#!/usr/bin/env python
# Copyright (c) 2016 Baidu, Inc. 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.

from paddle.trainer_config_helpers import *

######################## data source ################################
dict_path = 'gserver/tests/Sequence/tour_dict_phrase.dict'
dict_file = dict()
for line_count, line in enumerate(open(dict_path, "r")):
    dict_file[line.strip()] = line_count

define_py_data_sources(train_list='gserver/tests/Sequence/train.list.nest',
                       test_list=None,
                       module='sequenceGen',
                       obj='process2',
                       args={"dict_file":dict_file})

settings(batch_size=2)
######################## network configure ################################
dict_dim = len(open(dict_path,'r').readlines())
word_dim = 128
hidden_dim = 256
label_dim = 3

data = data_layer(name="word", size=dict_dim)

emb_group = embedding_layer(input=data, size=word_dim)

# (lstm_input + lstm) is equal to lstmemory 
def lstm_group(lstm_group_input):
    with mixed_layer(size=hidden_dim*4) as group_input:
      group_input += full_matrix_projection(input=lstm_group_input)

    lstm_output = lstmemory_group(input=group_input,
                                  name="lstm_group",
                                  size=hidden_dim,
                                  act=TanhActivation(),
                                  gate_act=SigmoidActivation(),
                                  state_act=TanhActivation(),
                                  lstm_layer_attr=ExtraLayerAttribute(error_clipping_threshold=50))
    return lstm_output

lstm_nest_group = recurrent_group(input=SubsequenceInput(emb_group),
                                  step=lstm_group,
                                  name="lstm_nest_group")
# hasSubseq ->(seqlastins) seq
lstm_last = last_seq(input=lstm_nest_group, agg_level=AggregateLevel.EACH_SEQUENCE)

# seq ->(expand) hasSubseq
lstm_expand = expand_layer(input=lstm_last, expand_as=emb_group, expand_level=ExpandLevel.FROM_SEQUENCE)

# hasSubseq ->(average) seq
lstm_average = pooling_layer(input=lstm_expand,
                             pooling_type=AvgPooling(),
                             agg_level=AggregateLevel.EACH_SEQUENCE)

with mixed_layer(size=label_dim, 
                 act=SoftmaxActivation(), 
                 bias_attr=True) as output:
    output += full_matrix_projection(input=lstm_average)

outputs(classification_cost(input=output, label=data_layer(name="label", size=1)))