File size: 4,126 Bytes
18a82fb | 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 | #!/usr/bin/env python
"""
Debug script to check dataset output types
Save this as: scripts/debug_dataset.py
"""
import sys
from pathlib import Path
# Add project root to Python path
project_root = Path(__file__).parent.parent
sys.path.append(str(project_root))
import torch
from src.data.data_utils import create_data_loaders
def debug_dataset():
"""Debug dataset to find the data type issue"""
print("🏗️ Project structure debug")
print(f"📁 Script location: {Path(__file__).parent}")
print(f"📁 Project root: {project_root}")
print(f"📁 Data directory: {project_root / 'data'}")
# Check if data files exist
data_dir = project_root / 'data'
train_split = data_dir / 'splits' / 'train.json'
val_split = data_dir / 'splits' / 'val.json'
print(f"\n📋 File existence check:")
print(f" Train split: {train_split.exists()} - {train_split}")
print(f" Val split: {val_split.exists()} - {val_split}")
if not train_split.exists():
print("❌ Train split file not found!")
return
# Create minimal config for debugging
config = {
'data_dir': str(data_dir), # Use absolute path
'batch_size': 4, # Small batch for debugging
'num_workers': 0, # No multiprocessing for easier debugging
'use_cached': False,
'model_name': 'resnet50',
'num_classes': 5
}
print(f"\n🔍 Creating data loaders with config:")
for key, value in config.items():
print(f" {key}: {value}")
try:
train_loader, val_loader, class_weights, class_names = create_data_loaders(config)
print(f"✅ Data loaders created successfully")
print(f"📊 Class names: {class_names}")
print(f"📊 Train dataset size: {len(train_loader.dataset)}")
print(f"📊 Val dataset size: {len(val_loader.dataset)}")
# Check first batch
print(f"\n🔬 Checking first batch...")
for i, batch in enumerate(train_loader):
print(f"\nBatch {i}:")
print(f" Type of batch: {type(batch)}")
if isinstance(batch, (list, tuple)):
print(f" Batch length: {len(batch)}")
for j, item in enumerate(batch):
print(f" Item {j}: type={type(item)}")
if hasattr(item, 'shape'):
print(f" shape={item.shape}")
if hasattr(item, 'dtype'):
print(f" dtype={item.dtype}")
# If it's supposed to be a tensor but isn't, show content
if not isinstance(item, torch.Tensor):
try:
if hasattr(item, '__iter__') and not isinstance(item, str):
content = list(item)[:5] if len(item) > 5 else list(item)
print(f" content preview: {content}")
else:
print(f" content: {item}")
except Exception as e:
print(f" content preview error: {e}")
else:
print(f" Batch content: {batch}")
# Only check first batch
break
except Exception as e:
print(f"❌ Error creating data loaders: {e}")
import traceback
traceback.print_exc()
# Try to load just the JSON to see the data format
print(f"\n🔍 Checking raw JSON data...")
try:
import json
with open(train_split, 'r') as f:
data = json.load(f)
print(f"📄 JSON loaded successfully: {len(data)} items")
if data:
print(f"📋 First item keys: {list(data[0].keys())}")
print(f"📋 First item: {data[0]}")
except Exception as json_e:
print(f"❌ Error reading JSON: {json_e}")
if __name__ == '__main__':
debug_dataset() |