model.train方法的dataset_sink_mode参数设置为False时以step作为单位打印数据——(只在mode=context.GRAPH_MODE下成立,在mode=context.PYNATIVE_MODE模式下不成立)
如题:
官方中的内容支持:
使用summary功能时,建议将model.train
方法的dataset_sink_mode
参数设置为False
,从而以step
作为collect_freq
参数的单位收集数据。当dataset_sink_mode
为True
时,将以epoch
作为collect_freq
的单位,此时建议手动设置collect_freq
参数。collect_freq
参数默认值为10
。
从官方文档中我们可以知道:
如果model.train
方法的dataset_sink_mode
参数设置为False
,那么就是以step为单位打印数据。
如果model.train
方法的dataset_sink_mode
参数设置为True
,那么就是以episode为单位打印数据。
这里我们不过多解释,直接上代码:
(代码具体参看:https://www.cnblogs.com/devilmaycry812839668/p/14971668.html)
当 dataset_sink_mode=False 时:
model.train(epoch_size, ds_train, callbacks=[ckpoint_cb, LossMonitor(125)], dataset_sink_mode=False)
![](https://images.cnblogs.com/OutliningIndicators/ContractedBlock.gif)
#!/usr/bin python # encoding:UTF-8 """" 对输入的超参数进行处理 """ import os import argparse """ 设置运行的背景context """ from mindspore import context """ 对数据集进行预处理 """ import mindspore.dataset as ds import mindspore.dataset.transforms.c_transforms as C import mindspore.dataset.vision.c_transforms as CV from mindspore.dataset.vision import Inter from mindspore import dtype as mstype """ 构建神经网络 """ import mindspore.nn as nn from mindspore.common.initializer import Normal """ 训练时对模型参数的保存 """ from mindspore.train.callback import ModelCheckpoint, CheckpointConfig """ 导入模型训练需要的库 """ from mindspore.nn import Accuracy from mindspore.train.callback import LossMonitor from mindspore import Model parser = argparse.ArgumentParser(description='MindSpore LeNet Example') parser.add_argument('--device_target', type=str, default="GPU", choices=['Ascend', 'GPU', 'CPU']) args = parser.parse_known_args()[0] # 为mindspore设置运行背景context context.set_context(mode=context.GRAPH_MODE, device_target=args.device_target) def create_dataset(data_path, batch_size=32, repeat_size=1, num_parallel_workers=1): # 定义数据集 mnist_ds = ds.MnistDataset(data_path) resize_height, resize_width = 32, 32 rescale = 1.0 / 255.0 shift = 0.0 rescale_nml = 1 / 0.3081 shift_nml = -1 * 0.1307 / 0.3081 # 定义所需要操作的map映射 resize_op = CV.Resize((resize_height, resize_width), interpolation=Inter.LINEAR) rescale_nml_op = CV.Rescale(rescale_nml, shift_nml) rescale_op = CV.Rescale(rescale, shift) hwc2chw_op = CV.HWC2CHW() type_cast_op = C.TypeCast(mstype.int32) # 使用map映射函数,将数据操作应用到数据集 mnist_ds = mnist_ds.map(operations=type_cast_op, input_columns="label", num_parallel_workers=num_parallel_workers) mnist_ds = mnist_ds.map(operations=resize_op, input_columns="image", num_parallel_workers=num_parallel_workers) mnist_ds = mnist_ds.map(operations=rescale_op, input_columns="image", num_parallel_workers=num_parallel_workers) mnist_ds = mnist_ds.map(operations=rescale_nml_op, input_columns="image", num_parallel_workers=num_parallel_workers) mnist_ds = mnist_ds.map(operations=hwc2chw_op, input_columns="image", num_parallel_workers=num_parallel_workers) # 进行shuffle、batch、repeat操作 buffer_size = 10000 mnist_ds = mnist_ds.shuffle(buffer_size=buffer_size) mnist_ds = mnist_ds.batch(batch_size, drop_remainder=True) mnist_ds = mnist_ds.repeat(repeat_size) return mnist_ds class LeNet5(nn.Cell): """ Lenet网络结构 """ def __init__(self, num_class=10, num_channel=1): super(LeNet5, self).__init__() # 定义所需要的运算 self.conv1 = nn.Conv2d(num_channel, 6, 5, pad_mode='valid') self.conv2 = nn.Conv2d(6, 16, 5, pad_mode='valid') self.fc1 = nn.Dense(16 * 5 * 5, 120, weight_init=Normal(0.02)) self.fc2 = nn.Dense(120, 84, weight_init=Normal(0.02)) self.fc3 = nn.Dense(84, num_class, weight_init=Normal(0.02)) self.relu = nn.ReLU() self.max_pool2d = nn.MaxPool2d(kernel_size=2, stride=2) self.flatten = nn.Flatten() def construct(self, x): # 使用定义好的运算构建前向网络 x = self.conv1(x) x = self.relu(x) x = self.max_pool2d(x) x = self.conv2(x) x = self.relu(x) x = self.max_pool2d(x) x = self.flatten(x) x = self.fc1(x) x = self.relu(x) x = self.fc2(x) x = self.relu(x) x = self.fc3(x) return x # 实例化网络 net = LeNet5() # 定义损失函数 net_loss = nn.SoftmaxCrossEntropyWithLogits(sparse=True, reduction='mean') # 定义优化器 net_opt = nn.Momentum(net.trainable_params(), learning_rate=0.01, momentum=0.9) # 设置模型保存参数 # 每125steps保存一次模型参数,最多保留15个文件 config_ck = CheckpointConfig(save_checkpoint_steps=125, keep_checkpoint_max=15) # 应用模型保存参数 ckpoint = ModelCheckpoint(prefix="checkpoint_lenet", config=config_ck) def train_net(args, model, epoch_size, data_path, repeat_size, ckpoint_cb, sink_mode): """定义训练的方法""" # 加载训练数据集 ds_train = create_dataset(os.path.join(data_path, "train"), 32, repeat_size) model.train(epoch_size, ds_train, callbacks=[ckpoint_cb, LossMonitor(125)], dataset_sink_mode=sink_mode) def test_net(network, model, data_path): """定义验证的方法""" ds_eval = create_dataset(os.path.join(data_path, "test")) acc = model.eval(ds_eval, dataset_sink_mode=False) print("{}".format(acc)) mnist_path = "./datasets/MNIST_Data" train_epoch = 1 dataset_size = 1 model = Model(net, net_loss, net_opt, metrics={"Accuracy": Accuracy()}) train_net(args, model, train_epoch, mnist_path, dataset_size, ckpoint, False) test_net(net, model, mnist_path)
输出结果:
epoch: 1 step: 125, loss is 2.2959
epoch: 1 step: 250, loss is 2.2959309
epoch: 1 step: 375, loss is 2.2982068
epoch: 1 step: 500, loss is 2.2916625
epoch: 1 step: 625, loss is 2.3001077
epoch: 1 step: 750, loss is 1.9395046
epoch: 1 step: 875, loss is 0.728865
epoch: 1 step: 1000, loss is 0.2426785
epoch: 1 step: 1125, loss is 0.45475814
epoch: 1 step: 1250, loss is 0.1676599
epoch: 1 step: 1375, loss is 0.14273866
epoch: 1 step: 1500, loss is 0.030339874
epoch: 1 step: 1625, loss is 0.19792284
epoch: 1 step: 1750, loss is 0.09066871
epoch: 1 step: 1875, loss is 0.12958783
{'Accuracy': 0.9688501602564102}
当 dataset_sink_mode=True 时:
model.train(epoch_size, ds_train, callbacks=[ckpoint_cb, LossMonitor(125)], dataset_sink_mode=True)
![](https://images.cnblogs.com/OutliningIndicators/ContractedBlock.gif)
#!/usr/bin python # encoding:UTF-8 """" 对输入的超参数进行处理 """ import os import argparse """ 设置运行的背景context """ from mindspore import context """ 对数据集进行预处理 """ import mindspore.dataset as ds import mindspore.dataset.transforms.c_transforms as C import mindspore.dataset.vision.c_transforms as CV from mindspore.dataset.vision import Inter from mindspore import dtype as mstype """ 构建神经网络 """ import mindspore.nn as nn from mindspore.common.initializer import Normal """ 训练时对模型参数的保存 """ from mindspore.train.callback import ModelCheckpoint, CheckpointConfig """ 导入模型训练需要的库 """ from mindspore.nn import Accuracy from mindspore.train.callback import LossMonitor from mindspore import Model parser = argparse.ArgumentParser(description='MindSpore LeNet Example') parser.add_argument('--device_target', type=str, default="GPU", choices=['Ascend', 'GPU', 'CPU']) args = parser.parse_known_args()[0] # 为mindspore设置运行背景context context.set_context(mode=context.GRAPH_MODE, device_target=args.device_target) def create_dataset(data_path, batch_size=32, repeat_size=1, num_parallel_workers=1): # 定义数据集 mnist_ds = ds.MnistDataset(data_path) resize_height, resize_width = 32, 32 rescale = 1.0 / 255.0 shift = 0.0 rescale_nml = 1 / 0.3081 shift_nml = -1 * 0.1307 / 0.3081 # 定义所需要操作的map映射 resize_op = CV.Resize((resize_height, resize_width), interpolation=Inter.LINEAR) rescale_nml_op = CV.Rescale(rescale_nml, shift_nml) rescale_op = CV.Rescale(rescale, shift) hwc2chw_op = CV.HWC2CHW() type_cast_op = C.TypeCast(mstype.int32) # 使用map映射函数,将数据操作应用到数据集 mnist_ds = mnist_ds.map(operations=type_cast_op, input_columns="label", num_parallel_workers=num_parallel_workers) mnist_ds = mnist_ds.map(operations=resize_op, input_columns="image", num_parallel_workers=num_parallel_workers) mnist_ds = mnist_ds.map(operations=rescale_op, input_columns="image", num_parallel_workers=num_parallel_workers) mnist_ds = mnist_ds.map(operations=rescale_nml_op, input_columns="image", num_parallel_workers=num_parallel_workers) mnist_ds = mnist_ds.map(operations=hwc2chw_op, input_columns="image", num_parallel_workers=num_parallel_workers) # 进行shuffle、batch、repeat操作 buffer_size = 10000 mnist_ds = mnist_ds.shuffle(buffer_size=buffer_size) mnist_ds = mnist_ds.batch(batch_size, drop_remainder=True) mnist_ds = mnist_ds.repeat(repeat_size) return mnist_ds class LeNet5(nn.Cell): """ Lenet网络结构 """ def __init__(self, num_class=10, num_channel=1): super(LeNet5, self).__init__() # 定义所需要的运算 self.conv1 = nn.Conv2d(num_channel, 6, 5, pad_mode='valid') self.conv2 = nn.Conv2d(6, 16, 5, pad_mode='valid') self.fc1 = nn.Dense(16 * 5 * 5, 120, weight_init=Normal(0.02)) self.fc2 = nn.Dense(120, 84, weight_init=Normal(0.02)) self.fc3 = nn.Dense(84, num_class, weight_init=Normal(0.02)) self.relu = nn.ReLU() self.max_pool2d = nn.MaxPool2d(kernel_size=2, stride=2) self.flatten = nn.Flatten() def construct(self, x): # 使用定义好的运算构建前向网络 x = self.conv1(x) x = self.relu(x) x = self.max_pool2d(x) x = self.conv2(x) x = self.relu(x) x = self.max_pool2d(x) x = self.flatten(x) x = self.fc1(x) x = self.relu(x) x = self.fc2(x) x = self.relu(x) x = self.fc3(x) return x # 实例化网络 net = LeNet5() # 定义损失函数 net_loss = nn.SoftmaxCrossEntropyWithLogits(sparse=True, reduction='mean') # 定义优化器 net_opt = nn.Momentum(net.trainable_params(), learning_rate=0.01, momentum=0.9) # 设置模型保存参数 # 每125steps保存一次模型参数,最多保留15个文件 config_ck = CheckpointConfig(save_checkpoint_steps=125, keep_checkpoint_max=15) # 应用模型保存参数 ckpoint = ModelCheckpoint(prefix="checkpoint_lenet", config=config_ck) def train_net(args, model, epoch_size, data_path, repeat_size, ckpoint_cb, sink_mode): """定义训练的方法""" # 加载训练数据集 ds_train = create_dataset(os.path.join(data_path, "train"), 32, repeat_size) model.train(epoch_size, ds_train, callbacks=[ckpoint_cb, LossMonitor(125)], dataset_sink_mode=sink_mode) def test_net(network, model, data_path): """定义验证的方法""" ds_eval = create_dataset(os.path.join(data_path, "test")) acc = model.eval(ds_eval, dataset_sink_mode=False) print("{}".format(acc)) mnist_path = "./datasets/MNIST_Data" train_epoch = 1 dataset_size = 1 model = Model(net, net_loss, net_opt, metrics={"Accuracy": Accuracy()}) train_net(args, model, train_epoch, mnist_path, dataset_size, ckpoint, True) test_net(net, model, mnist_path)
输出结果:
epoch: 1 step: 1875, loss is 0.04107348
{'Accuracy': 0.9638421474358975}
==================================================================
可以看到在mindspore中进行训练时如果设置 dataset_sink_mode=True
那么无论设置多少step打印一次结果,每个epoch中只会打印一次结果,即一个epoch中最后的那个打印结果
(比如数据集中一个epoch是100个数据,batch_size=10, 一个epoch的数据训练需要10个steps, 如果设置dataset_sink_mode=True那么只会打印第10step的结果,前9次step的结果不打印)。
=====================================================================
经过进一步发现,上面的描述都是在 运行背景设置为:
context.set_context(mode=context.GRAPH_MODE, device_target='GPU')
下才成立的。
如果设置为:
context.set_context(mode=context.PYNATIVE_MODE, device_target='GPU')
那么,无论 dataset_sink_mode 设置为False 还是True , 都是执行 以step为单位打印数据。
代码如下:
#!/usr/bin python # encoding:UTF-8 """" 对输入的超参数进行处理 """ import os import argparse """ 设置运行的背景context """ from mindspore import context """ 对数据集进行预处理 """ import mindspore.dataset as ds import mindspore.dataset.transforms.c_transforms as C import mindspore.dataset.vision.c_transforms as CV from mindspore.dataset.vision import Inter from mindspore import dtype as mstype """ 构建神经网络 """ import mindspore.nn as nn from mindspore.common.initializer import Normal """ 训练时对模型参数的保存 """ from mindspore.train.callback import ModelCheckpoint, CheckpointConfig """ 导入模型训练需要的库 """ from mindspore.nn import Accuracy from mindspore.train.callback import LossMonitor from mindspore import Model parser = argparse.ArgumentParser(description='MindSpore LeNet Example') parser.add_argument('--device_target', type=str, default="GPU", choices=['Ascend', 'GPU', 'CPU']) args = parser.parse_known_args()[0] # 为mindspore设置运行背景context context.set_context(mode=context.PYNATIVE_MODE, device_target=args.device_target) def create_dataset(data_path, batch_size=32, repeat_size=1, num_parallel_workers=1): # 定义数据集 mnist_ds = ds.MnistDataset(data_path) resize_height, resize_width = 32, 32 rescale = 1.0 / 255.0 shift = 0.0 rescale_nml = 1 / 0.3081 shift_nml = -1 * 0.1307 / 0.3081 # 定义所需要操作的map映射 resize_op = CV.Resize((resize_height, resize_width), interpolation=Inter.LINEAR) rescale_nml_op = CV.Rescale(rescale_nml, shift_nml) rescale_op = CV.Rescale(rescale, shift) hwc2chw_op = CV.HWC2CHW() type_cast_op = C.TypeCast(mstype.int32) # 使用map映射函数,将数据操作应用到数据集 mnist_ds = mnist_ds.map(operations=type_cast_op, input_columns="label", num_parallel_workers=num_parallel_workers) mnist_ds = mnist_ds.map(operations=resize_op, input_columns="image", num_parallel_workers=num_parallel_workers) mnist_ds = mnist_ds.map(operations=rescale_op, input_columns="image", num_parallel_workers=num_parallel_workers) mnist_ds = mnist_ds.map(operations=rescale_nml_op, input_columns="image", num_parallel_workers=num_parallel_workers) mnist_ds = mnist_ds.map(operations=hwc2chw_op, input_columns="image", num_parallel_workers=num_parallel_workers) # 进行shuffle、batch、repeat操作 buffer_size = 10000 mnist_ds = mnist_ds.shuffle(buffer_size=buffer_size) mnist_ds = mnist_ds.batch(batch_size, drop_remainder=True) mnist_ds = mnist_ds.repeat(repeat_size) return mnist_ds class LeNet5(nn.Cell): """ Lenet网络结构 """ def __init__(self, num_class=10, num_channel=1): super(LeNet5, self).__init__() # 定义所需要的运算 self.conv1 = nn.Conv2d(num_channel, 6, 5, pad_mode='valid') self.conv2 = nn.Conv2d(6, 16, 5, pad_mode='valid') self.fc1 = nn.Dense(16 * 5 * 5, 120, weight_init=Normal(0.02)) self.fc2 = nn.Dense(120, 84, weight_init=Normal(0.02)) self.fc3 = nn.Dense(84, num_class, weight_init=Normal(0.02)) self.relu = nn.ReLU() self.max_pool2d = nn.MaxPool2d(kernel_size=2, stride=2) self.flatten = nn.Flatten() def construct(self, x): # 使用定义好的运算构建前向网络 x = self.conv1(x) x = self.relu(x) x = self.max_pool2d(x) x = self.conv2(x) x = self.relu(x) x = self.max_pool2d(x) x = self.flatten(x) x = self.fc1(x) x = self.relu(x) x = self.fc2(x) x = self.relu(x) x = self.fc3(x) return x # 实例化网络 net = LeNet5() # 定义损失函数 net_loss = nn.SoftmaxCrossEntropyWithLogits(sparse=True, reduction='mean') # 定义优化器 net_opt = nn.Momentum(net.trainable_params(), learning_rate=0.01, momentum=0.9) # 设置模型保存参数 # 每125steps保存一次模型参数,最多保留15个文件 config_ck = CheckpointConfig(save_checkpoint_steps=125, keep_checkpoint_max=15) # 应用模型保存参数 ckpoint = ModelCheckpoint(prefix="checkpoint_lenet", config=config_ck) def train_net(args, model, epoch_size, data_path, repeat_size, ckpoint_cb, sink_mode): """定义训练的方法""" # 加载训练数据集 ds_train = create_dataset(os.path.join(data_path, "train"), 32, repeat_size) model.train(epoch_size, ds_train, callbacks=[ckpoint_cb, LossMonitor(125)], dataset_sink_mode=sink_mode) def test_net(network, model, data_path): """定义验证的方法""" ds_eval = create_dataset(os.path.join(data_path, "test")) acc = model.eval(ds_eval, dataset_sink_mode=False) print("{}".format(acc)) mnist_path = "./datasets/MNIST_Data" train_epoch = 1 dataset_size = 1 model = Model(net, net_loss, net_opt, metrics={"Accuracy": Accuracy()}) train_net(args, model, train_epoch, mnist_path, dataset_size, ckpoint, True) test_net(net, model, mnist_path)
不得不说,对于新框架 MindSpore 来说,还是坑蛮多的,稍有不注意就会出意料以外的结果。
posted on 2021-07-08 11:45 Angry_Panda 阅读(692) 评论(0) 编辑 收藏 举报