Class updates and extras

vanpelt · vanpelt · commit 160bee7d747e · 2019-03-26T23:33:03.000-07:00
diff --git a/keras-audio/gru-composer.py b/keras-audio/gru-composer.py
@@ -135,7 +135,7 @@ class Midi(Callback):
     Callback for sampling a midi file
     """
 
-    def sample(self, preds, temperature=1.0):
+    def sample(self, preds, temperature=1):
         # helper function to sample an index from a probability array
         preds = np.asarray(preds).astype('float64')
         preds = np.log(preds) / temperature
@@ -163,7 +163,7 @@ def generate_notes(self, network_input, pitchnames, n_vocab):
 
             prediction = model.predict(prediction_input, verbose=0)
 
-            index = self.sample(prediction[0], temperature=0.5)#np.argmax
+            index = self.sample(prediction[0], temperature=0.5)  # np.argmax
             result = int_to_note[index]
             prediction_output.append(result)
 
diff --git a/keras-autoencoder/variational_autoencoder.py b/keras-autoencoder/variational_autoencoder.py
@@ -17,6 +17,8 @@
 # reparameterization trick
 # instead of sampling from Q(z|X), sample eps = N(0,I)
 # z = z_mean + sqrt(var)*eps
+
+
 def sampling(args):
     """Reparameterization trick by sampling fr an isotropic unit Gaussian.
     # Arguments:
@@ -32,6 +34,7 @@ def sampling(args):
     epsilon = K.random_normal(shape=(batch, dim))
     return z_mean + K.exp(0.5 * z_log_var) * epsilon
 
+
 # MNIST dataset
 (x_train, y_train), (x_test, y_test) = mnist.load_data()
 
@@ -78,7 +81,7 @@ def sampling(args):
 data = (x_test, y_test)
 
 reconstruction_loss = binary_crossentropy(inputs,
-                                              outputs)
+                                          outputs)
 
 reconstruction_loss *= original_dim
 kl_loss = 1 + z_log_var - K.square(z_mean) - K.exp(z_log_var)
@@ -88,10 +91,10 @@ def sampling(args):
 vae.add_loss(vae_loss)
 vae.compile(optimizer='adam')
 
+
 vae.fit(x_train,
-            epochs=epochs,
-            batch_size=batch_size,
-            validation_data=(x_test, None),
-       callbacks=[WandbCallback(), PlotCallback(encoder, decoder, (x_test, y_test))] )
+        epochs=epochs,
+        batch_size=batch_size,
+        validation_data=(x_test, None),
+        callbacks=[WandbCallback(), PlotCallback(encoder, decoder, (x_test, y_test))])
 vae.save_weights('vae_mlp_mnist.h5')
-
diff --git a/keras-cnn/cnn.py b/keras-cnn/cnn.py
@@ -4,6 +4,7 @@
 from keras.utils import np_utils
 from wandb.keras import WandbCallback
 import wandb
+import os
 
 run = wandb.init()
 config = run.config
@@ -23,31 +24,33 @@
 X_test = X_test.astype('float32')
 X_test /= 255.
 
-#reshape input data
-X_train = X_train.reshape(X_train.shape[0], config.img_width, config.img_height, 1)
-X_test = X_test.reshape(X_test.shape[0], config.img_width, config.img_height, 1)
+# reshape input data
+X_train = X_train.reshape(
+    X_train.shape[0], config.img_width, config.img_height, 1)
+X_test = X_test.reshape(
+    X_test.shape[0], config.img_width, config.img_height, 1)
 
 # one hot encode outputs
 y_train = np_utils.to_categorical(y_train)
 y_test = np_utils.to_categorical(y_test)
 num_classes = y_test.shape[1]
-labels=range(10)
+labels = range(10)
 
 # build model
 model = Sequential()
 model.add(Conv2D(32,
-    (config.first_layer_conv_width, config.first_layer_conv_height),
-    input_shape=(28, 28,1),
-    activation='relu'))
+                 (config.first_layer_conv_width, config.first_layer_conv_height),
+                 input_shape=(28, 28, 1),
+                 activation='relu'))
 model.add(MaxPooling2D(pool_size=(2, 2)))
 model.add(Flatten())
 model.add(Dense(config.dense_layer_size, activation='relu'))
 model.add(Dense(num_classes, activation='softmax'))
 
 model.compile(loss='categorical_crossentropy', optimizer='adam',
-                metrics=['accuracy'])
+              metrics=['accuracy'])
 
 
 model.fit(X_train, y_train, validation_data=(X_test, y_test),
-        epochs=config.epochs,
-        callbacks=[WandbCallback(data_type="image", save_model=False)])
+          epochs=config.epochs,
+          callbacks=[WandbCallback(data_type="image", save_model=False)])
diff --git a/keras-cnn/wandb/settings b/keras-cnn/wandb/settings
@@ -1,4 +1,4 @@
 [default]
 entity: qualcomm
-project: digits-feb1
+project: digits-mar26
 base_url: https://api.wandb.ai
diff --git a/keras-mlp/wandb/settings b/keras-mlp/wandb/settings
@@ -1,4 +1,4 @@
 [default]
 entity: qualcomm
-project: digits-feb1
+project: digits-mar26
 base_url: https://api.wandb.ai
diff --git a/keras-perceptron/wandb/settings b/keras-perceptron/wandb/settings
@@ -1,4 +1,4 @@
 [default]
 entity: qualcomm
-project: digits-feb1
+project: digits-mar26
 base_url: https://api.wandb.ai
diff --git a/time-series/plotutil.py b/time-series/plotutil.py
@@ -1,41 +1,42 @@
 import matplotlib
 matplotlib.use('Agg')
-import matplotlib.pyplot as plt
-from matplotlib.backends.backend_agg import FigureCanvasAgg as FigureCanvas
-from matplotlib.figure import Figure
-import keras
-import numpy as np
 import wandb
+import numpy as np
+import keras
+from matplotlib.figure import Figure
+from matplotlib.backends.backend_agg import FigureCanvasAgg as FigureCanvas
+import matplotlib.pyplot as plt
 
-def fig2data ( fig ):
+
+def fig2data(fig):
     """
     @brief Convert a Matplotlib figure to a 4D numpy array with RGBA channels and return it
     @param fig a matplotlib figure
     @return a numpy 3D array of RGBA values
     """
     # draw the renderer
-    fig.canvas.draw ( )
- 
+    fig.canvas.draw()
+
     # Get the RGBA buffer from the figure
-    w,h = fig.canvas.get_width_height()
-    buf = np.fromstring ( fig.canvas.tostring_argb(), dtype=np.uint8 )
-    buf.shape = ( w, h,4 )
- 
+    w, h = fig.canvas.get_width_height()
+    buf = np.fromstring(fig.canvas.tostring_argb(), dtype=np.uint8)
+    buf.shape = (w, h, 4)
+
     # canvas.tostring_argb give pixmap in ARGB mode. Roll the ALPHA channel to have it in RGBA mode
-    buf = np.roll ( buf, 3, axis = 2 )
+    buf = np.roll(buf, 3, axis=2)
     return buf
 
 
-
-def repeated_predictions(model, data, look_back, steps=100 ):
+def repeated_predictions(model, data, look_back, steps=100):
     predictions = []
     for i in range(steps):
-        input_data = data[np.newaxis,:,np.newaxis]
+        input_data = data[np.newaxis, :, np.newaxis]
         generated = model.predict(input_data)[0]
         data = np.append(data, generated)[-look_back:]
         predictions.append(generated)
     return predictions
 
+
 class PlotCallback(keras.callbacks.Callback):
     def __init__(self, trainX, trainY, testX, testY, look_back):
         self.repeat_predictions = True
@@ -44,22 +45,24 @@ def __init__(self, trainX, trainY, testX, testY, look_back):
         self.testX = testX
         self.testY = testY
         self.look_back = look_back
-        
+
     def on_epoch_end(self, epoch, logs):
         if self.repeat_predictions:
-            preds = repeated_predictions(self.model, self.trainX[-1,:,0], self.look_back, self.testX.shape[0])
+            preds = repeated_predictions(
+                self.model, self.trainX[-1, :, 0], self.look_back, self.testX.shape[0])
         else:
             preds = model.predict(testX)
 
         # Generate a figure with matplotlib</font>
-        figure = matplotlib.pyplot.figure( figsize=(10,10) )
-        plot   = figure.add_subplot ( 111 )
+        figure = matplotlib.pyplot.figure(figsize=(10, 10))
+        plot = figure.add_subplot(111)
 
-        plot.plot ( self.trainY )
-        plot.plot ( np.append(np.empty_like(self.trainY) * np.nan, self.testY))
-        plot.plot ( np.append(np.empty_like(self.trainY) * np.nan, preds))
+        plot.plot(self.trainY)
+        plot.plot(np.append(np.empty_like(self.trainY) * np.nan, self.testY))
+        plot.plot(np.append(np.empty_like(self.trainY) * np.nan, preds))
 
-        data = fig2data ( figure )
+        data = fig2data(figure)
         matplotlib.pyplot.close(figure)
 
-        wandb.log({"image": wandb.Image(data)}, commit=False)
+        if epoch % 4 == 0:
+            wandb.log({"image": wandb.Image(data)}, commit=False)
diff --git a/time-series/rnn.py b/time-series/rnn.py
@@ -18,6 +18,7 @@
 config.repeated_predictions = False
 config.look_back = 20
 
+
 def load_data(data_type="airline"):
     if data_type == "flu":
         df = pd.read_csv('flusearches.csv')
@@ -31,6 +32,8 @@ def load_data(data_type="airline"):
     return data
 
 # convert an array of values into a dataset matrix
+
+
 def create_dataset(dataset):
     dataX, dataY = [], []
     for i in range(len(dataset)-config.look_back-1):
@@ -39,12 +42,13 @@ def create_dataset(dataset):
         dataY.append(dataset[i + config.look_back])
     return np.array(dataX), np.array(dataY)
 
+
 data = load_data()
-    
+
 # normalize data to between 0 and 1
 max_val = max(data)
 min_val = min(data)
-data=(data-min_val)/(max_val-min_val)
+data = (data-min_val)/(max_val-min_val)
 
 # split into train and test sets
 split = int(len(data) * 0.70)
@@ -59,11 +63,7 @@ def create_dataset(dataset):
 
 # create and fit the RNN
 model = Sequential()
-model.add(SimpleRNN(1, input_shape=(config.look_back,1 )))
+model.add(SimpleRNN(1, input_shape=(config.look_back, 1)))
 model.compile(loss='mse', optimizer='adam')
-model.fit(trainX, trainY, epochs=1000, batch_size=1, validation_data=(testX, testY),  callbacks=[WandbCallback(), PlotCallback(trainX, trainY, testX, testY, config.look_back)])
-
-
-
-
-
+model.fit(trainX, trainY, epochs=500, batch_size=1, validation_data=(testX, testY),  callbacks=[
+          WandbCallback(), PlotCallback(trainX, trainY, testX, testY, config.look_back)])
