diff --git a/demo/mnist/api_train_v2.py b/demo/mnist/api_train_v2.py index 767dffdad82f59d73b1505586260a61f5008f1f8..8a612cbc66d54ba3a3a9a444f85626812940cf6a 100644 --- a/demo/mnist/api_train_v2.py +++ b/demo/mnist/api_train_v2.py @@ -1,4 +1,3 @@ -import numpy import paddle.v2 as paddle import mnist_util @@ -40,17 +39,14 @@ def main(): trainer = paddle.trainer.SGD(update_equation=adam_optimizer) - trainer.train(train_data_reader=train_reader, - topology=cost, - parameters=parameters, - event_handler=event_handler, - batch_size=32, # batch size should be refactor in Data reader - data_types=[ # data_types will be removed, It should be in - # network topology - ('pixel', images.type), - ('label', label.type)], - reader_dict={'pixel':0, 'label':1} - ) + trainer.train( + train_data_reader=train_reader, + cost=cost, + parameters=parameters, + event_handler=event_handler, + batch_size=32, # batch size should be refactor in Data reader + reader_dict={images.name: 0, + label.name: 1}) if __name__ == '__main__': diff --git a/python/paddle/v2/__init__.py b/python/paddle/v2/__init__.py index 1dff754edf1faafcabd3bcea733970235964d344..90321defef7674dde96d57f0e0455031ee64ee86 100644 --- a/python/paddle/v2/__init__.py +++ b/python/paddle/v2/__init__.py @@ -18,6 +18,7 @@ import parameters import trainer import event import data_type +import topology import data_feeder import attr import pooling @@ -25,7 +26,7 @@ import py_paddle.swig_paddle as api __all__ = [ 'optimizer', 'layer', 'activation', 'parameters', 'init', 'trainer', - 'event', 'data_type', 'attr', 'pooling', 'data_feeder' + 'event', 'data_type', 'attr', 'pooling', 'data_feeder', 'topology' ] diff --git a/python/paddle/v2/data_feeder.py b/python/paddle/v2/data_feeder.py index 2a16d46dda47f822dd2d6c168528dd6cec53ab4e..3b106e100cff7539611d95bb4123b4e0dfbfa6cb 100644 --- a/python/paddle/v2/data_feeder.py +++ b/python/paddle/v2/data_feeder.py @@ -23,7 +23,7 @@ class DataFeeder(DataProviderConverter): """ DataFeeder converts the data returned by paddle.reader into a data structure of Arguments which is defined in the API. The paddle.reader usually returns - a list of mini-batch data entries. Each data entry in the list is one sampe. + a list of mini-batch data entries. Each data entry in the list is one sample. Each sample is a list or a tuple with one feature or multiple features. DataFeeder converts this mini-batch data entries into Arguments in order to feed it to C++ interface. diff --git a/python/paddle/v2/data_type.py b/python/paddle/v2/data_type.py index dd3ebfcb4267e1bb59011c81cb5a2716b8e45a6d..522ddfdaacce44be7cf27bdbfc1009d4a0c0bbe6 100644 --- a/python/paddle/v2/data_type.py +++ b/python/paddle/v2/data_type.py @@ -13,10 +13,10 @@ # limitations under the License. from paddle.trainer.PyDataProvider2 import \ - InputType, dense_vector, sparse_binary_vector,\ + InputType, DataType, dense_vector, sparse_binary_vector,\ sparse_vector, integer_value, integer_value_sequence __all__ = [ - 'InputType', 'dense_vector', 'sparse_binary_vector', 'sparse_vector', - 'integer_value', 'integer_value_sequence' + 'InputType', 'DataType', 'dense_vector', 'sparse_binary_vector', + 'sparse_vector', 'integer_value', 'integer_value_sequence' ] diff --git a/python/paddle/v2/layer.py b/python/paddle/v2/layer.py index d15e6398f51f43c1eeab67bba654f91cc56135a4..faf5b8bd87d7b8afe653a821607929589c5abc55 100644 --- a/python/paddle/v2/layer.py +++ b/python/paddle/v2/layer.py @@ -284,6 +284,7 @@ def mixed(size=0, return MixedLayerV2(size, input, name, act, bias_attr, layer_attr) +LayerV2 = Layer data = DataLayerV2 AggregateLevel = conf_helps.layers.AggregateLevel ExpandLevel = conf_helps.layers.ExpandLevel diff --git a/python/paddle/v2/parameters.py b/python/paddle/v2/parameters.py index f2c5a4d49cfc44d3fda3b19a7e2d52fbf9b04129..2a6026bcab1c8a373d8dd5eac480dec62a8eb3b9 100644 --- a/python/paddle/v2/parameters.py +++ b/python/paddle/v2/parameters.py @@ -1,26 +1,21 @@ import numpy as np -from . import layer as v2_layer import py_paddle.swig_paddle as api from paddle.proto.ParameterConfig_pb2 import ParameterConfig +from topology import Topology + __all__ = ['Parameters', 'create'] -def create(*layers): +def create(layers): """ - Create parameter pool by layers. In paddle, layer can be represent a - model config. - + Create parameter pool by topology. :param layers: :return: """ - for layer in layers: - if not isinstance(layer, v2_layer.Layer): - raise ValueError( - 'create must pass a topologies which type is paddle.layer.Layer') - model_config = v2_layer.parse_network(*layers) + topology = Topology(layers) pool = Parameters() - for param in model_config.parameters: + for param in topology.proto().parameters: pool.__append_config__(param) return pool diff --git a/python/paddle/v2/tests/CMakeLists.txt b/python/paddle/v2/tests/CMakeLists.txt index 2f08ceed534c58c3353be7861f45d024b7c60328..46b5d08b8761ea58530f5aefe5d1947408727f85 100644 --- a/python/paddle/v2/tests/CMakeLists.txt +++ b/python/paddle/v2/tests/CMakeLists.txt @@ -2,5 +2,11 @@ add_test(NAME test_v2_layer COMMAND ${PROJ_ROOT}/paddle/.set_python_path.sh -d ${PROJ_ROOT}/python/ ${PYTHON_EXECUTABLE} ${PROJ_ROOT}/python/paddle/v2/tests/test_layer.py WORKING_DIRECTORY ${PROJ_ROOT}/python/paddle) + add_test(NAME test_v2_api - COMMAND bash ${PROJ_ROOT}/python/paddle/v2/tests/run_tests.sh ${PYTHON_EXECUTABLE}) + COMMAND bash ${PROJ_ROOT}/python/paddle/v2/tests/run_tests.sh ${PYTHON_EXECUTABLE}) + +add_test(NAME topology_test + COMMAND ${PROJ_ROOT}/paddle/.set_python_path.sh -d ${PROJ_ROOT}/python/ + ${PYTHON_EXECUTABLE} ${PROJ_ROOT}/python/paddle/v2/tests/test_topology.py + WORKING_DIRECTORY ${PROJ_ROOT}/python/paddle) diff --git a/python/paddle/v2/tests/test_topology.py b/python/paddle/v2/tests/test_topology.py new file mode 100644 index 0000000000000000000000000000000000000000..1bf55a5bc68dfdb837773b3120e5b55d304f644d --- /dev/null +++ b/python/paddle/v2/tests/test_topology.py @@ -0,0 +1,83 @@ +# Copyright PaddlePaddle contributors. All Rights Reserved +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. +import unittest +import paddle.v2.layer as layer +import paddle.v2.topology as topology +import paddle.v2.data_type as data_type +import paddle.trainer_config_helpers as conf_helps + + +class TestTopology(unittest.TestCase): + def test_data_type(self): + pixel = layer.data(name='pixel', type=data_type.dense_vector(784)) + label = layer.data(name='label', type=data_type.integer_value(10)) + hidden = layer.fc(input=pixel, + size=100, + act=conf_helps.SigmoidActivation()) + inference = layer.fc(input=hidden, + size=10, + act=conf_helps.SoftmaxActivation()) + cost = layer.classification_cost(input=inference, label=label) + topo = topology.Topology(cost) + data_types = topo.data_type() + self.assertEqual(len(data_types), 2) + pixel_data_type = filter(lambda type: type[0] == "pixel", data_types) + self.assertEqual(len(pixel_data_type), 1) + pixel_data_type = pixel_data_type[0] + self.assertEqual(pixel_data_type[1].type, data_type.DataType.Dense) + self.assertEqual(pixel_data_type[1].dim, 784) + + label_data_type = filter(lambda type: type[0] == "label", data_types) + self.assertEqual(len(label_data_type), 1) + label_data_type = label_data_type[0] + self.assertEqual(label_data_type[1].type, data_type.DataType.Index) + self.assertEqual(label_data_type[1].dim, 10) + + def test_get_layer(self): + pixel = layer.data(name='pixel', type=data_type.dense_vector(784)) + label = layer.data(name='label', type=data_type.integer_value(10)) + hidden = layer.fc(input=pixel, + size=100, + act=conf_helps.SigmoidActivation()) + inference = layer.fc(input=hidden, + size=10, + act=conf_helps.SoftmaxActivation()) + cost = layer.classification_cost(input=inference, label=label) + topo = topology.Topology(cost) + pixel_layer = topo.get_layer("pixel") + label_layer = topo.get_layer("label") + self.assertEqual(pixel_layer, pixel) + self.assertEqual(label_layer, label) + + def test_parse(self): + pixel = layer.data(name='pixel', type=data_type.dense_vector(784)) + label = layer.data(name='label', type=data_type.integer_value(10)) + hidden = layer.fc(input=pixel, + size=100, + act=conf_helps.SigmoidActivation()) + inference = layer.fc(input=hidden, + size=10, + act=conf_helps.SoftmaxActivation()) + maxid = layer.max_id(input=inference) + cost1 = layer.classification_cost(input=inference, label=label) + cost2 = layer.cross_entropy_cost(input=inference, label=label) + + topology.Topology(cost2).proto() + topology.Topology([cost1]).proto() + topology.Topology([cost1, cost2]).proto() + topology.Topology([inference, maxid]).proto() + + +if __name__ == '__main__': + unittest.main() diff --git a/python/paddle/v2/topology.py b/python/paddle/v2/topology.py new file mode 100644 index 0000000000000000000000000000000000000000..a51b1073b4fc4fd3ac44c355e050b0d720944645 --- /dev/null +++ b/python/paddle/v2/topology.py @@ -0,0 +1,95 @@ +# Copyright (c) 2016 PaddlePaddle Authors. All Rights Reserved +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +import collections + +from paddle.proto.ModelConfig_pb2 import ModelConfig + +import layer as v2_layer + +__all__ = ['Topology'] + + +class Topology(object): + """ + Topology is used to store the information about all layers + and network configs. + """ + + def __init__(self, layers): + if not isinstance(layers, collections.Sequence): + __check_layer_type__(layers) + layers = [layers] + for layer in layers: + __check_layer_type__(layer) + self.layers = layers + self.__model_config__ = v2_layer.parse_network(*layers) + assert isinstance(self.__model_config__, ModelConfig) + + def proto(self): + return self.__model_config__ + + def get_layer(self, name): + """ + get v2.Layer Class instance by layer name + :param name: + :return: + """ + result_layer = [] + + def find_layer_by_name(layer, layer_name): + if len(result_layer) == 1: + return + elif layer.name == layer_name: + result_layer.append(layer) + else: + for parent_layer in layer.__parent_layers__.values(): + find_layer_by_name(parent_layer, layer_name) + + for layer in self.layers: + find_layer_by_name(layer, name) + + assert len(result_layer) == 1 + return result_layer[0] + + def data_layers(self): + """ + get all data layer + :return: + """ + data_layers = set() + + def find_data_layer(layer): + if isinstance(layer, v2_layer.DataLayerV2): + data_layers.add(layer) + for parent_layer in layer.__parent_layers__.values(): + find_data_layer(parent_layer) + + for layer in self.layers: + find_data_layer(layer) + + return data_layers + + def data_type(self): + """ + get data_type from proto, such as: + [('image', dense_vector(768)), ('label', integer_value(10))] + """ + return [(data_layer.name, data_layer.type) + for data_layer in self.data_layers()] + + +def __check_layer_type__(layer): + if not isinstance(layer, v2_layer.LayerV2): + raise ValueError('layer should have type paddle.layer.Layer') diff --git a/python/paddle/v2/trainer.py b/python/paddle/v2/trainer.py index 097814d2f4619797470668cbd0ea95f112a1fde6..2aeddaff89748ef6c769e1793345f5b143b941c6 100644 --- a/python/paddle/v2/trainer.py +++ b/python/paddle/v2/trainer.py @@ -1,11 +1,10 @@ import collections import py_paddle.swig_paddle as api -from paddle.proto.ModelConfig_pb2 import ModelConfig -from data_feeder import DataFeeder +from data_feeder import DataFeeder +from topology import Topology from . import event as v2_event -from . import layer as v2_layer from . import optimizer as v2_optimizer from . import parameters as v2_parameters @@ -30,7 +29,7 @@ class ITrainer(object): def train(self, train_data_reader, - topology, + cost, parameters, test_data_reader=None, event_handler=None): @@ -38,7 +37,7 @@ class ITrainer(object): train method. :param train_data_reader: - :param topology: + :param cost: :param parameters: :param test_data_reader: :param event_handler: @@ -63,19 +62,18 @@ class SGD(ITrainer): def train(self, train_data_reader, - topology, + cost, parameters, num_passes=1, test_data_reader=None, event_handler=None, batch_size=32, - data_types=None, reader_dict=None): """ Training method. Will train num_passes of input data. :param train_data_reader: - :param topology: Network Topology, use one or more Layers to represent it. + :param cost: cost layers, to be optimized. :param parameters: The parameter pools. :param num_passes: The total train passes. :param test_data_reader: @@ -83,18 +81,18 @@ class SGD(ITrainer): occurred. :type event_handler: (BaseEvent) => None :param batch_size: Not important, will be removed after data refactor. - :param data_types: Not important, will be removed after data refactor. :return: """ if event_handler is None: event_handler = default_event_handler - topology = v2_layer.parse_network(topology) + topology = Topology(cost) __check_train_args__(**locals()) gm = api.GradientMachine.createFromConfigProto( - topology, api.CREATE_MODE_NORMAL, self.__optimizer__.enable_types()) + topology.proto(), api.CREATE_MODE_NORMAL, + self.__optimizer__.enable_types()) assert isinstance(gm, api.GradientMachine) parameters.append_gradient_machine(gm) gm.randParameters() @@ -108,7 +106,7 @@ class SGD(ITrainer): assert isinstance(pass_evaluator, api.Evaluator) out_args = api.Arguments.createArguments(0) - feeder = DataFeeder(data_types, reader_dict) + feeder = DataFeeder(topology.data_type(), reader_dict) for pass_id in xrange(num_passes): event_handler(v2_event.BeginPass(pass_id)) @@ -154,7 +152,7 @@ def __data_reader_to_batch__(reader, batch_size, topology): def input_reorder(func): for item in func(): retv = [] - for __layer_name__ in topology.input_layer_names: + for __layer_name__ in topology.proto().input_layer_names: retv.append(item[__layer_name__]) yield retv @@ -191,7 +189,7 @@ def __check_train_args__(train_data_reader, topology, parameters, raise ValueError('test_data_reader should be a function, which can ' 'return a iterator') - if not isinstance(topology, ModelConfig): + if not isinstance(topology, Topology): raise ValueError('topology should be a model config') if not isinstance(parameters, v2_parameters.Parameters):