Create demo.py
Browse files
demo.py
ADDED
@@ -0,0 +1,135 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
#!/usr/bin/env python3
|
2 |
+
"""
|
3 |
+
Quick demo script to test Mamba Graph implementation
|
4 |
+
Device-safe version
|
5 |
+
"""
|
6 |
+
|
7 |
+
import torch
|
8 |
+
import os
|
9 |
+
from core.graph_mamba import GraphMamba
|
10 |
+
from data.loader import GraphDataLoader
|
11 |
+
from utils.metrics import GraphMetrics
|
12 |
+
|
13 |
+
def main():
|
14 |
+
print("π§ Testing Mamba Graph Neural Network")
|
15 |
+
print("=" * 50)
|
16 |
+
|
17 |
+
# Configuration
|
18 |
+
config = {
|
19 |
+
'model': {
|
20 |
+
'd_model': 128,
|
21 |
+
'd_state': 8,
|
22 |
+
'd_conv': 4,
|
23 |
+
'expand': 2,
|
24 |
+
'n_layers': 3,
|
25 |
+
'dropout': 0.1
|
26 |
+
},
|
27 |
+
'data': {
|
28 |
+
'batch_size': 16,
|
29 |
+
'test_split': 0.2
|
30 |
+
},
|
31 |
+
'ordering': {
|
32 |
+
'strategy': 'bfs',
|
33 |
+
'preserve_locality': True
|
34 |
+
}
|
35 |
+
}
|
36 |
+
|
37 |
+
# Setup device
|
38 |
+
if os.getenv('SPACE_ID'):
|
39 |
+
device = torch.device('cpu')
|
40 |
+
else:
|
41 |
+
device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')
|
42 |
+
print(f"Device: {device}")
|
43 |
+
|
44 |
+
# Load dataset
|
45 |
+
print("\nπ Loading Cora dataset...")
|
46 |
+
try:
|
47 |
+
data_loader = GraphDataLoader()
|
48 |
+
dataset = data_loader.load_node_classification_data('Cora')
|
49 |
+
data = dataset[0].to(device)
|
50 |
+
|
51 |
+
# Dataset info
|
52 |
+
info = data_loader.get_dataset_info(dataset)
|
53 |
+
print(f"β
Success!")
|
54 |
+
print(f"Nodes: {data.num_nodes}")
|
55 |
+
print(f"Edges: {data.num_edges}")
|
56 |
+
print(f"Features: {info['num_features']}")
|
57 |
+
print(f"Classes: {info['num_classes']}")
|
58 |
+
|
59 |
+
except Exception as e:
|
60 |
+
print(f"β Error loading dataset: {e}")
|
61 |
+
return
|
62 |
+
|
63 |
+
# Initialize model
|
64 |
+
print("\nποΈ Initializing GraphMamba...")
|
65 |
+
try:
|
66 |
+
model = GraphMamba(config).to(device)
|
67 |
+
total_params = sum(p.numel() for p in model.parameters())
|
68 |
+
print(f"β
Model initialized!")
|
69 |
+
print(f"Parameters: {total_params:,}")
|
70 |
+
|
71 |
+
except Exception as e:
|
72 |
+
print(f"β Error initializing model: {e}")
|
73 |
+
return
|
74 |
+
|
75 |
+
# Forward pass test
|
76 |
+
print("\nπ Testing forward pass...")
|
77 |
+
try:
|
78 |
+
model.eval()
|
79 |
+
with torch.no_grad():
|
80 |
+
h = model(data.x, data.edge_index)
|
81 |
+
print(f"β
Forward pass successful!")
|
82 |
+
print(f"Input shape: {data.x.shape}")
|
83 |
+
print(f"Output shape: {h.shape}")
|
84 |
+
print(f"Output range: [{h.min():.3f}, {h.max():.3f}]")
|
85 |
+
|
86 |
+
except Exception as e:
|
87 |
+
print(f"β Forward pass failed: {e}")
|
88 |
+
return
|
89 |
+
|
90 |
+
# Test different ordering strategies
|
91 |
+
print("\nπ Testing ordering strategies...")
|
92 |
+
|
93 |
+
strategies = ['bfs', 'spectral', 'degree', 'community']
|
94 |
+
|
95 |
+
for strategy in strategies:
|
96 |
+
try:
|
97 |
+
config['ordering']['strategy'] = strategy
|
98 |
+
model_test = GraphMamba(config).to(device)
|
99 |
+
model_test.eval()
|
100 |
+
|
101 |
+
with torch.no_grad():
|
102 |
+
h = model_test(data.x, data.edge_index)
|
103 |
+
print(f"β
{strategy}: Success - Shape {h.shape}")
|
104 |
+
|
105 |
+
except Exception as e:
|
106 |
+
print(f"β {strategy}: Failed - {str(e)}")
|
107 |
+
|
108 |
+
# Test evaluation
|
109 |
+
print("\nπ Testing evaluation...")
|
110 |
+
try:
|
111 |
+
# Initialize classifier
|
112 |
+
num_classes = info['num_classes']
|
113 |
+
model._init_classifier(num_classes, device)
|
114 |
+
|
115 |
+
# Create test mask if not available
|
116 |
+
if hasattr(data, 'test_mask'):
|
117 |
+
mask = data.test_mask
|
118 |
+
else:
|
119 |
+
mask = torch.zeros(data.num_nodes, dtype=torch.bool, device=device)
|
120 |
+
mask[data.num_nodes//2:] = True
|
121 |
+
|
122 |
+
metrics = GraphMetrics.evaluate_node_classification(model, data, mask, device)
|
123 |
+
print("β
Evaluation successful!")
|
124 |
+
for metric, value in metrics.items():
|
125 |
+
if isinstance(value, float):
|
126 |
+
print(f" {metric}: {value:.4f}")
|
127 |
+
|
128 |
+
except Exception as e:
|
129 |
+
print(f"β Evaluation failed: {e}")
|
130 |
+
|
131 |
+
print("\n⨠Demo completed!")
|
132 |
+
print("π Ready for production deployment!")
|
133 |
+
|
134 |
+
if __name__ == "__main__":
|
135 |
+
main()
|