cleanup

keras-encoding/cnn.py

@@ -73,23 +73,26 @@
         embedding_matrix[i] = embedding_vector
 
 from keras.layers import Embedding, Input, Dense, Flatten, Conv1D
-from keras.layers import MaxPooling1D
+from keras.layers import MaxPooling1D, Dropout
 from keras.models import Model
 
 embedding_layer = Embedding(len(word_index) + 1,
                             embedding_dim,
                             weights=[embedding_matrix],
                             input_length=config.max_sequence_length,
-                            trainable=True)
+                            trainable=False)
 
 sequence_input = Input(shape=(config.max_sequence_length,), dtype='int32')
 embedded_sequences = embedding_layer(sequence_input)
-x = Conv1D(128, 5, activation='relu')(embedded_sequences)
-x = MaxPooling1D(5)(x)
+x = Dropout(0.3)(embedded_sequences)
 x = Conv1D(128, 5, activation='relu')(x)
-x = MaxPooling1D(5)(x)  # global max pooling
+x = MaxPooling1D(5)(x)
+x = Dropout(0.3)(x)
+#x = Conv1D(128, 5, activation='relu')(x)
+#x = MaxPooling1D(5)(x)  # global max pooling
 x = Flatten()(x)
 x = Dense(128, activation='relu')(x)
+x = Dropout(0.3)(x)
 preds = Dense(num_labels, activation='softmax')(x)
 
 model = Model(sequence_input, preds)

keras-mlp/dropout.py

@@ -41,6 +41,3 @@
 # Fit the model
 model.fit(X_train, y_train, validation_data=(X_test, y_test),
         callbacks=[WandbKerasCallback()], epochs=config.epochs)
-
-# Save the run
-model.save(run.dir+'/two-layer.h5')

keras-mlp/mlp.py

@@ -30,9 +30,7 @@
 # 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'])
@@ -41,6 +39,3 @@
 # Fit the model
 model.fit(X_train, y_train, validation_data=(X_test, y_test),
         callbacks=[WandbKerasCallback()], epochs=config.epochs)
-
-# Save the run
-model.save(run.dir+'/two-layer.h5')

keras-perceptron/perceptron.py

@@ -21,14 +21,14 @@
 y_test = np_utils.to_categorical(y_test)
 
 num_classes = y_train.shape[1]
-print(y_train[1])
+
 # create model
 model=Sequential()
 model.add(Flatten(input_shape=(img_width,img_height)))
-model.add(Dense(num_classes, activation='softmax', kernel_initializer='zeros'))
+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),
-                    batch_size=200,callbacks=[WandbKerasCallback()])
+                    callbacks=[WandbKerasCallback()])

movie-reviews-lstm/lstm.py

@@ -2,7 +2,7 @@
 import numpy
 from keras.datasets import imdb
 from keras.models import Sequential
-from keras.layers import Dense
+from keras.layers import Dense, Conv1D, MaxPooling1D
 from keras.layers import LSTM
 from keras.layers.embeddings import Embedding
 from keras.preprocessing import sequence
@@ -20,12 +20,15 @@
 X_train = sequence.pad_sequences(X_train, maxlen=config.max_length)
 X_test = sequence.pad_sequences(X_test, maxlen=config.max_length)
 # create the model
-embedding_vecor_length = 32
+#embedding_vecor_length = 32
 model = Sequential()
 model.add(Embedding(config.num_words, config.embedding_vector_length, input_length=max_review_length))
+model.add(Conv1D(32,3,activation='relu'))
+model.add(MaxPooling1D(2))
 model.add(LSTM(config.lstm_output_size))
 model.add(Dense(1, activation='sigmoid'))
 model.compile(loss='binary_crossentropy', optimizer='adam', metrics=['accuracy'])
 print(model.summary())
 model.fit(X_train, y_train, epochs=config.epochs, batch_size=config.batch_size,
-    callbacks=[WandbKerasCallback()])
+    callbacks=[WandbKerasCallback()],
+    validation_data=(X_test, y_test))