relationship_loss': True,
'curriculum_stages': [
{'epochs': 20, 'classes': ['ancient', 'medieval', 'modern']},
{'epochs': 80, 'classes': list(range(25))}
],
'use_wandb': False
},
# Advanced experiments
'hierarchical_advanced': {
'experiment_name': 'hierarchical_advanced',
'model_type': 'hierarchical',
'num_classes': 25,
'num_broad_classes': 5,
'num_fine_classes': 25,
'learning_rate': 5e-5,
'max_epochs': 150,
'batch_size': 16,
'use_hierarchical_loss': True,
'use_contrastive_loss': True,
'use_style_relationship_loss': True,
'use_mixed_precision': True,
'gradient_clip_val': 1.0,
'accumulate_grad_batches': 2,
'curriculum_stages': [
{'epochs': 30, 'classes': ['ancient', 'medieval', 'modern']},
{'epochs': 60, 'classes': list(range(25))},
{'epochs': 60, 'classes': list(range(25))}
],
'use_wandb': True
}
}
return configs
def run_single_experiment(config: Dict[str, Any], data_path: str = None):
"""Run a single experiment."""
print(f"Starting experiment: {config['experiment_name']}")
print(f"Model type: {config['model_type']}")
print(f"Configuration: {json.dumps(config, indent=2)}")
# Initialize experiment runner
runner = ExperimentRunner(config)
# Run experiment
try:
trainer, pl_trainer = runner.run_experiment()
print(f"Experiment {config['experiment_name']} completed successfully!")
return trainer, pl_trainer
except Exception as e:
print(f"Experiment {config['experiment_name']} failed: {str(e)}")
raise
def run_experiment_suite(experiment_names: list = None, data_path: str = None):
"""Run a suite of experiments."""
configs = create_experiment_configs()
if experiment_names is None:
experiment_names = list(configs.keys())
results = {}
for exp_name in experiment_names:
if exp_name not in configs:
print(f"Warning: Experiment {exp_name} not found in configurations")
continue
print(f"\n{'='*50}")
print(f"Running experiment: {exp_name}")
print(f"{'='*50}")
try:
trainer, pl_trainer = run_single_experiment(configs[exp_name], data_path)
results[exp_name] = {
'status': 'success',
'trainer': trainer,
'pl_trainer': pl_trainer
}
except Exception as e:
print(f"Experiment {exp_name} failed: {str(e)}")
results[exp_name] = {
'status': 'failed',
'error': str(e)
}
# Save results summary
save_experiment_results(results)
return results
def save_experiment_results(results: Dict[str, Any]):
"""Save experiment results summary."""
summary = {}
for exp_name, result in results.items():
if result['status'] == 'success':
summary[exp_name] = {
'status': 'success',
'model_type': result['trainer'].model.__class__.__name__,
'hyperparameters': result['trainer'].hparams
}
else:
summary[exp_name] = {
'status': 'failed',
'error': result.get('error', 'Unknown error')
}
# Save to file
os.makedirs('results', exist_ok=True)
with open('results/experiment_summary.json', 'w') as f:
json.dump(summary, f, indent=2, default=str)
print(f"\nExperiment summary saved to results/experiment_summary.json")
def test_model_creation():
"""Test model creation to ensure everything works."""
print("Testing model creation...")
try:
# Test hierarchical model
hierarchical_model = HierarchicalArchitecturalClassifier()
print(f"✓ Hierarchical model created successfully")
print(f" Parameters: {sum(p.numel() for p in hierarchical_model.parameters()):,}")
# Test baseline models
resnet_model = BaselineModels.resnet50()
print(f"✓ ResNet-50 model created successfully")
print(f" Parameters: {sum(p.numel() for p in resnet_model.parameters()):,}")
efficientnet_model = BaselineModels.efficientnet_b4()
print(f"✓ EfficientNet-B4 model created successfully")
print(f" Parameters: {sum(p.numel() for p in efficientnet_model.parameters()):,}")
vit_model = BaselineModels.vit_base()
print(f"✓ ViT-Base model created successfully")
print(f" Parameters: {sum(p.numel() for p in vit_model.parameters()):,}")
# Test loss functions
combined_loss = CombinedLoss()
print(f"✓ Combined loss function created successfully")
print("\nAll model tests passed! ✓")
return True
except Exception as e:
print(f"Model test failed: {str(e)}")
return False
def main():
"""Main function."""
parser = argparse.ArgumentParser(description='Architectural Style Classification Training')
parser.add_argument('--experiment', type=str, default=None,
help='Specific experiment to run')
parser.add_argument('--suite', action='store_true',
help='Run the full experiment suite')
parser.add_argument('--test', action='store_true',
help='Test model creation and setup')
parser.add_argument('--data_path', type=str, default=None,
help='Path to dataset')
parser.add_argument('--config', type=str, default=None,
help='Path to custom config file')
args = parser.parse_args()
# Set random seeds for reproducibility
torch.manual_seed(42)
pl.seed_everything(42)
print("Architectural Style Classification Training")
print("=" * 50)
# Test mode
if args.test:
if test_model_creation():
print("Setup test completed successfully!")
else:
print("Setup test failed!")
return 1
# Load custom config if provided
if args.config:
config = load_config(args.config)
run_single_experiment(config, args.data_path)
return 0
# Run specific experiment
if args.experiment:
configs = create_experiment_configs()
if args.experiment not in configs:
print(f"Experiment '{args.experiment}' not found!")
print(f"Available experiments: {list(configs.keys())}")
return 1
run_single_experiment(configs[args.experiment], args.data_path)
return 0
# Run experiment suite
if args.suite:
run_experiment_suite(data_path=args.data_path)
return 0
# Default: run basic hierarchical experiment
print("No specific experiment specified. Running basic hierarchical experiment...")
configs = create_experiment_configs()
run_single_experiment(configs['hierarchical_basic'], args.data_path)
return 0
if __name__ == "__main__":
exit(main())
|