import tensorflow as tf from tensorflow.keras.models import Sequential from tensorflow.keras.layers import Dense, Reshape, Conv1D, MaxPooling1D, Dropout, Flatten from tensorflow.keras.optimizers.legacy import Adam from tensorflow.keras import regularizers # Hyperparameters EPOCHS = args.epochs or 100 LEARNING_RATE = args.learning_rate or 0.005 ENSURE_DETERMINISM = args.ensure_determinism BATCH_SIZE = args.batch_size or 32 # Shuffle and batch datasets if not ENSURE_DETERMINISM: train_dataset = train_dataset.shuffle(buffer_size=BATCH_SIZE * 4) train_dataset = train_dataset.batch(BATCH_SIZE, drop_remainder=False) validation_dataset = validation_dataset.batch(BATCH_SIZE, drop_remainder=False) # Model architecture model = Sequential() model.add(Reshape((int(input_length / 13), 13), input_shape=(input_length, ))) # Conv1D with L2 Regularization model.add(Conv1D(8, kernel_size=3, padding='same', activation='relu', kernel_regularizer=regularizers.l2(0.01))) model.add(MaxPooling1D(pool_size=2, strides=2, padding='same')) model.add(Dropout(0.25)) # Conv1D with L1 Regularization model.add(Conv1D(16, kernel_size=3, padding='same', activation='relu', kernel_regularizer=regularizers.l1(0.01))) model.add(MaxPooling1D(pool_size=2, strides=2, padding='same')) model.add(Dropout(0.25)) # Flatten layer model.add(Flatten()) # Dense with combined L1 and L2 Regularization model.add(Dense(classes, name='y_pred', activation='softmax', kernel_regularizer=regularizers.l1_l2(l1=0.01, l2=0.01))) # Optimizer opt = Adam(learning_rate=LEARNING_RATE, beta_1=0.9, beta_2=0.999) # Compile model model.compile(loss='categorical_crossentropy', optimizer=opt, metrics=['accuracy']) # Callbacks callbacks = [] callbacks.append(BatchLoggerCallback(BATCH_SIZE, train_sample_count, epochs=EPOCHS, ensure_determinism=ENSURE_DETERMINISM)) # Train the neural network model.fit(train_dataset, epochs=EPOCHS, validation_data=validation_dataset, verbose=2, callbacks=callbacks) # Optional: Disable per-channel quantization disable_per_channel_quantization = False