scripts/CNN_cantilever.py

@@ -18,27 +18,10 @@ from keras.callbacks import ModelCheckpoint, LearningRateScheduler
 
 np.random.seed(123)  # for reproducibility
 
-#def _get_available_gpus():
-#    """Get a list of available gpu devices (formatted as strings).
-#
-#    # Returns
-#        A list of available GPU devices.
-#    """
-#    #global _LOCAL_DEVICES
-#    if tfback._LOCAL_DEVICES is None:
-#        devices = tf.config.list_logical_devices()
-#        tfback._LOCAL_DEVICES = [x.name for x in devices]
-#    return [x for x in tfback._LOCAL_DEVICES if 'device:gpu' in x.lower()]
-#
-#
-#tfback._get_available_gpus = _get_available_gpus
-#tfback._get_available_gpus()
-#
-##tf.config.experimental_list_devices()
-#tf.config.list_logical_devices()
-
 ################ Let us import the data to train our ANN ###########################
 
+#This is not uploaded on gitlab because of heavy size of training data
+
 df = pd.read_csv('/home/users/sdeshpande/Deep_CNN/CSV_files/topface_converged_nov.csv', names=['f', 'u'],
                  header=None)
 
@@ -117,13 +100,9 @@ conv9 = Conv3D(64, (3,3, 3), activation='relu', padding='same',data_format='chan
 conv9 = Conv3D(64, (3,3, 3), activation='relu', padding='same',data_format='channels_first')(conv9)
 conv9 = Conv3D(3, (1,1, 1),  activation= None, padding= 'same',data_format='channels_first')(conv9)
 conv9 = Lambda(lambda x: x[:,:,2:30,2:14,2:14])(conv9)
-#conv9 = Conv3D(3, (5, 5, 5), activation= None, padding='valid',data_format='channels_first')(conv9)
 
 UNET = Model(inputs=inputs, outputs=conv9)
 
-#Load the best weights obtained previously
-#UNET.load_weights("saved_models/topfacelrs_25_09_2.h5")
-
 #Let's use a decaying learning rate as per the below law 
 def lr_scheduler(epoch, lr):
     k = 0.001
@@ -170,39 +149,6 @@ plt.ylabel('Loss')
 plt.legend()
 plt.savefig('/home/users/sdeshpande/Deep_CNN/To_clusteroct/plots/topfacee_15_12.png')
 
-#################  Save the Mean Absolute Error and export text file ############
-
-e = np.zeros((n_test,1))
-data_predicted = np.zeros((12096 * n_test, 3))
-
-for i in range(n_test):
-      pred = UNET.predict(X_test[i].reshape(1,3,28,12,12))
-      pred = pred.reshape(12096)
-      actual = Y_test[i].reshape(12096)
-      e[i] = np.mean(abs(pred - actual))
-
-      data_predicted[12096 * i:12096 * (i + 1), 0] = X_test[i].reshape(12096)
-      data_predicted[12096 * i:12096 * (i + 1), 1] = pred
-      data_predicted[12096 * i:12096 * (i + 1), 2] = actual
-
-MAE = np.mean(e) #came out as 0.00219 for
-std = np.std(e,ddof=1) #To get unbiased estimate
-
-
-#Let's get single prediction time
-tic_test = time.time()
-single_out = UNET.predict(X_test[0].reshape(1,3,28,12,12))
-toc_test = time.time()
-
-output = np.array([MAE,std,toc_train-tic_train,toc_test-tic_test])
-
-np.savetxt("/home/users/sdeshpande/Deep_CNN/To_clusteroct/Txt_outputs/MST_topfacee_15_12.txt", output , delimiter=",")
-np.savetxt("/home/users/sdeshpande/Deep_CNN/To_clusteroct/Predictions/topface_15_12.csv", data_predicted, delimiter=",")
-
-####### Dump all the test predictions in the a csv file #####
-
-#Here I can even use the saved weights file to later get predictions from local server
-