Remove profiling code

Author: vanpelt

examples/keras-perf/cnn.py

@@ -46,10 +46,11 @@
 model.add(tf.keras.layers.Dense(config.dense_layer_size, activation='relu'))
 model.add(tf.keras.layers.Dense(num_classes, activation='softmax'))
 
-run_options = tf.RunOptions(trace_level=tf.RunOptions.FULL_TRACE)
-run_metadata = tf.RunMetadata()
+# optional profiling setup...
+#run_options = tf.RunOptions(trace_level=tf.RunOptions.FULL_TRACE)
+#run_metadata = tf.RunMetadata()
+# options=run_options, run_metadata=run_metadata,
 model.compile(loss='categorical_crossentropy', optimizer='adam',
-              options=run_options, run_metadata=run_metadata,
               metrics=['accuracy'])
 # log the number of total parameters
 config.total_params = model.count_params()
@@ -61,10 +62,10 @@
                      tf.keras.callbacks.TensorBoard(log_dir=wandb.run.dir)])
 model.save('cnn.h5')
 
-# Write performance profile
-tl = timeline.Timeline(run_metadata.step_stats)
-with open('profile.json', 'w') as f:
-    f.write(tl.generate_chrome_trace_format())
+# optional profiling setup continued
+#tl = timeline.Timeline(run_metadata.step_stats)
+# with open('profile.json', 'w') as f:
+#    f.write(tl.generate_chrome_trace_format())
 
 # Convert to TensorFlow Lite model.
 converter = tf.lite.TFLiteConverter.from_keras_model_file('cnn.h5')

examples/lstm/imdb-classifier/imdb-bow.py

@@ -1,4 +1,4 @@
-import imdb
+import util
 import numpy as np
 import tensorflow as tf
 from tensorflow.keras.preprocessing import text
@@ -8,7 +8,7 @@
 config = wandb.config
 config.vocab_size = 1000
 
-(X_train, y_train), (X_test, y_test) = imdb.load_imdb()
+(X_train, y_train), (X_test, y_test) = util.load_imdb()
 
 tokenizer = text.Tokenizer(num_words=config.vocab_size)
 tokenizer.fit_on_texts(X_train)

examples/lstm/imdb-classifier/imdb-cnn.py

@@ -1,11 +1,10 @@
 import wandb
-import imdb
 import numpy as np
 import tensorflow as tf
 from tensorflow.keras.preprocessing import text, sequence
 from tensorflow.python.client import device_lib
 from tensorflow.keras.layers import LSTM, GRU, CuDNNLSTM, CuDNNGRU
-
+from tensorflow.keras.datasets import imdb
 
 # set parameters:
 wandb.init()
@@ -19,12 +18,7 @@
 config.hidden_dims = 250
 config.epochs = 10
 
-(X_train, y_train), (X_test, y_test) = imdb.load_imdb()
-print("Tokenizing text")
-tokenizer = text.Tokenizer(num_words=config.vocab_size)
-tokenizer.fit_on_texts(X_train)
-X_train = tokenizer.texts_to_sequences(X_train)
-X_test = tokenizer.texts_to_sequences(X_test)
+(X_train, y_train), (X_test, y_test) = imdb.load_data(num_words=config.vocab_size)
 
 X_train = sequence.pad_sequences(X_train, maxlen=config.maxlen)
 X_test = sequence.pad_sequences(X_test, maxlen=config.maxlen)
@@ -60,4 +54,4 @@
 model.fit(X_train, y_train,
           batch_size=config.batch_size,
           epochs=config.epochs,
-          validation_data=(X_test, y_test), callbacks=[wandb.keras.WandbCallback()])
+          validation_data=(X_test, y_test), callbacks=[wandb.keras.WandbCallback(save_mode=False)])

examples/lstm/imdb-classifier/imdb-embedding.py

@@ -5,7 +5,8 @@
 from tensorflow.keras.preprocessing import text, sequence
 from tensorflow.python.client import device_lib
 from tensorflow.keras.layers import LSTM, GRU, CuDNNLSTM, CuDNNGRU
-import imdb
+from tensorflow.keras.datasets import imdb
+import util
 import os
 
 # set parameters:
@@ -20,7 +21,7 @@
 config.hidden_dims = 100
 config.epochs = 10
 
-(X_train, y_train), (X_test, y_test) = imdb.load_imdb()
+(X_train, y_train), (X_test, y_test) = util.load_imdb()
 
 if not os.path.exists("glove.6B.100d.txt"):
     print("Downloading glove embeddings...")
@@ -75,4 +76,4 @@
 model.fit(X_train, y_train,
           batch_size=config.batch_size,
           epochs=config.epochs,
-          validation_data=(X_test, y_test), callbacks=[wandb.keras.WandbCallback(input_type="time")])
+          validation_data=(X_test, y_test), callbacks=[wandb.keras.WandbCallback(save_model=False)])

examples/lstm/imdb-classifier/imdb-lstm.py

@@ -1,10 +1,10 @@
 import wandb
-import imdb
 import numpy as np
 import tensorflow as tf
 from tensorflow.keras.layers import LSTM, GRU, CuDNNLSTM, CuDNNGRU
 from tensorflow.python.client import device_lib
 from tensorflow.keras.preprocessing import text, sequence
+from tensorflow.keras.datasets import imdb
 
 # set parameters:
 wandb.init()
@@ -19,12 +19,17 @@
 config.epochs = 10
 
 # Load and tokenize input
-(X_train, y_train), (X_test, y_test) = imdb.load_imdb()
-print("Tokenizing text")
+(X_train, y_train), (X_test, y_test) = imdb.load_data(num_words=config.vocab_size)
+
+# Example of manual data loading
+"""
+import util
+(X_train, y_train), (X_test, y_test) = util.load_imdb()
 tokenizer = text.Tokenizer(num_words=config.vocab_size)
 tokenizer.fit_on_texts(X_train)
 X_train = tokenizer.texts_to_sequences(X_train)
 X_test = tokenizer.texts_to_sequences(X_test)
+"""
 
 # Ensure all input is the same size
 X_train = sequence.pad_sequences(
@@ -51,6 +56,6 @@
 model.fit(X_train, y_train,
           batch_size=config.batch_size,
           epochs=config.epochs,
-          validation_data=(X_test, y_test), callbacks=[wandb.keras.WandbCallback()])
+          validation_data=(X_test, y_test), callbacks=[wandb.keras.WandbCallback(save_model=False)])
 
 model.save("seniment.h5")

examples/lstm/imdb-classifier/imdb.py

@@ -0,0 +1,47 @@
+import os
+import subprocess
+import numpy as np
+
+
+def load_imdb():
+    if not os.path.exists("./aclImdb"):
+        print("Downloading imdb dataset...")
+        subprocess.check_output(
+            "curl -SL http://ai.stanford.edu/~amaas/data/sentiment/aclImdb_v1.tar.gz | tar xz", shell=True)
+
+    if not os.path.exists('./aclImdb/cache.npz'):
+        X_train = []
+        y_train = []
+
+        path = './aclImdb/train/pos/'
+        X_train.extend([open(path + f).read()
+                        for f in os.listdir(path) if f.endswith('.txt')])
+        y_train.extend([1 for _ in range(12500)])
+
+        path = './aclImdb/train/neg/'
+        X_train.extend([open(path + f).read()
+                        for f in os.listdir(path) if f.endswith('.txt')])
+        y_train.extend([0 for _ in range(12500)])
+
+        X_test = []
+        y_test = []
+
+        path = './aclImdb/test/pos/'
+        X_test.extend([open(path + f).read()
+                       for f in os.listdir(path) if f.endswith('.txt')])
+        y_test.extend([1 for _ in range(12500)])
+
+        path = './aclImdb/test/neg/'
+        X_test.extend([open(path + f).read()
+                       for f in os.listdir(path) if f.endswith('.txt')])
+        y_test.extend([0 for _ in range(12500)])
+
+        np.savez('./aclImdb/cache.npz', X_train=X_train, y_train=y_train,
+                 X_test=X_test, y_test=y_test)
+
+    cached = np.load('./aclImdb/cache.npz')
+    return (cached['X_train'], cached['y_train']), (cached['X_test'], cached['y_test'])
+
+
+if __name__ == '__main__':
+    load_imdb()