MindSpore平台的体验与部署

第一章 MindSpore框架的体验与部署

---

目录

MindSpore系列第一篇

前言

一、昇思是什么

二、使用步骤

1.在anaconda上部署MindSpore

2.简单体验昇思框架

总结

---

---

前言

随着人工智能的不断发展，国内外的ai平台越来越多，今天就来体验一下华为旗下的国产ai平台，昇思（MindSpore）。

---

一、昇思是什么

MindSpore是由华为于2019年8月推出的新一代全场景AI计算框架，2020年3月28日，华为宣布MindSpore正式开源。昇思作为新一代ai框架有着自己得天独厚的优势。

二、使用步骤

1.在anaconda上部署MindSpore

在昇思官网安装指导页面获取自己对应的安装代码。这里，笔者使用的是Windows系统的anaconda环境,使用gpu进行运算。读者可以根据自己的机器环境进行配置。

在这里我们使用，anaconda navigator新建一个新的虚拟环境。

在新的环境的python环境下，输入刚才复制的代码。 

出现问题，报错显示 

 让我们跟着教程重新安装一遍环境，依旧失败。

百度一下，发现是安装源的问题。

几经周折，数次调整后依旧无法安装，可能是清华源的问题。

最终选择使用pip安装。

 成功安装并验证

pip代码如下：

pip install https://ms-release.obs.cn-north-4.myhuaweicloud.com/1.9.0/MindSpore/cpu/x86_64/mindspore-1.9.0-cp39-cp39-win_amd64.whl --trusted-host ms-release.obs.cn-north-4.myhuaweicloud.com -i https://pypi.tuna.tsinghua.edu.cn/simple

2.简单体验昇思框架

继续在navigator中，选择安装notebook

 安装完成后打开notebook

 简单运行官方的代码样例

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

导入成功 

继续运行，下载数据集。

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)

出现报错

---------------------------------------------------------------------------
ModuleNotFoundError                       Traceback (most recent call last)
Cell In [2], line 1
----> 1 from download import download
      3 url = "https://mindspore-website.obs.cn-north-4.myhuaweicloud.com/" \
      4       "notebook/datasets/MNIST_Data.zip"
      5 path = download(url, "./", kind="zip", replace=True)

ModuleNotFoundError: No module named 'download'

使用pip安装download模块

pip install download

安装失败

选择conda，找不到对应的模块。

最终选择手动下载

https://mindspore-website.obs.cn-north-4.myhuaweicloud.com/notebook/datasets/MNIST_Data.zip

移动到笔记本对应的个人文件夹目录并解压

成功跳过下载并运行下面的代码

train_dataset = MnistDataset('MNIST_Data/train')
test_dataset = MnistDataset('MNIST_Data/test')
print(train_dataset.get_col_names())

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

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

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

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

# 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)

Network<
  (flatten): Flatten<>
  (dense_relu_sequential): SequentialCell<
    (0): Dense
    (1): ReLU<>
    (2): Dense
    (3): ReLU<>
    (4): Dense
    >
  >

定义损失函数与优化器

loss_fn = nn.CrossEntropyLoss()
optimizer = nn.SGD(model.trainable_params(), 1e-2)

定义训练函数：

def train(model, dataset, loss_fn, optimizer):
    # Define forward function
    def forward_fn(data, label):
        logits = model(data)
        loss = loss_fn(logits, label)
        return loss, logits

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

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

    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}]")

定义测试函数，评估性能

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")

最终运行模型：

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

Epoch 1
-------------------------------
loss: 2.303244  [  0/938]
loss: 2.285310  [100/938]
loss: 2.254808  [200/938]
loss: 2.173993  [300/938]
loss: 1.815277  [400/938]
loss: 1.376053  [500/938]
loss: 1.102494  [600/938]
loss: 0.804129  [700/938]
loss: 0.569727  [800/938]
loss: 0.633577  [900/938]
Test: 
 Accuracy: 83.7%, Avg loss: 0.537655 

Epoch 2
-------------------------------
loss: 0.456971  [  0/938]
loss: 0.687430  [100/938]
loss: 0.434879  [200/938]
loss: 0.502015  [300/938]
loss: 0.356903  [400/938]
loss: 0.405563  [500/938]
loss: 0.350419  [600/938]
loss: 0.280745  [700/938]
loss: 0.249667  [800/938]
loss: 0.418286  [900/938]
Test: 
 Accuracy: 90.2%, Avg loss: 0.336636 

Epoch 3
-------------------------------
loss: 0.382209  [  0/938]
loss: 0.376853  [100/938]
loss: 0.325569  [200/938]
loss: 0.358031  [300/938]
loss: 0.303792  [400/938]
loss: 0.420408  [500/938]
loss: 0.470213  [600/938]
loss: 0.286934  [700/938]
loss: 0.188889  [800/938]
loss: 0.246530  [900/938]
Test: 
 Accuracy: 91.9%, Avg loss: 0.281553 

Done!

---

总结

本文对mindspore框架的部署以及会遇到的问题进行了简单的介绍，同时通过运行官网介绍的例子初步体验了昇思框架的优越性。

总的来说，在安装方面依旧有着部分的兼容问题，但是可以通过pip解决。而MindSpore的优点则是较高的性能与对用户友好。在实际运行阶段，快速完成了对模型的训练。将来还期望能够进一步体验它的多层api、与昇腾计算的高效联动、动态图和静态图统一的编码方式等其他优点。

作为一款国产机器学习框架，昇思未来十分光明。