Spaces:
Sleeping
Sleeping
Update app.py
Browse files
app.py
CHANGED
@@ -11,8 +11,12 @@ from sklearn.preprocessing import LabelEncoder
|
|
11 |
# Load dataset
|
12 |
dataset = load_dataset('thefcraft/civitai-stable-diffusion-337k', split='train[:10000]')
|
13 |
|
14 |
-
# Text preprocessing function
|
15 |
def preprocess_text(text, max_length=100):
|
|
|
|
|
|
|
|
|
16 |
# Convert text to lowercase and split into words
|
17 |
words = text.lower().split()
|
18 |
# Truncate or pad to max_length
|
@@ -29,32 +33,55 @@ class CustomDataset(Dataset):
|
|
29 |
transforms.Resize((224, 224)),
|
30 |
transforms.ToTensor(),
|
31 |
])
|
|
|
|
|
|
|
|
|
|
|
32 |
self.label_encoder = LabelEncoder()
|
33 |
-
self.labels = self.label_encoder.fit_transform(
|
34 |
|
35 |
# Create vocabulary from all prompts
|
36 |
self.vocab = set()
|
37 |
-
for item in
|
38 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
39 |
self.vocab = list(self.vocab)
|
40 |
self.word_to_idx = {word: idx for idx, word in enumerate(self.vocab)}
|
41 |
|
42 |
def __len__(self):
|
43 |
-
return len(self.
|
44 |
|
45 |
def text_to_vector(self, text):
|
46 |
-
|
47 |
-
|
48 |
-
|
49 |
-
|
50 |
-
|
51 |
-
|
|
|
|
|
|
|
|
|
52 |
|
53 |
def __getitem__(self, idx):
|
54 |
-
|
55 |
-
|
56 |
-
|
57 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
58 |
|
59 |
# Define CNN for image processing
|
60 |
class ImageModel(nn.Module):
|
@@ -85,11 +112,11 @@ class TextMLP(nn.Module):
|
|
85 |
|
86 |
# Combined model
|
87 |
class CombinedModel(nn.Module):
|
88 |
-
def __init__(self, vocab_size):
|
89 |
super(CombinedModel, self).__init__()
|
90 |
self.image_model = ImageModel()
|
91 |
self.text_model = TextMLP(vocab_size)
|
92 |
-
self.fc = nn.Linear(1024,
|
93 |
|
94 |
def forward(self, image, text):
|
95 |
image_features = self.image_model(image)
|
@@ -97,9 +124,15 @@ class CombinedModel(nn.Module):
|
|
97 |
combined = torch.cat((image_features, text_features), dim=1)
|
98 |
return self.fc(combined)
|
99 |
|
100 |
-
# Create dataset instance
|
|
|
101 |
custom_dataset = CustomDataset(dataset)
|
102 |
-
|
|
|
|
|
|
|
|
|
|
|
103 |
|
104 |
def get_recommendations(image):
|
105 |
model.eval()
|
@@ -111,7 +144,7 @@ def get_recommendations(image):
|
|
111 |
])
|
112 |
image_tensor = transform(image).unsqueeze(0)
|
113 |
|
114 |
-
# Create dummy text vector
|
115 |
dummy_text = torch.zeros((1, len(custom_dataset.vocab)))
|
116 |
|
117 |
# Get model output
|
@@ -121,9 +154,13 @@ def get_recommendations(image):
|
|
121 |
# Get recommended images and their information
|
122 |
recommendations = []
|
123 |
for idx in indices[0]:
|
124 |
-
|
125 |
-
|
126 |
-
|
|
|
|
|
|
|
|
|
127 |
|
128 |
return recommendations
|
129 |
|
@@ -137,4 +174,5 @@ interface = gr.Interface(
|
|
137 |
)
|
138 |
|
139 |
# Launch the app
|
140 |
-
|
|
|
|
11 |
# Load dataset
|
12 |
dataset = load_dataset('thefcraft/civitai-stable-diffusion-337k', split='train[:10000]')
|
13 |
|
14 |
+
# Text preprocessing function with None handling
|
15 |
def preprocess_text(text, max_length=100):
|
16 |
+
# Handle None or empty text
|
17 |
+
if text is None or not isinstance(text, str):
|
18 |
+
text = ""
|
19 |
+
|
20 |
# Convert text to lowercase and split into words
|
21 |
words = text.lower().split()
|
22 |
# Truncate or pad to max_length
|
|
|
33 |
transforms.Resize((224, 224)),
|
34 |
transforms.ToTensor(),
|
35 |
])
|
36 |
+
|
37 |
+
# Filter out None values from Model column
|
38 |
+
valid_indices = [i for i, model in enumerate(dataset['Model']) if model is not None]
|
39 |
+
self.valid_dataset = dataset.select(valid_indices)
|
40 |
+
|
41 |
self.label_encoder = LabelEncoder()
|
42 |
+
self.labels = self.label_encoder.fit_transform(self.valid_dataset['Model'])
|
43 |
|
44 |
# Create vocabulary from all prompts
|
45 |
self.vocab = set()
|
46 |
+
for item in self.valid_dataset['prompt']:
|
47 |
+
try:
|
48 |
+
self.vocab.update(preprocess_text(item))
|
49 |
+
except Exception as e:
|
50 |
+
print(f"Error processing prompt: {e}")
|
51 |
+
continue
|
52 |
+
|
53 |
+
# Remove empty string from vocabulary if present
|
54 |
+
self.vocab.discard('')
|
55 |
self.vocab = list(self.vocab)
|
56 |
self.word_to_idx = {word: idx for idx, word in enumerate(self.vocab)}
|
57 |
|
58 |
def __len__(self):
|
59 |
+
return len(self.valid_dataset)
|
60 |
|
61 |
def text_to_vector(self, text):
|
62 |
+
try:
|
63 |
+
words = preprocess_text(text)
|
64 |
+
vector = torch.zeros(len(self.vocab))
|
65 |
+
for word in words:
|
66 |
+
if word in self.word_to_idx:
|
67 |
+
vector[self.word_to_idx[word]] += 1
|
68 |
+
return vector
|
69 |
+
except Exception as e:
|
70 |
+
print(f"Error converting text to vector: {e}")
|
71 |
+
return torch.zeros(len(self.vocab))
|
72 |
|
73 |
def __getitem__(self, idx):
|
74 |
+
try:
|
75 |
+
image = self.transform(self.valid_dataset[idx]['image'])
|
76 |
+
text_vector = self.text_to_vector(self.valid_dataset[idx]['prompt'])
|
77 |
+
label = self.labels[idx]
|
78 |
+
return image, text_vector, label
|
79 |
+
except Exception as e:
|
80 |
+
print(f"Error getting item at index {idx}: {e}")
|
81 |
+
# Return zero tensors as fallback
|
82 |
+
return (torch.zeros((3, 224, 224)),
|
83 |
+
torch.zeros(len(self.vocab)),
|
84 |
+
0)
|
85 |
|
86 |
# Define CNN for image processing
|
87 |
class ImageModel(nn.Module):
|
|
|
112 |
|
113 |
# Combined model
|
114 |
class CombinedModel(nn.Module):
|
115 |
+
def __init__(self, vocab_size, num_classes):
|
116 |
super(CombinedModel, self).__init__()
|
117 |
self.image_model = ImageModel()
|
118 |
self.text_model = TextMLP(vocab_size)
|
119 |
+
self.fc = nn.Linear(1024, num_classes)
|
120 |
|
121 |
def forward(self, image, text):
|
122 |
image_features = self.image_model(image)
|
|
|
124 |
combined = torch.cat((image_features, text_features), dim=1)
|
125 |
return self.fc(combined)
|
126 |
|
127 |
+
# Create dataset instance
|
128 |
+
print("Creating dataset...")
|
129 |
custom_dataset = CustomDataset(dataset)
|
130 |
+
print(f"Vocabulary size: {len(custom_dataset.vocab)}")
|
131 |
+
print(f"Number of valid samples: {len(custom_dataset)}")
|
132 |
+
|
133 |
+
# Create model
|
134 |
+
num_classes = len(custom_dataset.label_encoder.classes_)
|
135 |
+
model = CombinedModel(len(custom_dataset.vocab), num_classes)
|
136 |
|
137 |
def get_recommendations(image):
|
138 |
model.eval()
|
|
|
144 |
])
|
145 |
image_tensor = transform(image).unsqueeze(0)
|
146 |
|
147 |
+
# Create dummy text vector
|
148 |
dummy_text = torch.zeros((1, len(custom_dataset.vocab)))
|
149 |
|
150 |
# Get model output
|
|
|
154 |
# Get recommended images and their information
|
155 |
recommendations = []
|
156 |
for idx in indices[0]:
|
157 |
+
try:
|
158 |
+
recommended_image = custom_dataset.valid_dataset[idx.item()]['image']
|
159 |
+
model_name = custom_dataset.valid_dataset[idx.item()]['Model']
|
160 |
+
recommendations.append((recommended_image, f"{model_name}"))
|
161 |
+
except Exception as e:
|
162 |
+
print(f"Error getting recommendation for index {idx}: {e}")
|
163 |
+
continue
|
164 |
|
165 |
return recommendations
|
166 |
|
|
|
174 |
)
|
175 |
|
176 |
# Launch the app
|
177 |
+
if __name__ == "__main__":
|
178 |
+
interface.launch()
|