Feature/python.paddle.v2

demo/mnist/api_train.py

@@ -6,42 +6,33 @@
 
 The user api could be simpler and carefully designed.
 """
-import py_paddle.swig_paddle as api
-from py_paddle import DataProviderConverter
-import paddle.trainer.PyDataProvider2 as dp
-import numpy as np
 import random
+
+import paddle.v2 as paddle
+import py_paddle.swig_paddle as api
+
 from mnist_util import read_from_mnist
-from paddle.trainer_config_helpers import *
 
 
 def optimizer_config():
-    settings(
+    paddle.config.settings(
         learning_rate=1e-4,
-        learning_method=AdamOptimizer(),
+        learning_method=paddle.config.AdamOptimizer(),
         batch_size=1000,
-        model_average=ModelAverage(average_window=0.5),
-        regularization=L2Regularization(rate=0.5))
+        model_average=paddle.config.ModelAverage(average_window=0.5),
+        regularization=paddle.config.L2Regularization(rate=0.5))
 
 
 def network_config():
-    imgs = data_layer(name='pixel', size=784)
-    hidden1 = fc_layer(input=imgs, size=200)
-    hidden2 = fc_layer(input=hidden1, size=200)
-    inference = fc_layer(input=hidden2, size=10, act=SoftmaxActivation())
-    cost = classification_cost(
-        input=inference, label=data_layer(
+    imgs = paddle.config.data_layer(name='pixel', size=784)
+    hidden1 = paddle.config.fc_layer(input=imgs, size=200)
+    hidden2 = paddle.config.fc_layer(input=hidden1, size=200)
+    inference = paddle.config.fc_layer(
+        input=hidden2, size=10, act=paddle.config.SoftmaxActivation())
+    cost = paddle.config.classification_cost(
+        input=inference, label=paddle.config.data_layer(
             name='label', size=10))
-    outputs(cost)
-
-
-def init_parameter(network):
-    assert isinstance(network, api.GradientMachine)
-    for each_param in network.getParameters():
-        assert isinstance(each_param, api.Parameter)
-        array_size = len(each_param)
-        array = np.random.uniform(-1.0, 1.0, array_size).astype('float32')
-        each_param.getBuf(api.PARAMETER_VALUE).copyFromNumpyArray(array)
+    paddle.config.outputs(cost)
 
 
 def generator_to_batch(generator, batch_size):
@@ -73,42 +64,44 @@ def input_order_converter(generator):
 
 
 def main():
-    api.initPaddle("-use_gpu=false", "-trainer_count=4")  # use 4 cpu cores
+    paddle.raw.initPaddle("-use_gpu=false",
+                          "-trainer_count=4")  # use 4 cpu cores
 
     # get enable_types for each optimizer.
     # enable_types = [value, gradient, momentum, etc]
     # For each optimizer(SGD, Adam), GradientMachine should enable different
     # buffers.
-    opt_config_proto = parse_optimizer_config(optimizer_config)
-    opt_config = api.OptimizationConfig.createFromProto(opt_config_proto)
-    _temp_optimizer_ = api.ParameterOptimizer.create(opt_config)
+    opt_config_proto = paddle.config.parse_optimizer(optimizer_config)
+    opt_config = paddle.raw.OptimizationConfig.createFromProto(opt_config_proto)
+    _temp_optimizer_ = paddle.raw.ParameterOptimizer.create(opt_config)
     enable_types = _temp_optimizer_.getParameterTypes()
 
     # Create Simple Gradient Machine.
-    model_config = parse_network_config(network_config)
-    m = api.GradientMachine.createFromConfigProto(
+    model_config = paddle.config.parse_network(network_config)
+    m = paddle.raw.GradientMachine.createFromConfigProto(
         model_config, api.CREATE_MODE_NORMAL, enable_types)
 
     # This type check is not useful. Only enable type hint in IDE.
     # Such as PyCharm
-    assert isinstance(m, api.GradientMachine)
+    assert isinstance(m, paddle.raw.GradientMachine)
 
     # Initialize Parameter by numpy.
-    init_parameter(network=m)
+    m.randParameters()
 
     # Create Local Updater. Local means not run in cluster.
     # For a cluster training, here we can change to createRemoteUpdater
     # in future.
-    updater = api.ParameterUpdater.createLocalUpdater(opt_config)
-    assert isinstance(updater, api.ParameterUpdater)
+    updater = paddle.raw.ParameterUpdater.createLocalUpdater(opt_config)
+    assert isinstance(updater, paddle.raw.ParameterUpdater)
 
     # Initialize ParameterUpdater.
     updater.init(m)
 
     # DataProvider Converter is a utility convert Python Object to Paddle C++
     # Input. The input format is as same as Paddle's DataProvider.
-    converter = DataProviderConverter(
-        input_types=[dp.dense_vector(784), dp.integer_value(10)])
+    converter = paddle.data.DataProviderConverter(input_types=[
+        paddle.data.dense_vector(784), paddle.data.integer_value(10)
+    ])
 
     train_file = './data/raw_data/train'
     test_file = './data/raw_data/t10k'

python/CMakeLists.txt

@@ -4,11 +4,12 @@ set(OUTPUT_DIR
 file(GLOB TRAINER_PY_FILES . ./paddle/trainer/*.py)
 file(GLOB HELPERS_PY_FILES . ./paddle/trainer_config_helpers/*.py)
 file(GLOB UTILS_PY_FILES . ./paddle/utils/*.py)
-
+file(GLOB V2_PY_FILES . ./paddle/v2/*.py)
 set(PY_FILES paddle/__init__.py
              ${TRAINER_PY_FILES}
              ${HELPERS_PY_FILES}
-             ${UTILS_PY_FILES})
+             ${UTILS_PY_FILES}
+	     ${V2_PY_FILES})
 
 configure_file(${CMAKE_CURRENT_SOURCE_DIR}/setup.py.in
     ${CMAKE_CURRENT_BINARY_DIR}/setup.py)

python/paddle/trainer_config_helpers/__init__.py

@@ -21,5 +21,6 @@
 from optimizers import *
 from attrs import *
 from config_parser_utils import *
+
 # This will enable operator overload for LayerOutput
 import layer_math

python/paddle/v2/__init__.py

@@ -0,0 +1,18 @@
+"""
+This is an experimental package for Paddle new API.
+
+Currently, we use should always use
+
+..  code-block: python
+
+    import paddle.v2 as paddle
+
+as our import statement. The API is in flux, never use this package in
+production.
+"""
+
+import py_paddle.swig_paddle as raw
+import config
+import data
+
+__all__ = ['config', 'data', 'raw']

python/paddle/v2/config.py

@@ -0,0 +1,12 @@
+from paddle.trainer_config_helpers import *
+from paddle.trainer.config_parser import parse_config as parse
+from paddle.trainer_config_helpers.config_parser_utils import \
+    parse_network_config as parse_network
+from paddle.trainer_config_helpers.config_parser_utils import \
+    parse_optimizer_config as parse_optimizer
+
+import paddle.trainer_config_helpers as tmp
+
+__all__ = ['parse', 'parse_network', 'parse_optimizer']
+
+__all__.extend(filter(lambda x: x[:2] != '__', dir(tmp)))

python/paddle/v2/data.py

@@ -0,0 +1,11 @@
+from paddle.trainer.PyDataProvider2 import *
+from py_paddle.dataprovider_converter import DataProviderConverter
+
+__all__ = [
+    'dense_vector', 'dense_vector_sequence', 'dense_vector_sub_sequence',
+    'integer_value', 'integer_sequence', 'integer_value_sub_sequence',
+    'sparse_binary_vector', 'sparse_binary_vector_sequence',
+    'sparse_binary_vector_sub_sequence', 'sparse_vector',
+    'sparse_vector_sequence', 'sparse_vector_sub_sequence', 'provider',
+    'CacheType', 'DataProviderConverter'
+]

python/setup.py.in

@@ -4,7 +4,8 @@ packages=['paddle',
           'paddle.proto',
           'paddle.trainer',
           'paddle.trainer_config_helpers',
-          'paddle.utils']
+          'paddle.utils',
+	  'paddle.v2']
 
 setup(name='paddle',
       version='${PADDLE_VERSION}',