Spaces:
Runtime error
Runtime error
Update app.py
Browse files
app.py
CHANGED
|
@@ -1,5 +1,7 @@
|
|
| 1 |
with open("app.py", "w") as f:
|
| 2 |
f.write("""
|
|
|
|
|
|
|
| 3 |
import torch
|
| 4 |
import torch.nn as nn
|
| 5 |
import torch.nn.functional as F
|
|
@@ -7,6 +9,11 @@ import gradio as gr
|
|
| 7 |
import pickle
|
| 8 |
import pandas as pd
|
| 9 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 10 |
# TabTransformer Model Tanımı
|
| 11 |
class TabTransformer(nn.Module):
|
| 12 |
def __init__(self, input_dim, num_classes=2, d_model=64, nhead=4, num_layers=3, dropout=0.1):
|
|
@@ -25,102 +32,122 @@ class TabTransformer(nn.Module):
|
|
| 25 |
|
| 26 |
def forward(self, x):
|
| 27 |
x = self.embedding(x)
|
| 28 |
-
x = x.unsqueeze(0)
|
| 29 |
x = self.transformer_encoder(x)
|
| 30 |
-
x = x.squeeze(0)
|
| 31 |
return self.fc(x)
|
| 32 |
|
| 33 |
-
#
|
| 34 |
categorical_features = ['Multifocal_PVC', 'Nonsustained_VT', 'gender', 'HTN', 'DM', 'Fullcompansasion']
|
| 35 |
numeric_features = ['pvc_percent', 'PVCQRS', 'EF', 'Age', 'PVC_Prematurity_index', 'QRS_ratio',
|
| 36 |
-
|
| 37 |
-
|
| 38 |
|
| 39 |
-
# Mean değerleri ile varsayılanlar
|
| 40 |
numeric_means = {
|
| 41 |
-
'pvc_percent': 11.96,
|
| 42 |
-
'
|
| 43 |
-
'
|
| 44 |
-
'
|
| 45 |
-
'PVC_Prematurity_index': 0.6158,
|
| 46 |
-
'QRS_ratio': 1.933,
|
| 47 |
-
'mean_HR': 71.28,
|
| 48 |
-
'symptom_duration': 14.91,
|
| 49 |
-
'QTc_sinus': 425.0,
|
| 50 |
-
'PVCCI_dispersion': 57.1,
|
| 51 |
-
'CI_variability': 22.98,
|
| 52 |
-
'PVC_Peak_QRS_duration': 76.13,
|
| 53 |
-
'PVCCI': 513.4,
|
| 54 |
'PVC_Compansatory_interval': 1044
|
| 55 |
}
|
| 56 |
|
| 57 |
-
|
| 58 |
-
|
| 59 |
-
|
| 60 |
-
scaler_path = "trans_scaler.pkl"
|
| 61 |
-
|
| 62 |
-
# Model tanımı
|
| 63 |
-
input_dim = len(categorical_features) + len(numeric_features) # Toplam giriş boyutu
|
| 64 |
-
model = TabTransformer(input_dim=input_dim)
|
| 65 |
-
model.load_state_dict(torch.load(model_path, map_location=torch.device('cpu'))) # Model ağırlıklarını yükle
|
| 66 |
-
model.eval() # Değerlendirme moduna al
|
| 67 |
-
|
| 68 |
-
# Scaler yükleme
|
| 69 |
-
with open(scaler_path, "rb") as f:
|
| 70 |
-
scaler = pickle.load(f)
|
| 71 |
|
| 72 |
-
|
| 73 |
-
|
| 74 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 75 |
|
| 76 |
def predict(*inputs):
|
|
|
|
|
|
|
|
|
|
| 77 |
try:
|
| 78 |
-
# Girdileri
|
| 79 |
cat_inputs = inputs[:len(categorical_features)]
|
| 80 |
num_inputs = inputs[len(categorical_features):]
|
| 81 |
-
|
| 82 |
-
# Kategorik
|
| 83 |
cat_data = [1 if val == "Yes" else 0 for val in cat_inputs]
|
| 84 |
-
|
| 85 |
-
# Sayısal
|
| 86 |
num_data = [float(val) for val in num_inputs]
|
| 87 |
-
|
| 88 |
-
#
|
| 89 |
data = pd.DataFrame([cat_data + num_data], columns=categorical_features + numeric_features)
|
|
|
|
|
|
|
| 90 |
scaled_data = scaler.transform(data)
|
| 91 |
-
|
| 92 |
-
#
|
| 93 |
-
tensor_data = torch.FloatTensor(scaled_data)
|
| 94 |
with torch.no_grad():
|
|
|
|
| 95 |
logits = model(tensor_data)
|
| 96 |
probabilities = F.softmax(logits, dim=1).numpy()
|
| 97 |
-
|
| 98 |
return {
|
| 99 |
"Probability of Response": float(probabilities[0][0]),
|
| 100 |
"Probability of Non-Response": float(probabilities[0][1])
|
| 101 |
}
|
| 102 |
except Exception as e:
|
| 103 |
print(f"Tahmin hatası: {str(e)}")
|
| 104 |
-
|
| 105 |
|
| 106 |
# Gradio arayüzü
|
| 107 |
-
|
| 108 |
-
[gr.Dropdown(choices=['Yes', 'No'], label=
|
| 109 |
-
[gr.Number(label=
|
| 110 |
-
|
| 111 |
-
outputs = gr.Label(label="Prediction")
|
| 112 |
-
|
| 113 |
-
|
| 114 |
-
|
| 115 |
-
|
| 116 |
-
|
| 117 |
-
|
| 118 |
-
|
|
|
|
| 119 |
|
| 120 |
-
# Spaces için başlatma
|
| 121 |
if __name__ == "__main__":
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 122 |
try:
|
| 123 |
-
|
|
|
|
| 124 |
except Exception as e:
|
| 125 |
print(f"Arayüz başlatma hatası: {str(e)}")
|
| 126 |
-
|
|
|
|
| 1 |
with open("app.py", "w") as f:
|
| 2 |
f.write("""
|
| 3 |
+
import os
|
| 4 |
+
import sys
|
| 5 |
import torch
|
| 6 |
import torch.nn as nn
|
| 7 |
import torch.nn.functional as F
|
|
|
|
| 9 |
import pickle
|
| 10 |
import pandas as pd
|
| 11 |
|
| 12 |
+
# Debug için print fonksiyonları
|
| 13 |
+
print("Python version:", sys.version)
|
| 14 |
+
print("Current working directory:", os.getcwd())
|
| 15 |
+
print("Directory contents:", os.listdir())
|
| 16 |
+
|
| 17 |
# TabTransformer Model Tanımı
|
| 18 |
class TabTransformer(nn.Module):
|
| 19 |
def __init__(self, input_dim, num_classes=2, d_model=64, nhead=4, num_layers=3, dropout=0.1):
|
|
|
|
| 32 |
|
| 33 |
def forward(self, x):
|
| 34 |
x = self.embedding(x)
|
| 35 |
+
x = x.unsqueeze(0)
|
| 36 |
x = self.transformer_encoder(x)
|
| 37 |
+
x = x.squeeze(0)
|
| 38 |
return self.fc(x)
|
| 39 |
|
| 40 |
+
# Özellikler
|
| 41 |
categorical_features = ['Multifocal_PVC', 'Nonsustained_VT', 'gender', 'HTN', 'DM', 'Fullcompansasion']
|
| 42 |
numeric_features = ['pvc_percent', 'PVCQRS', 'EF', 'Age', 'PVC_Prematurity_index', 'QRS_ratio',
|
| 43 |
+
'mean_HR', 'symptom_duration', 'QTc_sinus', 'PVCCI_dispersion',
|
| 44 |
+
'CI_variability', 'PVC_Peak_QRS_duration', 'PVCCI', 'PVC_Compansatory_interval']
|
| 45 |
|
|
|
|
| 46 |
numeric_means = {
|
| 47 |
+
'pvc_percent': 11.96, 'PVCQRS': 155.1, 'EF': 59.93, 'Age': 52.19,
|
| 48 |
+
'PVC_Prematurity_index': 0.6158, 'QRS_ratio': 1.933, 'mean_HR': 71.28,
|
| 49 |
+
'symptom_duration': 14.91, 'QTc_sinus': 425.0, 'PVCCI_dispersion': 57.1,
|
| 50 |
+
'CI_variability': 22.98, 'PVC_Peak_QRS_duration': 76.13, 'PVCCI': 513.4,
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 51 |
'PVC_Compansatory_interval': 1044
|
| 52 |
}
|
| 53 |
|
| 54 |
+
# Global değişkenler
|
| 55 |
+
model = None
|
| 56 |
+
scaler = None
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 57 |
|
| 58 |
+
def load_model_and_scaler():
|
| 59 |
+
global model, scaler
|
| 60 |
+
try:
|
| 61 |
+
print("Model ve scaler yükleniyor...")
|
| 62 |
+
|
| 63 |
+
# Model dosyası kontrolü
|
| 64 |
+
model_path = "tabtransformer_model.pth"
|
| 65 |
+
if not os.path.exists(model_path):
|
| 66 |
+
raise FileNotFoundError(f"Model dosyası bulunamadı: {model_path}")
|
| 67 |
+
|
| 68 |
+
# Scaler dosyası kontrolü
|
| 69 |
+
scaler_path = "trans_scaler.pkl"
|
| 70 |
+
if not os.path.exists(scaler_path):
|
| 71 |
+
raise FileNotFoundError(f"Scaler dosyası bulunamadı: {scaler_path}")
|
| 72 |
+
|
| 73 |
+
# Model yükleme
|
| 74 |
+
input_dim = len(categorical_features) + len(numeric_features)
|
| 75 |
+
model = TabTransformer(input_dim=input_dim)
|
| 76 |
+
model.load_state_dict(torch.load(model_path, map_location='cpu'))
|
| 77 |
+
model.eval()
|
| 78 |
+
|
| 79 |
+
# Scaler yükleme
|
| 80 |
+
with open(scaler_path, 'rb') as f:
|
| 81 |
+
scaler = pickle.load(f)
|
| 82 |
+
|
| 83 |
+
print("Model ve scaler başarıyla yüklendi!")
|
| 84 |
+
return True
|
| 85 |
+
except Exception as e:
|
| 86 |
+
print(f"Model yükleme hatası: {str(e)}")
|
| 87 |
+
return False
|
| 88 |
|
| 89 |
def predict(*inputs):
|
| 90 |
+
if model is None or scaler is None:
|
| 91 |
+
return {"Error": "Model henüz yüklenmedi"}
|
| 92 |
+
|
| 93 |
try:
|
| 94 |
+
# Girdileri ayır
|
| 95 |
cat_inputs = inputs[:len(categorical_features)]
|
| 96 |
num_inputs = inputs[len(categorical_features):]
|
| 97 |
+
|
| 98 |
+
# Kategorik verileri dönüştür
|
| 99 |
cat_data = [1 if val == "Yes" else 0 for val in cat_inputs]
|
| 100 |
+
|
| 101 |
+
# Sayısal verileri dönüştür
|
| 102 |
num_data = [float(val) for val in num_inputs]
|
| 103 |
+
|
| 104 |
+
# DataFrame oluştur
|
| 105 |
data = pd.DataFrame([cat_data + num_data], columns=categorical_features + numeric_features)
|
| 106 |
+
|
| 107 |
+
# Veriyi ölçeklendir
|
| 108 |
scaled_data = scaler.transform(data)
|
| 109 |
+
|
| 110 |
+
# Tahmin
|
|
|
|
| 111 |
with torch.no_grad():
|
| 112 |
+
tensor_data = torch.FloatTensor(scaled_data)
|
| 113 |
logits = model(tensor_data)
|
| 114 |
probabilities = F.softmax(logits, dim=1).numpy()
|
| 115 |
+
|
| 116 |
return {
|
| 117 |
"Probability of Response": float(probabilities[0][0]),
|
| 118 |
"Probability of Non-Response": float(probabilities[0][1])
|
| 119 |
}
|
| 120 |
except Exception as e:
|
| 121 |
print(f"Tahmin hatası: {str(e)}")
|
| 122 |
+
return {"Error": str(e)}
|
| 123 |
|
| 124 |
# Gradio arayüzü
|
| 125 |
+
def create_interface():
|
| 126 |
+
inputs = [gr.Dropdown(choices=['Yes', 'No'], label=feat) for feat in categorical_features]
|
| 127 |
+
inputs.extend([gr.Number(label=feat, value=numeric_means[feat]) for feat in numeric_features])
|
| 128 |
+
|
| 129 |
+
outputs = gr.Label(label="Prediction")
|
| 130 |
+
|
| 131 |
+
return gr.Interface(
|
| 132 |
+
fn=predict,
|
| 133 |
+
inputs=inputs,
|
| 134 |
+
outputs=outputs,
|
| 135 |
+
title="TabTransformer Prediction",
|
| 136 |
+
description="Enter the features to predict the response probability"
|
| 137 |
+
)
|
| 138 |
|
|
|
|
| 139 |
if __name__ == "__main__":
|
| 140 |
+
print("Uygulama başlatılıyor...")
|
| 141 |
+
|
| 142 |
+
# Model ve scaler'ı yükle
|
| 143 |
+
if not load_model_and_scaler():
|
| 144 |
+
print("Model yüklenemedi. Uygulama sonlandırılıyor.")
|
| 145 |
+
sys.exit(1)
|
| 146 |
+
|
| 147 |
+
# Arayüzü oluştur ve başlat
|
| 148 |
try:
|
| 149 |
+
demo = create_interface()
|
| 150 |
+
demo.launch(server_name="0.0.0.0", server_port=7860)
|
| 151 |
except Exception as e:
|
| 152 |
print(f"Arayüz başlatma hatası: {str(e)}")
|
| 153 |
+
sys.exit(1)""")
|