test 233

skin_cancer_classification.py

@@ -29,6 +29,10 @@ Original file is located at
 import sys
 IN_COLAB = 'google.colab' in sys.modules
 
+# if IN_COLAB:
+#  !pip install pandas numpy matplotlib seaborn pillow scikit-learn tensorflow
+#  !pip install --upgrade kagglehub[pandas-datasets,hf-datasets]
+
 import kagglehub
 
 # Download latest version
@@ -180,6 +184,18 @@ df['target'] = df['dx'].map(benign_malignant_dict)
 # Preview
 print(df[['dx', 'target']].head())
 
+from sklearn.model_selection import train_test_split
+
+train_df, val_df = train_test_split(
+    df,
+    test_size=0.2,
+    stratify=df['target'],
+    random_state=42
+)
+
+print("Train size:", len(train_df))
+print("Validation size:", len(val_df))
+
 # Convert 'target' column to string type for ImageDataGenerator
 train_df['target'] = train_df['target'].astype(str)
 val_df['target'] = val_df['target'].astype(str)
@@ -198,21 +214,10 @@ plt.show()
 
 print(df['target'].value_counts())
 
-"""### Train / Validation split"""
+"""### Train / Validation split
 
-from sklearn.model_selection import train_test_split
+### Class weight (class imbalance)
-
+"""
-train_df, val_df = train_test_split(
-    df,
-    test_size=0.2,
-    stratify=df['target'],
-    random_state=42
-)
-
-print("Train size:", len(train_df))
-print("Validation size:", len(val_df))
-
-"""### Class weight (class imbalance)"""
 
 from sklearn.utils.class_weight import compute_class_weight
 
@@ -236,6 +241,12 @@ from tensorflow.keras.applications import DenseNet121
 from tensorflow.keras.applications.densenet import preprocess_input
 from tensorflow.keras.layers import Dense, GlobalAveragePooling2D
 from tensorflow.keras.models import Model
+from tensorflow.keras.optimizers import Adam
+
+gpus = tf.config.list_physical_devices('GPU')
+print("GPUs:", gpus)
+
+strategy = tf.distribute.MirroredStrategy()
 
 data_augmentation = tf.keras.Sequential([
     tf.keras.layers.RandomFlip("horizontal"),
@@ -260,7 +271,9 @@ x = Dense(512, activation='relu')(x) # Added another Dense layer
 x = Dense(256, activation='relu')(x) # Existing Dense layer
 predictions = Dense(1, activation='sigmoid')(x) # Output layer for binary classification
 
+with strategy.scope(): # Use all gpus
   model = Model(inputs=base_model.input, outputs=predictions)
+  model.compile(optimizer=Adam(learning_rate=0.0001), loss='binary_crossentropy', metrics=['accuracy'])
 
 """## 4. Data Generators
 
@@ -308,21 +321,7 @@ val_generator = val_datagen.flow_from_dataframe(
     seed=42
 )
 
-"""## 5. Compile the Model
+"""## 6. Train the Model"""
-
-I will compile the model using the Adam optimizer, binary cross-entropy loss (suitable for binary classification), and track accuracy as a metric.
-"""
-
-
-
-from tensorflow.keras.optimizers import Adam
-
-model.compile(optimizer=Adam(learning_rate=0.0001), loss='binary_crossentropy', metrics=['accuracy'])
-
-"""## 6. Train the Model
-
-I will now train the model using the prepared data generators. I'll also add callbacks for early stopping to prevent overfitting and to save the best model.
-"""
 
 from tensorflow.keras.callbacks import EarlyStopping, ModelCheckpoint
 
@@ -349,7 +348,7 @@ history = model.fit(
     class_weight=class_weights # Use class weights to handle imbalance
 )
 
-"""## X. Evaluation
+"""## 7. Evaluation
 
 ### Load best model
 """