Merge 67b0a59 into 4dfe0f3

autokeras/__init__.py

@@ -1,4 +1,4 @@
-from autokeras.image.image_supervised import ImageClassifier, ImageRegressor
+from autokeras.image.image_supervised import ImageClassifier, ImageRegressor, PortableImageSupervised
 from autokeras.text.text_supervised import TextClassifier, TextRegressor
 from autokeras.tabular.tabular_supervised import TabularClassifier, TabularRegressor
 from autokeras.net_module import CnnModule, MlpModule

autokeras/utils.py

@@ -18,7 +18,6 @@
 from scipy.ndimage import zoom
 
 
-
 class NoImprovementError(Exception):
     def __init__(self, message):
         self.message = message
@@ -49,12 +48,11 @@ def pickle_to_file(obj, path):
     """Save the pickle file to the specified path."""
     pickle.dump(obj, open(path, 'wb'))
 
+
 # TODO cannot detect nvidia-smi in Windows normally. We need a fall back for windows
 def get_device():
     """ If CUDA is available, use CUDA device, else use CPU device.
-
     When choosing from CUDA devices, this function will choose the one with max memory available.
-
     Returns: string device name.
     """
     # TODO: could use gputil in the future
@@ -140,7 +138,7 @@ def verbose_print(new_father_id, new_graph, new_model_id):
     """Print information about the operation performed on father model to obtain current model and father's id."""
     cell_size = [24, 49]
 
-    logging.info('New Model Id - '+str(new_model_id))
+    logging.info('New Model Id - ' + str(new_model_id))
     header = ['Father Model ID', 'Added Operation']
     line = '|'.join(str(x).center(cell_size[i]) for i, x in enumerate(header))
     logging.info('\n' + '+' + '-' * len(line) + '+')
@@ -155,6 +153,7 @@ def verbose_print(new_father_id, new_graph, new_model_id):
         logging.info('|' + line + '|')
     logging.info('+' + '-' * len(line) + '+')
 
+
 def validate_xy(x_train, y_train):
     """Validate `x_train`'s type and the shape of `x_train`, `y_train`."""
     try:
@@ -264,4 +263,5 @@ def get_system():
         return Constant.SYS_LINUX
     if os.name == 'nt':
         return Constant.SYS_WINDOWS
+
     raise EnvironmentError('Unsupported environment')

examples/portable_load.py

@@ -0,0 +1,22 @@
+import os
+
+from keras.datasets import mnist
+
+from autokeras import ImageClassifier
+from autokeras.utils import pickle_from_file
+
+# Customer temp dir by your own
+TEMP_DIR = '/tmp/autokeras_U8KEOQ'
+model_file_name = os.path.join(TEMP_DIR, 'test_autokeras_model.pkl')
+
+if __name__ == '__main__':
+    (x_train, y_train), (x_test, y_test) = mnist.load_data()
+    x_train = x_train.reshape(x_train.shape + (1,))
+    x_test = x_test.reshape(x_test.shape + (1,))
+    clf = ImageClassifier(verbose=True, augment=False, path=TEMP_DIR, resume=True)
+    clf.fit(x_train, y_train, time_limit=30 * 60)
+    clf.final_fit(x_train, y_train, x_test, y_test)
+    clf.export_autokeras_model(model_file_name)
+    model = pickle_from_file(model_file_name)
+    results = model.evaluate(x_test, y_test)
+    print(results)

mkdocs/docs/start.md

@@ -17,7 +17,6 @@ You need to download the code from the GitHub repo and run the following command
     pip install -r requirements.txt
     python setup.py install
 
-=======
 
 #### How to visualize the best selected architecture ?
 
@@ -105,13 +104,25 @@ The returned values `x_train` and `y_train` are the numpy arrays,
 which can be directly feed into the `fit` function of `ImageClassifier`.
 
 #### How to export keras models?
+    clf.load_searcher().load_best_model().produce_keras_model().save('my_model.h5')
+This uses the keras function model.save() to export a single HDF5 file containing the architecture of the model, the weights of the model, the training configuration, and the state of the optimizer. See https://keras.io/getting-started/faq/#how-can-i-save-a-keras-model
+
+Note: This is being built into AutoKeras as ImageClassifier().export_keras_model() 
 
+#### how to export Portable model
     from autokeras import ImageClassifier
     clf = ImageClassifier(verbose=True, augment=False)
-    clf.load_searcher().load_best_model().produce_keras_model().save('my_model.h5')
+    clf.export_autokeras_model(model_file_name)
+The model will be stored into the path `model_file_name`. 
+
+#### How to load exported Portable model?
+    from autokeras.utils import pickle_from_file
+    model = pickle_from_file(model_file_name)
+    results = model.evaluate(x_test, y_test)
+    print(results)
+
+The model will be loaded from the path `model_file_name` and then you can use the functions listed in `PortableImageSupervised`.
 
-This uses the keras function model.save() to export a single HDF5 file containing the architecture of the model, the weights of the model, the training configuration, and the state of the optimizer. See https://keras.io/getting-started/faq/#how-can-i-save-a-keras-model
-Note: This is being built into AutoKeras as ImageClassifier().export_keras_model() 
 
 #### How to visualize keras models?
 
@@ -122,6 +133,7 @@ This is not specific to AutoKeras, however, the following will generate a .PNG v
     from keras.utils import plot_model
     plot_model(model, to_file='my_model.png')
 
+        
 # CnnGenerator tutorial
 
 `CnnGenerator` in `net_module.py` is a child class of `Networkmodule`. It can generates neural architecture with basic cnn modules
@@ -208,7 +220,7 @@ Normally, there's two place to call the MlpGenerator, one is call `MlpGenerator.
 For example, in a image classification class `ImageClassifier`, one can initialize the cnn module as:
 
 ```python
-self.mlp = MlpGenerator(loss, metric, searcher_args, path, verbose)
+mlpModule = MlpModule(loss, metric, searcher_args, path, verbose)
 ```
 Where:
 * `loss` and `metric` determines by the type of training model(classification or regression or others)
@@ -218,7 +230,7 @@ Where:
 
 Then, for the searching part, one can call:
 ```python
-self.mlp.fit(n_output_node, input_shape, train_data, test_data, time_limit=24 * 60 * 60)
+mlpModule.fit(n_output_node, input_shape, train_data, test_data, time_limit=24 * 60 * 60)
 ```
 where:
 * n_output_node: A integer value represent the number of output node in the final layer.
@@ -230,13 +242,10 @@ where:
 
 And for final testing(testing the best searched model), one can call:
 ```python
-self.mlp.final_fit(train_data, test_data, trainer_args=None, retrain=False)
+mlpModule.final_fit(train_data, test_data, trainer_args=None, retrain=False)
 ```
 where:
 * train_data: A DataLoader instance representing the training data.
 * test_data: A DataLoader instance representing the testing data.
 * trainer_args: A dictionary containing the parameters of the ModelTrainer constructor.
 * retrain: A boolean of whether reinitialize the weights of the model.
-
-
-