mindspore/tests/st/runtime/test_runtime_optimize.py

193 lines
6.5 KiB
Python

# Copyright 2022 Huawei Technologies Co., Ltd
#
# 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 time
import numpy as np
import pytest
import mindspore
from mindspore import context, ops, nn, Tensor
class NetNonConcurrent(nn.Cell):
def __init__(self):
super().__init__()
self.relu = ops.ReLU()
self.add = ops.Add()
def construct(self, input_x, input_x1, input_x2, input_x3):
output = self.relu(input_x)
for _ in range(50):
output = self.add(output, 1)
for _ in range(50):
output = self.add(output, 1)
for _ in range(50):
output = self.add(output, 1)
for _ in range(50):
output = self.add(output, 1)
return output
class NetNonConcurrentWithWhile(nn.Cell):
def __init__(self):
super().__init__()
self.relu = ops.ReLU()
self.add = ops.Add()
def construct(self, input_x, input_loop, input_x1, input_loop1):
output = self.relu(input_x)
while input_loop < 4:
input_loop = input_loop + 1
for _ in range(50):
output = self.add(output, 1)
return output
class NetConcurrent(nn.Cell):
def __init__(self):
super().__init__()
self.relu = ops.ReLU()
self.add = ops.Add()
def construct(self, input_x1, input_x2, input_x3, input_x4):
output1 = self.relu(input_x1)
output2 = self.relu(input_x2)
output3 = self.relu(input_x3)
output4 = self.relu(input_x4)
for _ in range(50):
output1 = self.add(output1, 1)
for _ in range(50):
output2 = self.add(output2, 1)
for _ in range(50):
output3 = self.add(output3, 1)
for _ in range(50):
output4 = self.add(output4, 1)
output = output1 + output2 + output3 + output4
return output
class NetConcurrentWithWhile(nn.Cell):
def __init__(self):
super().__init__()
self.relu = ops.ReLU()
self.add = ops.Add()
def construct(self, input_x1, input_loop1, input_x2, input_loop2):
output1 = self.relu(input_x1)
while input_loop1 < 2:
input_loop1 = input_loop1 + 1
for _ in range(50):
output1 = self.add(output1, 1)
output2 = self.relu(input_x2)
while input_loop2 < 2:
input_loop2 = input_loop2 + 1
for _ in range(50):
output2 = self.add(output2, 1)
output = output1 + output2
return output
def run_multi_actor_fusion(net_name, net, input1, input2, input3, input4, expect_output):
context.set_context(mode=context.GRAPH_MODE)
total_time = 0
total_count = 0
for i in range(200):
time1 = time.time()
output = net(input1, input2, input3, input4).asnumpy()
time2 = time.time()
if i > 1:
total_count += 1
total_time += (time2 - time1) * 1000
assert (output == expect_output).all()
print(net_name + " avg_time:", total_time/total_count)
@pytest.mark.level1
@pytest.mark.platform_x86_cpu
@pytest.mark.platform_x86_gpu_training
@pytest.mark.platform_arm_ascend_training
@pytest.mark.platform_x86_ascend_training
@pytest.mark.env_onecard
def test_non_concurrent():
"""
Feature: Multi actor fusion with non concurrent.
Description: Test the net which is non concurrent, that can trigger the function of multi actor fusion.
Expectation: The value and shape of output are the expected values.
"""
input_x = Tensor(np.ones(2), mindspore.float32)
net = NetNonConcurrent()
expect = np.array([201, 201])
run_multi_actor_fusion("non_concurrent", net, input_x, input_x, input_x, input_x, expect)
@pytest.mark.level1
@pytest.mark.platform_x86_gpu_training
@pytest.mark.platform_arm_ascend_training
@pytest.mark.platform_x86_ascend_training
@pytest.mark.env_onecard
def test_non_concurrent_with_while():
"""
Feature: Multi actor fusion with non concurrent and while.
Description: Test the net which is non concurrent with while, that can trigger the function of multi actor fusion.
Expectation: The value and shape of output are the expected values.
"""
input_x = Tensor(np.ones(2), mindspore.float32)
input_loop = Tensor([0], mindspore.float32)
net = NetNonConcurrentWithWhile()
expect = np.array([201, 201])
run_multi_actor_fusion("non_concurrent_with_while", net, input_x, input_loop, input_x, input_loop, expect)
@pytest.mark.level1
@pytest.mark.platform_x86_cpu
@pytest.mark.platform_x86_gpu_training
@pytest.mark.platform_arm_ascend_training
@pytest.mark.platform_x86_ascend_training
@pytest.mark.env_onecard
def test_concurrent():
"""
Feature: Multi actor fusion with concurrent.
Description: Test the net which is concurrent, that can trigger the function of multi actor fusion.
Expectation: The value and shape of output are the expected values.
"""
input1 = Tensor(np.ones(2), mindspore.float32)
input2 = Tensor(np.ones(2), mindspore.float32)
input3 = Tensor(np.ones(2), mindspore.float32)
input4 = Tensor(np.ones(2), mindspore.float32)
net = NetConcurrent()
expect = np.array([204, 204])
run_multi_actor_fusion("concurrent", net, input1, input2, input3, input4, expect)
@pytest.mark.level1
@pytest.mark.platform_x86_gpu_training
@pytest.mark.platform_arm_ascend_training
@pytest.mark.platform_x86_ascend_training
@pytest.mark.env_onecard
def test_concurrent_with_while():
"""
Feature: Multi actor fusion with concurrent and while.
Description: Test the net which is concurrent with while, that can trigger the function of multi actor fusion.
Expectation: The value and shape of output are the expected values.
"""
input1 = Tensor(np.ones(2), mindspore.float32)
input_loop1 = Tensor([0], mindspore.float32)
input2 = Tensor(np.ones(2), mindspore.float32)
input_loop2 = Tensor([0], mindspore.float32)
net = NetConcurrentWithWhile()
expect = np.array([202, 202])
run_multi_actor_fusion("concurrent_with_while", net, input1, input_loop1, input2, input_loop2, expect)