#edit-mode: -*- 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.

#Todo(luotao02) This config is only used for unitest. It is out of date now, and will be updated later.

default_initial_std(0.5)

model_type("nn")

DataLayer(
    name = "input",
    size = 3,
)

DataLayer(
    name = "weight",
    size = 1,
)

Layer(
    name = "layer1_1",
    type = "fc",
    size = 5,
    active_type = "sigmoid",
    inputs = "input",
)

Layer(
    name = "layer1_2",
    type = "fc",
    size = 12,
    active_type = "linear",
    inputs = Input("input", parameter_name='sharew'),
)

Layer(
    name = "layer1_3",
    type = "fc",
    size = 3,
    active_type = "tanh",
    inputs = "input",
)

Layer(
    name = "layer1_5",
    type = "fc",
    size = 3,
    active_type = "tanh",
    inputs = Input("input",
              learning_rate=0.01,
              momentum=0.9,
              decay_rate=0.05,
              initial_mean=0.0,
              initial_std=0.01,
              format = "csc",
              nnz = 4)
)

FCLayer(
    name = "layer1_4",
    size = 5,
    active_type = "square",
    inputs = "input",
    drop_rate = 0.5,
)

Layer(
    name = "pool",
    type = "pool",
    inputs = Input("layer1_2",
                   pool = Pool(pool_type="cudnn-avg-pool",
                               channels = 1,
                               size_x = 2,
                               size_y = 3,
                               img_width = 3,
                               padding = 1,
                               padding_y = 2,
                               stride = 2,
                               stride_y = 3))
)

Layer(
    name = "concat",
    type = "concat",
    inputs = ["layer1_3", "layer1_4"],
)

MixedLayer(
    name = "output",
    size = 3,
    active_type = "softmax",
    inputs = [
        FullMatrixProjection("layer1_1",
              learning_rate=0.1),
        TransposedFullMatrixProjection("layer1_2", parameter_name='sharew'),
        FullMatrixProjection("concat"),
        IdentityProjection("layer1_3"),
    ],
)

Layer(
    name = "label",
    type = "data",
    size = 1,
)

Layer(
    name = "cost",
    type = "multi-class-cross-entropy",
    inputs = ["output", "label", "weight"],
)

Layer(
    name = "cost2",
    type = "nce",
    num_classes = 3,
    active_type = "sigmoid",
    neg_sampling_dist = [0.1, 0.3, 0.6],
    inputs = ["layer1_2", "label", "weight"],
)

Evaluator(
    name = "error",
    type = "classification_error",
    inputs = ["output", "label", "weight"]
)

Inputs("input", "label", "weight")
Outputs("cost", "cost2")

TrainData(
    ProtoData(
        files = "dummy_list",
        constant_slots = [1.0],
        async_load_data = True,
    )
)

TestData(
    SimpleData(
        files = "trainer/tests/sample_filelist.txt",
        feat_dim = 3,
        context_len = 0,
        buffer_capacity = 1000000,
        async_load_data = False,
    ),
)

Settings(
    algorithm = "sgd",
    num_batches_per_send_parameter = 1,
    num_batches_per_get_parameter = 1,
    batch_size = 100,
    learning_rate = 0.001,
    learning_rate_decay_a = 1e-5,
    learning_rate_decay_b = 0.5,
)