昇思MindSpore学习笔记2--快速入门

一、环境准备

安装minspore模块

!pip uninstall mindspore -y
!pip install -i https://pypi.mirrors.ustc.edu.cn/simple mindspore==2.3.0rc1

导入minsporei

mport mindspore
from mindspore import nn
from mindspore.dataset import vision, transforms
from mindspore.dataset import MnistDataset

二、学习内容

用MindSpore API实现一个简单的深度学习模型。

1.使用download下载数据

from download import download
url = (
    "https://mindspore-website.obs.cn-north-4.myhuaweicloud.com/"
    "notebook/datasets/MNIST_Data.zip"
)
path = download(url, "./", kind="zip", replace=True

)输出：

Downloading data from https://mindspore-website.obs.cn-north-4.myhuaweicloud.com/notebook/datasets/MNIST_Data.zip (10.3 MB)

file_sizes: 100%|███████████████████████████| 10.8M/10.8M [00:00<00:00, 150MB/s]
Extracting zip file...
Successfully downloaded / unzipped to ./

下载的MNIST数据集目录结构：

MNIST_Data

└── train

    ├── train-images-idx3-ubyte (60000个训练图片)

    ├── train-labels-idx1-ubyte (60000个训练标签)

└── test

    ├── t10k-images-idx3-ubyte (10000个测试图片)

    ├── t10k-labels-idx1-ubyte (10000个测试标签)

2.加载数据

train_dataset = MnistDataset("MNIST_Data/train")
test_dataset = MnistDataset("MNIST_Data/test")

打印训练数据的列名

print(train_dataset.get_col_names())

输出：

['image', 'label']

3.数据预处理

定义数据处理流水线

用到的模块：数据集Dataset、数据变换Transforms

用到的操作：map、batch

def datapipe(dataset, batch_size):
    image_transforms = [
        vision.Rescale(1.0 / 255.0, 0),
        vision.Normalize(mean=(0.1307,), std=(0.3081,)),
        vision.HWC2CHW(),
    ]
    label_transform = transforms.TypeCast(mindspore.int32)

    dataset = dataset.map(image_transforms, "image")
    dataset = dataset.map(label_transform, "label")
    dataset = dataset.batch(batch_size)
    return dataset
# Map vision transforms and batch dataset

预处理训练数据和测试数据

train_dataset = datapipe(train_dataset, 64)
test_dataset = datapipe(test_dataset, 64)

迭代查看数据和标签的shape和datatype

for image, label in test_dataset.create_tuple_iterator():
    print(f"Shape of image [N, C, H, W]: {image.shape} {image.dtype}")
    print(f"Shape of label: {label.shape} {label.dtype}")
Break

输出：

Shape of image [N, C, H, W]: (64, 1, 28, 28) Float32
Shape of label: (64,) Int32

for data in test_dataset.create_dict_iterator():
    print(f"Shape of image [N, C, H, W]: {data['image'].shape} {data['image'].dtype}")
    print(f"Shape of label: {data['label'].shape} {data['label'].dtype}")
Break

输出：

Shape of image [N, C, H, W]: (64, 1, 28, 28) Float32
Shape of label: (64,) Int32

4.构建神经网络

mindspore.nn类        所有神经网络的基类。

nn.Cell类                 自定义网络时可以继承

        __init__            包含所有网络层的定义

        Construct          数据（Tensor）的变换过程

# Define model
class Network(nn.Cell):
    def __init__(self):
        super().__init__()
        self.flatten = nn.Flatten()
        self.dense_relu_sequential = nn.SequentialCell(
            nn.Dense(28 * 28, 512),
            nn.ReLU(),
            nn.Dense(512, 512),
            nn.ReLU(),
            nn.Dense(512, 10),
        )

    def construct(self, x):
        x = self.flatten(x)
        logits = self.dense_relu_sequential(x)
        return logits


model = Network()
print(model)

输出：

Network<
  (flatten): Flatten<>
  (dense_relu_sequential): SequentialCell<
    (0): Dense<input_channels=784, output_channels=512, has_bias=True>
    (1): ReLU<>
    (2): Dense<input_channels=512, output_channels=512, has_bias=True>
    (3): ReLU<>
    (4): Dense<input_channels=512, output_channels=10, has_bias=True>
    >
  >

5.模型训练

训练三步：

    （1）正向计算

模型预测结果（logits），与正确标签（label）求预测损失（loss）。

    （2）反向传播

自动微分求模型参数（parameters）对于loss的梯度（gradients）。

    （3）参数优化

将梯度更新到参数上。

# Instantiate loss function and optimizer
loss_fn = nn.CrossEntropyLoss()
optimizer = nn.SGD(model.trainable_params(), 1e-2)


# 1. Define forward function
def forward_fn(data, label):
    logits = model(data)
    loss = loss_fn(logits, label)
    return loss, logits


# 2. Get gradient function
grad_fn = mindspore.value_and_grad(forward_fn, None, optimizer.parameters, has_aux=True)


# 3. Define function of one-step training
def train_step(data, label):
    (loss, _), grads = grad_fn(data, label)
    optimizer(grads)
    return loss


def train(model, dataset):
    size = dataset.get_dataset_size()
    model.set_train()
    for batch, (data, label) in enumerate(dataset.create_tuple_iterator()):
        loss = train_step(data, label)

        if batch % 100 == 0:
            loss, current = loss.asnumpy(), batch
            print(f"loss: {loss:>7f}  [{current:>3d}/{size:>3d}]")

6.测试评估

定义测试函数

def test(model, dataset, loss_fn):
    num_batches = dataset.get_dataset_size()
    model.set_train(False)
    total, test_loss, correct = 0, 0, 0
    for data, label in dataset.create_tuple_iterator():
        pred = model(data)
        total += len(data)
        test_loss += loss_fn(pred, label).asnumpy()
        correct += (pred.argmax(1) == label).asnumpy().sum()
    test_loss /= num_batches
    correct /= total
    print(f"Test: \n Accuracy: {(100*correct):>0.1f}%, Avg loss: {test_loss:>8f} \n")

迭代测试，这里迭代3次

epochs = 3
for t in range(epochs):
    print(f"Epoch {t+1}\n-------------------------------")
    train(model, train_dataset)
    test(model, test_dataset, loss_fn)
print("Done!")

输出：

Epoch 1
-------------------------------
loss: 2.307619  [  0/938]
loss: 1.641935  [100/938]
loss: 0.906887  [200/938]
loss: 0.559807  [300/938]
loss: 0.402098  [400/938]
loss: 0.356145  [500/938]
loss: 0.349220  [600/938]
loss: 0.398980  [700/938]
loss: 0.428066  [800/938]
loss: 0.239198  [900/938]
Test: 
 Accuracy: 91.0%, Avg loss: 0.313863 

Epoch 2
-------------------------------
loss: 0.383306  [  0/938]
loss: 0.262217  [100/938]
loss: 0.261673  [200/938]
loss: 0.659807  [300/938]
loss: 0.188391  [400/938]
loss: 0.288512  [500/938]
loss: 0.188541  [600/938]
loss: 0.215372  [700/938]
loss: 0.321593  [800/938]
loss: 0.179319  [900/938]
Test: 
 Accuracy: 92.9%, Avg loss: 0.246736 

Epoch 3
-------------------------------
loss: 0.374724  [  0/938]
loss: 0.198583  [100/938]
loss: 0.129807  [200/938]
loss: 0.144729  [300/938]
loss: 0.140285  [400/938]
loss: 0.319090  [500/938]
loss: 0.408056  [600/938]
loss: 0.278954  [700/938]
loss: 0.206172  [800/938]
loss: 0.107220  [900/938]
Test: 
 Accuracy: 94.1%, Avg loss: 0.204920 

Done!

7.保存导出模型

# Save checkpoint
mindspore.save_checkpoint(model, "model.ckpt")
print("Saved Model to model.ckpt")

输出：

Saved Model to model.ckpt

8.加载模型

重构模型

加载模型参数

# Instantiate a random initialized model
model = Network()
# Load checkpoint and load parameter to model
param_dict = mindspore.load_checkpoint("model.ckpt")
param_not_load, _ = mindspore.load_param_into_net(model, param_dict)
print(param_not_load)

输出：

[]

param_not_load是未加载的参数，为空表示全部参数都已加载

9.应用模型

model.set_train(False)
for data, label in test_dataset:
    pred = model(data)
    predicted = pred.argmax(1)
    print(f'Predicted: "{predicted[:10]}", Actual: "{label[:10]}"')
    break

输出：

Predicted: "[9 2 5 1 1 9 1 3 0 8]", Actual: "[9 2 5 1 1 9 1 3 0 8]"