translation in lstm

Author: l2k2

keras-cnn/convolution.ipynb

@@ -12,7 +12,7 @@
 config = run.config
 config.optimizer = "adam"
 config.epochs = 50
-config.dropout = 10
+config.dropout = 0.4
 config.hidden_nodes = 100
 
 # load data

keras-cnn/maxpool.ipynb

@@ -37,7 +37,9 @@
 # create model
 model=Sequential()
 model.add(Flatten(input_shape=(img_width,img_height)))
+model.add(Dropout(0.4))
 model.add(Dense(config.hidden_nodes, activation='relu'))
+model.add(Dropout(0.4))
 model.add(Dense(num_classes, activation='softmax'))
 model.compile(loss='categorical_crossentropy', optimizer=config.optimizer,
                     metrics=['accuracy'])

keras-mlp/dropout.py

@@ -13,6 +13,10 @@
 # load data
 (X_train, y_train), (X_test, y_test) = mnist.load_data()
 
+# normalize data
+X_train = X_train.astype('float32') / 255.
+X_test = X_test.astype('float32') / 255.
+
 img_width = X_train.shape[1]
 img_height = X_train.shape[2]
 
@@ -26,12 +30,13 @@
 # create model
 model=Sequential()
 model.add(Flatten(input_shape=(img_width,img_height)))
-model.add(Dense(num_classes))
-model.compile(loss='mse', optimizer='adam',
+model.add(Dense(num_classes, activation="softmax"))
+model.compile(loss='categorical_crossentropy', optimizer='adam',
                 metrics=['accuracy'])
 
 # Fit the model
 model.fit(X_train, y_train, epochs=10, validation_data=(X_test, y_test),
                     callbacks=[WandbCallback(data_type="image", labels=labels, save_model=False)])
-
+print(model.predict(X_test[:10]))
+model.save('model.h5')
 

keras-mlp/mlp.py

@@ -37,5 +37,4 @@
                 metrics=['accuracy'])
 
 # Fit the model
-model.fit(X_train, y_train, epochs=config.epochs, validation_data=(X_test, y_test),
-                    callbacks=[WandbCallback(data_type="image", labels=signdata.letters)])
+model.fit(X_train, y_train, epochs=config.epochs, validation_data=(X_test, y_test), callbacks=[WandbCallback(data_type="image", labels=signdata.letters)])

keras-perceptron/perceptron-linear.py

@@ -7,7 +7,7 @@
 
 model = InceptionV3(weights='imagenet')
 
-img_path = 'elephant.jpg'
+img_path = 'crane.jpeg'
 img = image.load_img(img_path, target_size=(299, 299))
 x = image.img_to_array(img)
 x = np.expand_dims(x, axis=0)

keras-sign/perceptron.py

@@ -7,7 +7,7 @@
 
 model = ResNet50(weights='imagenet')
 
-img_path = 'elephant.jpg'
+img_path = 'crane.jpeg'
 img = image.load_img(img_path, target_size=(224, 224))
 x = image.img_to_array(img)
 x = np.expand_dims(x, axis=0)

keras-transfer/inception-inspect.py

@@ -65,12 +65,12 @@ def VGG_16(weights_path=None):
     model.summary()
     # Test pretrained model if weights exist
     if os.path.exists("vgg16_weights.h5"):
-        #im = cv2.resize(cv2.imread('elephant.jpg'), (224, 224)).astype(np.float32)
-        #im[:,:,0] -= 103.939
-        #im[:,:,1] -= 116.779
-        #im[:,:,2] -= 123.68
-        #im = im.transpose((2,0,1))
-        #im = np.expand_dims(im, axis=0)
+        im = cv2.resize(cv2.imread('elephant.jpg'), (224, 224)).astype(np.float32)
+        im[:,:,0] -= 103.939
+        im[:,:,1] -= 116.779
+        im[:,:,2] -= 123.68
+        im = im.transpose((2,0,1))
+        im = np.expand_dims(im, axis=0)
         model = VGG_16('vgg16_weights.h5')
         sgd = SGD(lr=0.1, decay=1e-6, momentum=0.9, nesterov=True)
         model.compile(optimizer=sgd, loss='categorical_crossentropy')

keras-transfer/resnet50-inspect.py

@@ -1,7 +1,10 @@
 import imdb
 import numpy as np
 from keras.preprocessing import text
+from keras.models import Sequential
+from keras.layers import Conv2D, MaxPooling2D, Dropout, Dense, Flatten
 import wandb
+from wandb.keras import WandbCallback
 from sklearn.linear_model import LogisticRegression
 
 wandb.init()
@@ -10,18 +13,21 @@
 
 (X_train, y_train), (X_test, y_test) = imdb.load_imdb()
 
-
 tokenizer = text.Tokenizer(num_words=config.vocab_size)
 tokenizer.fit_on_texts(X_train)
 X_train = tokenizer.texts_to_matrix(X_train)
 X_test = tokenizer.texts_to_matrix(X_test)
 
-bow_model = LogisticRegression()
-bow_model.fit(X_train, y_train)
+# one hot encode outputs
+y_train = np_utils.to_categorical(y_train)
+y_test = np_utils.to_categorical(y_test)
 
-pred_train = bow_model.predict(X_train)
-acc = np.sum(pred_train==y_train)/len(pred_train)
+# create model
+model=Sequential()
+model.add(Dense(2, activation="softmax", input_shape=(1000,)))
+model.compile(loss='binary_crossentropy', optimizer='adam',
+                metrics=['accuracy'])
 
-pred_test = bow_model.predict(X_test)
-val_acc = np.sum(pred_test==y_test)/len(pred_test)
-wandb.log({"val_acc": val_acc, "acc": acc})
+# Fit the model
+model.fit(X_train, y_train, epochs=10, validation_data=(X_test, y_test),
+                    callbacks=[WandbCallback(save_model=False)])

keras-transfer/vgg-inspect.py

@@ -24,6 +24,8 @@
 config.epochs = 10
 
 (X_train, y_train), (X_test, y_test) = imdb.load_imdb()
+print("Review", X_train[0])
+print("Label", y_train[0])
 
 tokenizer = text.Tokenizer(num_words=config.vocab_size)
 tokenizer.fit_on_texts(X_train)
@@ -32,6 +34,8 @@
 
 X_train = sequence.pad_sequences(X_train, maxlen=config.maxlen)
 X_test = sequence.pad_sequences(X_test, maxlen=config.maxlen)
+print(X_train.shape)
+print("After pre-processing", X_train[0])
 
 model = Sequential()
 model.add(Embedding(config.vocab_size,

lstm/imdb-classifier/imdb-bow.py

@@ -50,7 +50,7 @@
     y[i, char_indices[next_chars[i]]] = 1
 
 model = Sequential()
-model.add(SimpleRNN(128, input_shape=(config.maxlen, len(chars))))
+model.add(GRU(128, input_shape=(config.maxlen, len(chars))))
 model.add(Dense(len(chars), activation='softmax'))
 model.compile(loss='categorical_crossentropy', optimizer="rmsprop")
 

lstm/imdb-classifier/imdb-cnn.py

@@ -49,7 +49,7 @@ 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])
         else:
-            preds = model.predict(testX)
+            preds = self.model.predict(self.testX)
 
         # Generate a figure with matplotlib</font>
         figure = matplotlib.pyplot.figure( figsize=(10,10) )

lstm/text-gen/char-gen.py

@@ -15,8 +15,8 @@
 wandb.init()
 config = wandb.config
 
-config.repeated_predictions = False
-config.look_back = 20
+config.repeated_predictions = True
+config.look_back = 4
 
 def load_data(data_type="airline"):
     if data_type == "flu":
@@ -39,7 +39,7 @@ def create_dataset(dataset):
         dataY.append(dataset[i + config.look_back])
     return np.array(dataX), np.array(dataY)
 
-data = load_data()
+data = load_data("sin")
 
 # normalize data to between 0 and 1
 max_val = max(data)
@@ -60,8 +60,8 @@ def create_dataset(dataset):
 # create and fit the RNN
 model = Sequential()
 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.compile(loss='mae', optimizer='rmsprop')
+model.fit(trainX, trainY, epochs=1000, batch_size=20, validation_data=(testX, testY),  callbacks=[WandbCallback(), PlotCallback(trainX, trainY, testX, testY, config.look_back)])