test_profiler.py 4.0 KB
Newer Older
1
#   Copyright (c) 2018 PaddlePaddle Authors. All Rights Reserved.
D
dzhwinter 已提交
2
#
D
dzhwinter 已提交
3 4 5
# 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
D
dzhwinter 已提交
6
#
D
dzhwinter 已提交
7
#     http://www.apache.org/licenses/LICENSE-2.0
D
dzhwinter 已提交
8
#
D
dzhwinter 已提交
9 10 11 12 13 14
# 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.

15
import unittest
16
import os
17
import numpy as np
18 19 20 21
import paddle.fluid as fluid
import paddle.fluid.profiler as profiler
import paddle.fluid.layers as layers
import paddle.fluid.core as core
D
dangqingqing 已提交
22 23


24
class TestProfiler(unittest.TestCase):
X
Xin Pan 已提交
25
    def net_profiler(self, state, profile_path='/tmp/profile'):
26 27
        enable_if_gpu = state == 'GPU' or state == "All"
        if enable_if_gpu and not core.is_compiled_with_cuda():
28 29 30 31 32 33
            return
        startup_program = fluid.Program()
        main_program = fluid.Program()

        with fluid.program_guard(main_program, startup_program):
            image = fluid.layers.data(name='x', shape=[784], dtype='float32')
X
Xin Pan 已提交
34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49
            hidden1 = fluid.layers.fc(input=image, size=64, act='relu')
            i = layers.zeros(shape=[1], dtype='int64')
            counter = fluid.layers.zeros(
                shape=[1], dtype='int64', force_cpu=True)
            until = layers.fill_constant([1], dtype='int64', value=10)
            data_arr = layers.array_write(hidden1, i)
            cond = fluid.layers.less_than(x=counter, y=until)
            while_op = fluid.layers.While(cond=cond)
            with while_op.block():
                hidden_n = fluid.layers.fc(input=hidden1, size=64, act='relu')
                layers.array_write(hidden_n, i, data_arr)
                fluid.layers.increment(x=counter, value=1, in_place=True)
                layers.less_than(x=counter, y=until, cond=cond)

            hidden_n = layers.array_read(data_arr, i)
            hidden2 = fluid.layers.fc(input=hidden_n, size=64, act='relu')
50 51 52
            predict = fluid.layers.fc(input=hidden2, size=10, act='softmax')
            label = fluid.layers.data(name='y', shape=[1], dtype='int64')
            cost = fluid.layers.cross_entropy(input=predict, label=label)
Y
Yu Yang 已提交
53
            avg_cost = fluid.layers.mean(cost)
F
fengjiayi 已提交
54 55 56
            batch_size = fluid.layers.create_tensor(dtype='int64')
            batch_acc = fluid.layers.accuracy(
                input=predict, label=label, total=batch_size)
57

58
        optimizer = fluid.optimizer.Momentum(learning_rate=0.001, momentum=0.9)
59 60 61 62 63
        opts = optimizer.minimize(avg_cost, startup_program=startup_program)

        place = fluid.CPUPlace() if state == 'CPU' else fluid.CUDAPlace(0)
        exe = fluid.Executor(place)
        exe.run(startup_program)
64

F
fengjiayi 已提交
65
        pass_acc_calculator = fluid.average.WeightedAverage()
F
fengjiayi 已提交
66
        with profiler.profiler(state, 'total', profile_path) as prof:
67 68 69 70 71
            for iter in range(10):
                if iter == 2:
                    profiler.reset_profiler()
                x = np.random.random((32, 784)).astype("float32")
                y = np.random.randint(0, 10, (32, 1)).astype("int64")
72

73 74 75
                outs = exe.run(main_program,
                               feed={'x': x,
                                     'y': y},
F
fengjiayi 已提交
76
                               fetch_list=[avg_cost, batch_acc, batch_size])
77
                acc = np.array(outs[1])
F
fengjiayi 已提交
78 79 80
                b_size = np.array(outs[2])
                pass_acc_calculator.add(value=acc, weight=b_size)
                pass_acc = pass_acc_calculator.eval()
81

D
dangqingqing 已提交
82 83
    def test_cpu_profiler(self):
        self.net_profiler('CPU')
84

D
dangqingqing 已提交
85 86
    def test_cuda_profiler(self):
        self.net_profiler('GPU')
87

88
    def test_all_profiler(self):
X
Xin Pan 已提交
89 90 91
        self.net_profiler('All', '/tmp/profile_out')
        with open('/tmp/profile_out', 'r') as f:
            self.assertGreater(len(f.read()), 0)
92

D
dangqingqing 已提交
93

94 95
if __name__ == '__main__':
    unittest.main()