TeacherPuffy
commited on
Create eval_model.py
Browse files- eval_model.py +117 -0
eval_model.py
ADDED
@@ -0,0 +1,117 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
import argparse
|
2 |
+
import os
|
3 |
+
import torch
|
4 |
+
import torch.nn as nn
|
5 |
+
from datasets import load_dataset
|
6 |
+
from torch.utils.data import DataLoader
|
7 |
+
from collections import defaultdict
|
8 |
+
import numpy as np
|
9 |
+
from PIL import Image
|
10 |
+
|
11 |
+
# Define the MLP model (same as in the training script)
|
12 |
+
class MLP(nn.Module):
|
13 |
+
def __init__(self, input_size, hidden_sizes, output_size):
|
14 |
+
super(MLP, self).__init__()
|
15 |
+
layers = []
|
16 |
+
sizes = [input_size] + hidden_sizes + [output_size]
|
17 |
+
for i in range(len(sizes) - 1):
|
18 |
+
layers.append(nn.Linear(sizes[i], sizes[i+1]))
|
19 |
+
if i < len(sizes) - 2:
|
20 |
+
layers.append(nn.ReLU())
|
21 |
+
self.model = nn.Sequential(*layers)
|
22 |
+
|
23 |
+
def forward(self, x):
|
24 |
+
return self.model(x)
|
25 |
+
|
26 |
+
# Custom Dataset class to handle image preprocessing (same as in the training script)
|
27 |
+
class TinyImageNetDataset(Dataset):
|
28 |
+
def __init__(self, dataset):
|
29 |
+
self.dataset = dataset
|
30 |
+
|
31 |
+
def __len__(self):
|
32 |
+
return len(self.dataset)
|
33 |
+
|
34 |
+
def __getitem__(self, idx):
|
35 |
+
example = self.dataset[idx]
|
36 |
+
img = example['image']
|
37 |
+
img = np.array(img.convert('L')) # Convert PIL image to grayscale NumPy array
|
38 |
+
img = img.reshape(-1) # Flatten the image
|
39 |
+
img = torch.from_numpy(img).float() # Convert to tensor
|
40 |
+
label = torch.tensor(example['label'])
|
41 |
+
return img, label
|
42 |
+
|
43 |
+
# Function to evaluate the model on the validation set and compute class-wise accuracy
|
44 |
+
def evaluate_model(model, val_loader, num_classes):
|
45 |
+
device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
|
46 |
+
model.to(device)
|
47 |
+
model.eval()
|
48 |
+
|
49 |
+
class_correct = defaultdict(int)
|
50 |
+
class_total = defaultdict(int)
|
51 |
+
|
52 |
+
with torch.no_grad():
|
53 |
+
for inputs, labels in val_loader:
|
54 |
+
inputs, labels = inputs.to(device), labels.to(device)
|
55 |
+
|
56 |
+
outputs = model(inputs)
|
57 |
+
_, predicted = torch.max(outputs, 1)
|
58 |
+
|
59 |
+
for label, prediction in zip(labels, predicted):
|
60 |
+
if label == prediction:
|
61 |
+
class_correct[label.item()] += 1
|
62 |
+
class_total[label.item()] += 1
|
63 |
+
|
64 |
+
class_accuracies = {}
|
65 |
+
for class_idx in range(num_classes):
|
66 |
+
if class_total[class_idx] > 0:
|
67 |
+
class_accuracies[class_idx] = 100 * class_correct[class_idx] / class_total[class_idx]
|
68 |
+
else:
|
69 |
+
class_accuracies[class_idx] = 0.0
|
70 |
+
|
71 |
+
return class_accuracies
|
72 |
+
|
73 |
+
# Main function to load the model and evaluate it
|
74 |
+
def main():
|
75 |
+
parser = argparse.ArgumentParser(description='Evaluate the MLP model on the zh-plus/tiny-imagenet dataset.')
|
76 |
+
parser.add_argument('--checkpoint', type=str, required=True, help='Path to the model checkpoint')
|
77 |
+
parser.add_argument('--layer_count', type=int, default=2, help='Number of hidden layers (default: 2)')
|
78 |
+
parser.add_argument('--width', type=int, default=512, help='Number of neurons per hidden layer (default: 512)')
|
79 |
+
parser.add_argument('--output_file', type=str, default='class_accuracies.txt', help='Output file to save class-wise accuracies')
|
80 |
+
args = parser.parse_args()
|
81 |
+
|
82 |
+
# Load the zh-plus/tiny-imagenet dataset
|
83 |
+
dataset = load_dataset('zh-plus/tiny-imagenet')
|
84 |
+
val_dataset = dataset['valid'] # Assuming 'validation' is the correct key
|
85 |
+
|
86 |
+
# Determine the number of classes
|
87 |
+
num_classes = len(set(val_dataset['label']))
|
88 |
+
|
89 |
+
# Determine the fixed resolution of the images
|
90 |
+
image_size = 64 # Assuming the images are square
|
91 |
+
|
92 |
+
# Define the model
|
93 |
+
input_size = image_size * image_size # Since images are grayscale
|
94 |
+
hidden_sizes = [args.width] * args.layer_count
|
95 |
+
output_size = num_classes
|
96 |
+
|
97 |
+
model = MLP(input_size, hidden_sizes, output_size)
|
98 |
+
model.load_state_dict(torch.load(args.checkpoint))
|
99 |
+
|
100 |
+
# Create DataLoader for validation
|
101 |
+
val_loader = DataLoader(TinyImageNetDataset(val_dataset), batch_size=8, shuffle=False)
|
102 |
+
|
103 |
+
# Evaluate the model
|
104 |
+
class_accuracies = evaluate_model(model, val_loader, num_classes)
|
105 |
+
|
106 |
+
# Print the results
|
107 |
+
print("Class-wise accuracies:")
|
108 |
+
for class_idx, accuracy in class_accuracies.items():
|
109 |
+
print(f"Class {class_idx}: {accuracy:.2f}%")
|
110 |
+
|
111 |
+
# Save the results to a text file
|
112 |
+
with open(args.output_file, 'w') as f:
|
113 |
+
for class_idx, accuracy in class_accuracies.items():
|
114 |
+
f.write(f"Class {class_idx}: {accuracy:.2f}%\n")
|
115 |
+
|
116 |
+
if __name__ == '__main__':
|
117 |
+
main()
|