(续）在深度计算框架MindSpore中如何对不持续的计算进行处理——对数据集进行一定epoch数量的训练后，进行其他工作处理，再返回来接着进行一定epoch数量的训练——单步计算
内容接前文：
https://www.cnblogs.com/devilmaycry812839668/p/14988686.html
这里我们考虑的数据集是自建数据集，那么效果又会如何呢？？？
import mindspore
import numpy as np # 引入numpy科学计算库
import matplotlib.pyplot as plt # 引入绘图库 np.random.seed(123) # 随机数生成种子 import mindspore.nn as nn
import mindspore.ops as ops
from mindspore import Tensor
from mindspore import ParameterTuple, Parameter
from mindspore import dtype as mstype
from mindspore import Model
import mindspore.dataset as ds
from mindspore.train.callback import ModelCheckpoint, CheckpointConfig
from mindspore.train.callback import LossMonitor class Net(nn.Cell):
def __init__(self, input_dims, output_dims):
super(Net, self).__init__()
self.matmul = ops.MatMul() self.weight_1 = Parameter(Tensor(np.random.randn(input_dims, 128), dtype=mstype.float32), name='weight_1')
self.bias_1 = Parameter(Tensor(np.zeros(128), dtype=mstype.float32), name='bias_1')
self.weight_2 = Parameter(Tensor(np.random.randn(128, 64), dtype=mstype.float32), name='weight_2')
self.bias_2 = Parameter(Tensor(np.zeros(64), dtype=mstype.float32), name='bias_2')
self.weight_3 = Parameter(Tensor(np.random.randn(64, output_dims), dtype=mstype.float32), name='weight_3')
self.bias_3 = Parameter(Tensor(np.zeros(output_dims), dtype=mstype.float32), name='bias_3') def construct(self, x):
x1 = self.matmul(x, self.weight_1) + self.bias_1
x2 = self.matmul(x1, self.weight_2) + self.bias_2
x3 = self.matmul(x2, self.weight_3) + self.bias_3
return x3 def main():
net = Net(1, 1)
# loss function
loss = nn.MSELoss()
# optimizer
optim = nn.SGD(params=net.trainable_params(), learning_rate=0.000001)
# make net model
model = Model(net, loss, optim, metrics={'loss': nn.Loss()}) # 数据集
x, y = np.array([[0.1]], dtype=np.float32), np.array([[0.1]], dtype=np.float32) def generator_multidimensional():
for i in range(10):
a = x*i
b = y*i
print(a, b)
yield (a, b) dataset = ds.GeneratorDataset(source=generator_multidimensional, column_names=["input", "output"]) model.train(1, dataset, callbacks=LossMonitor(1), dataset_sink_mode=True) print('*' * 100) x, y = np.array([[0.2]], dtype=np.float32), np.array([[0.2]], dtype=np.float32)
model.train(1, dataset, callbacks=LossMonitor(1), dataset_sink_mode=False) # right
# False, False
# False, True
# True, True xxx # not right
# True, False if __name__ == '__main__':
""" 设置运行的背景context """
from mindspore import context # 为mindspore设置运行背景context
#context.set_context(mode=context.PYNATIVE_MODE, device_target='GPU')
context.set_context(mode=context.GRAPH_MODE, device_target='GPU') import time a = time.time()
main()
b = time.time()
# print(b-a)
运行结果：
WARNING: 'ControlDepend' is deprecated from version 1.1 and will be removed in a future version, use 'Depend' instead.
[WARNING] ME(22644:139765219266688,MainProcess):2021-07-09-03:51:42.606.193 [mindspore/ops/operations/array_ops.py:2302] WARN_DEPRECATED: The usage of Pack is deprecated. Please use Stack.
[[0.]] [[0.]]
[[0.1]] [[0.1]]
[[0.2]] [[0.2]]
[[0.3]] [[0.3]]
[[0.4]] [[0.4]]
[[0.5]] [[0.5]]
[[0.6]] [[0.6]]
[[0.7]] [[0.7]]
[[0.8]] [[0.8]]
[[0.90000004]] [[0.90000004]]
[[0.]] [[0.]]
[[0.1]] [[0.1]]
[[0.]] [[0.]]
[[0.1]] [[0.1]]
[[0.2]] [[0.2]]
[[0.3]] [[0.3]]
[[0.4]] [[0.4]]
[[0.5]] [[0.5]]
[[0.6]] [[0.6]]
[[0.7]] [[0.7]]
[[0.8]] [[0.8]]
[[0.90000004]] [[0.90000004]]
epoch: 1 step: 10, loss is 14095.578
****************************************************************************************************
[[0.]] [[0.]]
[[0.2]] [[0.2]]
[[0.4]] [ERROR] ANALYZER(22644,python):2021-07-09-03:51:44.281.599 [mindspore/ccsrc/pipeline/jit/static_analysis/evaluator.cc:74] Eval] Function construct_wrapper, The number of parameters of this function is 20, but the number of provided arguments is 22. NodeInfo:
Traceback (most recent call last):
File "/tmp/pycharm_project_753/second_nnnew_line_regression.py", line 85, in <module>
[[0.4]] main()
File "/tmp/pycharm_project_753/second_nnnew_line_regression.py", line 63, in main [[0.6]] [[0.6]]model.train(1, dataset, callbacks=LossMonitor(1), dataset_sink_mode=False) File "/usr/local/python-3.7.5/lib/python3.7/site-packages/mindspore/train/model.py", line 592, in train
[[0.8]] [[0.8]]
[[1.]] sink_size=sink_size)[[1.]] File "/usr/local/python-3.7.5/lib/python3.7/site-packages/mindspore/train/model.py", line 385, in _train
[[1.2]] [[1.2]]
[[1.4]] [[1.4]]self._train_process(epoch, train_dataset, list_callback, cb_params) File "/usr/local/python-3.7.5/lib/python3.7/site-packages/mindspore/train/model.py", line 513, in _train_process
[[1.6]] [[1.6]]
[[1.8000001]] [[1.8000001]]outputs = self._train_network(*next_element) File "/usr/local/python-3.7.5/lib/python3.7/site-packages/mindspore/nn/cell.py", line 322, in __call__
out = self.compile_and_run(*inputs)
File "/usr/local/python-3.7.5/lib/python3.7/site-packages/mindspore/nn/cell.py", line 578, in compile_and_run
self.compile(*inputs)
File "/usr/local/python-3.7.5/lib/python3.7/site-packages/mindspore/nn/cell.py", line 565, in compile
_executor.compile(self, *inputs, phase=self.phase, auto_parallel_mode=self._auto_parallel_mode)
File "/usr/local/python-3.7.5/lib/python3.7/site-packages/mindspore/common/api.py", line 505, in compile
result = self._executor.compile(obj, args_list, phase, use_vm)
TypeError: mindspore/ccsrc/pipeline/jit/static_analysis/evaluator.cc:74 Eval] Function construct_wrapper, The number of parameters of this function is 20, but the number of provided arguments is 22. NodeInfo: # 进程已结束，退出代码为 1
经过多次实验，发现：
设置为：
model.train(1, dataset, callbacks=LossMonitor(1), dataset_sink_mode=True)
print('*' * 100)
x, y = np.array([[0.2]], dtype=np.float32), np.array([[0.2]], dtype=np.float32)
model.train(1, dataset, callbacks=LossMonitor(1), dataset_sink_mode=False)
则会报错。
而：
model.train(1, dataset, callbacks=LossMonitor(1), dataset_sink_mode=True)
print('*' * 100)
x, y = np.array([[0.2]], dtype=np.float32), np.array([[0.2]], dtype=np.float32)
model.train(1, dataset, callbacks=LossMonitor(1), dataset_sink_mode=True)
model.train(1, dataset, callbacks=LossMonitor(1), dataset_sink_mode=False)
print('*' * 100)
x, y = np.array([[0.2]], dtype=np.float32), np.array([[0.2]], dtype=np.float32)
model.train(1, dataset, callbacks=LossMonitor(1), dataset_sink_mode=False)
model.train(1, dataset, callbacks=LossMonitor(1), dataset_sink_mode=False)
print('*' * 100)
x, y = np.array([[0.2]], dtype=np.float32), np.array([[0.2]], dtype=np.float32)
model.train(1, dataset, callbacks=LossMonitor(1), dataset_sink_mode=True)
则都不会报错。
可以看到如果两次调用同一个mode来训练持续数据，那么dataset_sink_mode 的设置还是很重要的，
这里面推荐的dataset_sink_mode的设置是如果同一个model多次训练，那么所有的训练时dataset_sink_mode都设置为False ,
这样经验上来说能够更大可能性保证正常运行。
====================================================================
那么我们对多次的 model.train 进行训练和一次的model.train 进行训练，那么在运算效率上会有多大区别呢？？？
1.
一次model.tain 进行20000epochs的训练， 代码如下：
import mindspore
import numpy as np # 引入numpy科学计算库
import matplotlib.pyplot as plt # 引入绘图库 np.random.seed(123) # 随机数生成种子 import mindspore.nn as nn
import mindspore.ops as ops
from mindspore import Tensor
from mindspore import ParameterTuple, Parameter
from mindspore import dtype as mstype
from mindspore import Model
import mindspore.dataset as ds
from mindspore.train.callback import ModelCheckpoint, CheckpointConfig
from mindspore.train.callback import LossMonitor class Net(nn.Cell):
def __init__(self, input_dims, output_dims):
super(Net, self).__init__()
self.matmul = ops.MatMul() self.weight_1 = Parameter(Tensor(np.random.randn(input_dims, 128), dtype=mstype.float32), name='weight_1')
self.bias_1 = Parameter(Tensor(np.zeros(128), dtype=mstype.float32), name='bias_1')
self.weight_2 = Parameter(Tensor(np.random.randn(128, 64), dtype=mstype.float32), name='weight_2')
self.bias_2 = Parameter(Tensor(np.zeros(64), dtype=mstype.float32), name='bias_2')
self.weight_3 = Parameter(Tensor(np.random.randn(64, output_dims), dtype=mstype.float32), name='weight_3')
self.bias_3 = Parameter(Tensor(np.zeros(output_dims), dtype=mstype.float32), name='bias_3') def construct(self, x):
x1 = self.matmul(x, self.weight_1) + self.bias_1
x2 = self.matmul(x1, self.weight_2) + self.bias_2
x3 = self.matmul(x2, self.weight_3) + self.bias_3
return x3 def main():
net = Net(1, 1)
# loss function
loss = nn.MSELoss()
# optimizer
optim = nn.SGD(params=net.trainable_params(), learning_rate=0.000001)
# make net model
model = Model(net, loss, optim, metrics={'loss': nn.Loss()}) # 数据集
x, y = np.array([[0.1]], dtype=np.float32), np.array([[0.1]], dtype=np.float32) def generator_multidimensional():
for i in range(100):
a = x*i
b = y*i
#print(a, b)
yield (a, b) dataset = ds.GeneratorDataset(source=generator_multidimensional, column_names=["input", "output"]) model.train(20000, dataset, callbacks=LossMonitor(100), dataset_sink_mode=False) """
print('*' * 100) x, y = np.array([[0.2]], dtype=np.float32), np.array([[0.2]], dtype=np.float32)
model.train(1, dataset, callbacks=LossMonitor(1), dataset_sink_mode=False)
""" # right
# False, False
# False, True
# True, True xxx # not right
# True, False if __name__ == '__main__':
""" 设置运行的背景context """
from mindspore import context # 为mindspore设置运行背景context
#context.set_context(mode=context.PYNATIVE_MODE, device_target='GPU')
context.set_context(mode=context.GRAPH_MODE, device_target='GPU') import time a = time.time()
main()
b = time.time()
print(b-a)
训练时间：
1464.53s
1460.07s
1452.64s
2.
2次model.tain 分别进行10000epochs的训练， 代码如下：
import mindspore
import numpy as np # 引入numpy科学计算库
import matplotlib.pyplot as plt # 引入绘图库 np.random.seed(123) # 随机数生成种子 import mindspore.nn as nn
import mindspore.ops as ops
from mindspore import Tensor
from mindspore import ParameterTuple, Parameter
from mindspore import dtype as mstype
from mindspore import Model
import mindspore.dataset as ds
from mindspore.train.callback import ModelCheckpoint, CheckpointConfig
from mindspore.train.callback import LossMonitor class Net(nn.Cell):
def __init__(self, input_dims, output_dims):
super(Net, self).__init__()
self.matmul = ops.MatMul() self.weight_1 = Parameter(Tensor(np.random.randn(input_dims, 128), dtype=mstype.float32), name='weight_1')
self.bias_1 = Parameter(Tensor(np.zeros(128), dtype=mstype.float32), name='bias_1')
self.weight_2 = Parameter(Tensor(np.random.randn(128, 64), dtype=mstype.float32), name='weight_2')
self.bias_2 = Parameter(Tensor(np.zeros(64), dtype=mstype.float32), name='bias_2')
self.weight_3 = Parameter(Tensor(np.random.randn(64, output_dims), dtype=mstype.float32), name='weight_3')
self.bias_3 = Parameter(Tensor(np.zeros(output_dims), dtype=mstype.float32), name='bias_3') def construct(self, x):
x1 = self.matmul(x, self.weight_1) + self.bias_1
x2 = self.matmul(x1, self.weight_2) + self.bias_2
x3 = self.matmul(x2, self.weight_3) + self.bias_3
return x3 def main():
net = Net(1, 1)
# loss function
loss = nn.MSELoss()
# optimizer
optim = nn.SGD(params=net.trainable_params(), learning_rate=0.000001)
# make net model
model = Model(net, loss, optim, metrics={'loss': nn.Loss()}) # 数据集
x, y = np.array([[0.1]], dtype=np.float32), np.array([[0.1]], dtype=np.float32) def generator_multidimensional():
for i in range(100):
a = x*i
b = y*i
#print(a, b)
yield (a, b) dataset = ds.GeneratorDataset(source=generator_multidimensional, column_names=["input", "output"]) model.train(10000, dataset, callbacks=LossMonitor(100), dataset_sink_mode=False) print('*' * 100) x, y = np.array([[0.1]], dtype=np.float32), np.array([[0.1]], dtype=np.float32)
model.train(10000, dataset, callbacks=LossMonitor(100), dataset_sink_mode=False) # right
# False, False
# False, True
# True, True xxx # not right
# True, False if __name__ == '__main__':
""" 设置运行的背景context """
from mindspore import context # 为mindspore设置运行背景context
#context.set_context(mode=context.PYNATIVE_MODE, device_target='GPU')
context.set_context(mode=context.GRAPH_MODE, device_target='GPU') import time a = time.time()
main()
b = time.time()
print(b-a)
训练时间：
1460.29s
1454.51s
1457.07s
3.
10次model.tain 分别进行2000epochs的训练， 代码如下：
import mindspore
import numpy as np # 引入numpy科学计算库
import matplotlib.pyplot as plt # 引入绘图库 np.random.seed(123) # 随机数生成种子 import mindspore.nn as nn
import mindspore.ops as ops
from mindspore import Tensor
from mindspore import ParameterTuple, Parameter
from mindspore import dtype as mstype
from mindspore import Model
import mindspore.dataset as ds
from mindspore.train.callback import ModelCheckpoint, CheckpointConfig
from mindspore.train.callback import LossMonitor class Net(nn.Cell):
def __init__(self, input_dims, output_dims):
super(Net, self).__init__()
self.matmul = ops.MatMul() self.weight_1 = Parameter(Tensor(np.random.randn(input_dims, 128), dtype=mstype.float32), name='weight_1')
self.bias_1 = Parameter(Tensor(np.zeros(128), dtype=mstype.float32), name='bias_1')
self.weight_2 = Parameter(Tensor(np.random.randn(128, 64), dtype=mstype.float32), name='weight_2')
self.bias_2 = Parameter(Tensor(np.zeros(64), dtype=mstype.float32), name='bias_2')
self.weight_3 = Parameter(Tensor(np.random.randn(64, output_dims), dtype=mstype.float32), name='weight_3')
self.bias_3 = Parameter(Tensor(np.zeros(output_dims), dtype=mstype.float32), name='bias_3') def construct(self, x):
x1 = self.matmul(x, self.weight_1) + self.bias_1
x2 = self.matmul(x1, self.weight_2) + self.bias_2
x3 = self.matmul(x2, self.weight_3) + self.bias_3
return x3 def main():
net = Net(1, 1)
# loss function
loss = nn.MSELoss()
# optimizer
optim = nn.SGD(params=net.trainable_params(), learning_rate=0.000001)
# make net model
model = Model(net, loss, optim, metrics={'loss': nn.Loss()}) # 数据集
x, y = np.array([[0.1]], dtype=np.float32), np.array([[0.1]], dtype=np.float32) def generator_multidimensional():
for i in range(100):
a = x*i
b = y*i
#print(a, b)
yield (a, b) dataset = ds.GeneratorDataset(source=generator_multidimensional, column_names=["input", "output"]) for i in range(10):
print(i, '\t', '*' * 100)
model.train(2000, dataset, callbacks=LossMonitor(100), dataset_sink_mode=False) # right
# False, False
# False, True
# True, True xxx # not right
# True, False if __name__ == '__main__':
""" 设置运行的背景context """
from mindspore import context # 为mindspore设置运行背景context
#context.set_context(mode=context.PYNATIVE_MODE, device_target='GPU')
context.set_context(mode=context.GRAPH_MODE, device_target='GPU') import time a = time.time()
main()
b = time.time()
print(b-a)
训练时间：
1457.52s
4.
100次model.tain 分别进行200epochs的训练， 代码如下：
import mindspore
import numpy as np # 引入numpy科学计算库
import matplotlib.pyplot as plt # 引入绘图库 np.random.seed(123) # 随机数生成种子 import mindspore.nn as nn
import mindspore.ops as ops
from mindspore import Tensor
from mindspore import ParameterTuple, Parameter
from mindspore import dtype as mstype
from mindspore import Model
import mindspore.dataset as ds
from mindspore.train.callback import ModelCheckpoint, CheckpointConfig
from mindspore.train.callback import LossMonitor class Net(nn.Cell):
def __init__(self, input_dims, output_dims):
super(Net, self).__init__()
self.matmul = ops.MatMul() self.weight_1 = Parameter(Tensor(np.random.randn(input_dims, 128), dtype=mstype.float32), name='weight_1')
self.bias_1 = Parameter(Tensor(np.zeros(128), dtype=mstype.float32), name='bias_1')
self.weight_2 = Parameter(Tensor(np.random.randn(128, 64), dtype=mstype.float32), name='weight_2')
self.bias_2 = Parameter(Tensor(np.zeros(64), dtype=mstype.float32), name='bias_2')
self.weight_3 = Parameter(Tensor(np.random.randn(64, output_dims), dtype=mstype.float32), name='weight_3')
self.bias_3 = Parameter(Tensor(np.zeros(output_dims), dtype=mstype.float32), name='bias_3') def construct(self, x):
x1 = self.matmul(x, self.weight_1) + self.bias_1
x2 = self.matmul(x1, self.weight_2) + self.bias_2
x3 = self.matmul(x2, self.weight_3) + self.bias_3
return x3 def main():
net = Net(1, 1)
# loss function
loss = nn.MSELoss()
# optimizer
optim = nn.SGD(params=net.trainable_params(), learning_rate=0.000001)
# make net model
model = Model(net, loss, optim, metrics={'loss': nn.Loss()}) # 数据集
x, y = np.array([[0.1]], dtype=np.float32), np.array([[0.1]], dtype=np.float32) def generator_multidimensional():
for i in range(100):
a = x*i
b = y*i
#print(a, b)
yield (a, b) dataset = ds.GeneratorDataset(source=generator_multidimensional, column_names=["input", "output"]) for i in range(100):
print(i, '\t', '*' * 100)
model.train(200, dataset, callbacks=LossMonitor(100), dataset_sink_mode=False) # right
# False, False
# False, True
# True, True xxx # not right
# True, False if __name__ == '__main__':
""" 设置运行的背景context """
from mindspore import context # 为mindspore设置运行背景context
#context.set_context(mode=context.PYNATIVE_MODE, device_target='GPU')
context.set_context(mode=context.GRAPH_MODE, device_target='GPU') import time a = time.time()
main()
b = time.time()
print(b-a)
训练时间：
1457.56s
5.
1000次model.tain 分别进行20epochs的训练， 代码如下：
6.
10000次model.tain 分别进行2epochs的训练， 代码如下：
import mindspore
import numpy as np # 引入numpy科学计算库
import matplotlib.pyplot as plt # 引入绘图库 np.random.seed(123) # 随机数生成种子 import mindspore.nn as nn
import mindspore.ops as ops
from mindspore import Tensor
from mindspore import ParameterTuple, Parameter
from mindspore import dtype as mstype
from mindspore import Model
import mindspore.dataset as ds
from mindspore.train.callback import ModelCheckpoint, CheckpointConfig
from mindspore.train.callback import LossMonitor class Net(nn.Cell):
def __init__(self, input_dims, output_dims):
super(Net, self).__init__()
self.matmul = ops.MatMul() self.weight_1 = Parameter(Tensor(np.random.randn(input_dims, 128), dtype=mstype.float32), name='weight_1')
self.bias_1 = Parameter(Tensor(np.zeros(128), dtype=mstype.float32), name='bias_1')
self.weight_2 = Parameter(Tensor(np.random.randn(128, 64), dtype=mstype.float32), name='weight_2')
self.bias_2 = Parameter(Tensor(np.zeros(64), dtype=mstype.float32), name='bias_2')
self.weight_3 = Parameter(Tensor(np.random.randn(64, output_dims), dtype=mstype.float32), name='weight_3')
self.bias_3 = Parameter(Tensor(np.zeros(output_dims), dtype=mstype.float32), name='bias_3') def construct(self, x):
x1 = self.matmul(x, self.weight_1) + self.bias_1
x2 = self.matmul(x1, self.weight_2) + self.bias_2
x3 = self.matmul(x2, self.weight_3) + self.bias_3
return x3 def main():
net = Net(1, 1)
# loss function
loss = nn.MSELoss()
# optimizer
optim = nn.SGD(params=net.trainable_params(), learning_rate=0.000001)
# make net model
model = Model(net, loss, optim, metrics={'loss': nn.Loss()}) # 数据集
x, y = np.array([[0.1]], dtype=np.float32), np.array([[0.1]], dtype=np.float32) def generator_multidimensional():
for i in range(100):
a = x*i
b = y*i
#print(a, b)
yield (a, b) dataset = ds.GeneratorDataset(source=generator_multidimensional, column_names=["input", "output"]) for i in range(10000):
print(i, '\t', '*' * 100)
model.train(2, dataset, callbacks=[LossMonitor(100)], dataset_sink_mode=False) # right
# False, False
# False, True
# True, True xxx # not right
# True, False if __name__ == '__main__':
""" 设置运行的背景context """
from mindspore import context # 为mindspore设置运行背景context
#context.set_context(mode=context.PYNATIVE_MODE, device_target='GPU')
context.set_context(mode=context.GRAPH_MODE, device_target='GPU') import time a = time.time()
main()
b = time.time()
print(b-a)
训练时间：
1464.64s
7.
20000次model.tain 分别进行1epochs的训练， 代码如下：
import mindspore
import numpy as np # 引入numpy科学计算库
import matplotlib.pyplot as plt # 引入绘图库 np.random.seed(123) # 随机数生成种子 import mindspore.nn as nn
import mindspore.ops as ops
from mindspore import Tensor
from mindspore import ParameterTuple, Parameter
from mindspore import dtype as mstype
from mindspore import Model
import mindspore.dataset as ds
from mindspore.train.callback import ModelCheckpoint, CheckpointConfig
from mindspore.train.callback import LossMonitor class Net(nn.Cell):
def __init__(self, input_dims, output_dims):
super(Net, self).__init__()
self.matmul = ops.MatMul() self.weight_1 = Parameter(Tensor(np.random.randn(input_dims, 128), dtype=mstype.float32), name='weight_1')
self.bias_1 = Parameter(Tensor(np.zeros(128), dtype=mstype.float32), name='bias_1')
self.weight_2 = Parameter(Tensor(np.random.randn(128, 64), dtype=mstype.float32), name='weight_2')
self.bias_2 = Parameter(Tensor(np.zeros(64), dtype=mstype.float32), name='bias_2')
self.weight_3 = Parameter(Tensor(np.random.randn(64, output_dims), dtype=mstype.float32), name='weight_3')
self.bias_3 = Parameter(Tensor(np.zeros(output_dims), dtype=mstype.float32), name='bias_3') def construct(self, x):
x1 = self.matmul(x, self.weight_1) + self.bias_1
x2 = self.matmul(x1, self.weight_2) + self.bias_2
x3 = self.matmul(x2, self.weight_3) + self.bias_3
return x3 def main():
net = Net(1, 1)
# loss function
loss = nn.MSELoss()
# optimizer
optim = nn.SGD(params=net.trainable_params(), learning_rate=0.000001)
# make net model
model = Model(net, loss, optim, metrics={'loss': nn.Loss()}) # 数据集
x, y = np.array([[0.1]], dtype=np.float32), np.array([[0.1]], dtype=np.float32) def generator_multidimensional():
for i in range(100):
a = x*i
b = y*i
#print(a, b)
yield (a, b) dataset = ds.GeneratorDataset(source=generator_multidimensional, column_names=["input", "output"]) for i in range(20000):
#print(i, '\t', '*' * 100)
model.train(1, dataset, callbacks=[LossMonitor(100)], dataset_sink_mode=False) # right
# False, False
# False, True
# True, True xxx # not right
# True, False if __name__ == '__main__':
""" 设置运行的背景context """
from mindspore import context # 为mindspore设置运行背景context
#context.set_context(mode=context.PYNATIVE_MODE, device_target='GPU')
context.set_context(mode=context.GRAPH_MODE, device_target='GPU') import time a = time.time()
main()
b = time.time()
print(b-a)
训练时间：
1475.06s
1469.48s
1475.70s
1471.46s
=========================================================
从测试的数据上来看使用多次的model.train确实要比一次调用model.train要耗费时间，不过考虑到这样能够更多的支持其他功能，这种性能损耗也是完全可以接受的。
本文实验环境为 MindSpore1.1 docker版本
宿主机：Ubuntu18.04系统
CPU:I7-8700
GPU:1060ti NVIDIA显卡
(续）在深度计算框架MindSpore中如何对不持续的计算进行处理——对数据集进行一定epoch数量的训练后，进行其他工作处理，再返回来接着进行一定epoch数量的训练——单步计算的更多相关文章
- 带你学习MindSpore中算子使用方法
摘要:本文分享下MindSpore中算子的使用和遇到问题时的解决方法. 本文分享自华为云社区<[MindSpore易点通]算子使用问题与解决方法>,作者:chengxiaoli. 简介 算 ...
- TensorFlow - 框架实现中的三种 Graph
文章目录 TensorFlow - 框架实现中的三种 Graph 1. Graph 2. GraphDef 3. MetaGraph 4. Checkpoint 5. 总结 TensorFlow - ...
- SSH框架应用中常用Jar包用途介绍
struts2需要的几个jar包:1)xwork-core-2.1.62)struts2-core-2.1.83)ognl-2.7.34)freemarker-2.3.155)commons-io-1 ...
- python运维开发(十七)----jQuery续(示例)web框架django
内容目录: jQuery示例 前端插件 web框架 Django框架 jQuery示例 dom事件绑定,dom绑定在form表单提交按钮地方都会绑定一个onclick事件,所有查看网站的人都能看到代码 ...
- 如何在Crystal框架项目中内置启动MetaQ服务？
当Crystal框架项目中需要使用消息机制,而项目规模不大.性能要求不高时,可内置启动MetaQ服务器. 分步指南 项目引入crystal-extend-metaq模块,如下: <depende ...
- 如何在Crystal框架项目中内置启动Zookeeper服务？
当Crystal框架项目需要使用到Zookeeper服务时(如使用Dubbo RPC时,需要注册服务到Zookeeper),而独立部署和启动Zookeeper服务不仅繁琐,也容易出现错误. 在小型项目 ...
- 浅入深出之Java集合框架（中）
Java中的集合框架(中) 由于Java中的集合框架的内容比较多,在这里分为三个部分介绍Java的集合框架,内容是从浅到深,如果已经有java基础的小伙伴可以直接跳到<浅入深出之Java集合框架 ...
- Javscript调用iframe框架页面中函数的方法
Javscript调用iframe框架页面中函数的方法,可以实现iframe之间传值或修改值了, 访问iframe里面的函数: window.frames['CallCenter_iframe'].h ...
- 游戏框架设计中的。绑定binding。。。命令 command 和消息message 以及MVVM
游戏框架设计中的.绑定binding...命令 command 和消息message
- 关于MFC框架程序中CWinApp::OnIdle
很早之前就发现,我写的图形引擎在MFC框架程序中的刷帧率始终在60FPS左右.好在自己的程序对刷帧率的要求不是很高,所以一直没有太过纠结此事.直到今天看了别人的程序才发现应该在函数CWinApp::O ...
随机推荐
- java.lang.NoClassDefFoundError: com/google/gson/GsonBuilder
Exception in thread "main" java.lang.NoClassDefFoundError: com/google/gson/GsonBuilder 解决方 ...
- redis高可用哨兵篇
https://redis.io/docs/manual/sentinel/#sentinels-and-replicas-auto-discovery 官网资料 在上文主从复制的基础上,如果注节点出 ...
- @RequestMapping 注解用在类上面有什么作用？
是一个用来处理请求地址映射的注解,可用于类或方法上.用于类上,表示类中的所有响应请求的方法都是以该地址作为父路径.
- Lru-k在Rust中的实现及源码解析
LRU-K 是一种缓存淘汰算法,旨在改进传统的LRU(Least Recently Used,最近最少使用)算法的性能.将其中高频的数据达到K次访问移入到另一个队列进行保护. 算法思想 LRU-K中的 ...
- 记录.Net部署Docker-v指令使用
记录Docker的-v指令使用 前言 之前我浅学了一下docker,方便部署.net项目(部署的是打包之后的项目) dockerfile文件如下: FROM mcr.microsoft.com/dot ...
- 千万别忽视基础！十张图带你一步步理解Java内存结构！
作为一个Java程序员,在日常的开发中,不必像C/C++程序员那样,为每一个内存的分配而操心,JVM会替我们进行自动的内存分配和回收,方便我们开发.但是一旦发生内存泄漏或者内存溢出,如果对Java内存 ...
- 【论文阅读】Exploring the Limitations of Behavior Cloning for Autonomous Driving
Column: January 16, 2022 11:11 PM Last edited time: January 21, 2022 12:23 PM Sensor/组织: 1 RGB Statu ...
- 实测14us，Linux-RT实时性能及开发案例分享—基于全志T507-H国产平台
本文带来的是基于全志T507-H(硬件平台:创龙科技TLT507-EVM评估板),Linux-RT内核的硬件GPIO输入和输出实时性测试及应用开发案例的分享.本次演示的开发环境如下: Windows开 ...
- FLUTTER 中 Isolate 的一个例子.
isolate 起因 最近看了一点isolate的东西, 自己写了一个例子. 普通的的 consummer-producer例子是只有前后两端的,实际上,会把前后两端再进行包装. 我这里这个例子,是把 ...
- 树莓派安装OpenCv
树莓派安装OpenCv 更换树莓派软件源 我们选择将树莓派的软件源切换到清华大学镜像站,据笔者亲测,通过此站可以顺利安装openCV. 切换软件源需要修改两个软件源配置文件的内容. 第一个需要修改是「 ...