{
"cells": [
{
"cell_type": "markdown",
"metadata": {},
"source": [
"# 评价指标\n",
"\n",
"[](https://openi.pcl.ac.cn/MindSpore/docs/src/branch/r1.9/tutorials/source_zh_cn/advanced/model/metric.ipynb?card=2&image=MindSpore1.8.1) [](https://mindspore-website.obs.cn-north-4.myhuaweicloud.com/notebook/r1.9/tutorials/zh_cn/advanced/model/mindspore_metric.ipynb) [](https://mindspore-website.obs.cn-north-4.myhuaweicloud.com/notebook/r1.9/tutorials/zh_cn/advanced/model/mindspore_metric.py) [](https://gitee.com/mindspore/docs/blob/r1.9/tutorials/source_zh_cn/advanced/model/metric.ipynb)\n",
"\n",
"当训练任务结束，常常需要评价函数（Metrics）来评估模型的好坏。不同的训练任务往往需要不同的Metrics函数。例如，对于二分类问题，常用的评价指标有precision（准确率）、recall（召回率）等，而对于多分类任务，可使用宏平均（Macro）和微平均（Micro）来评估。\n",
"\n",
"MindSpore提供了大部分常见任务的评价函数，如`nn.Accuracy`、`nn.Precision`、`nn.MAE`和`nn.MSE`等，由于MindSpore提供的评价函数无法满足所有任务的需求，很多情况下用户需要针对具体的任务自定义Metrics来评估训练的模型。\n",
"\n",
"本章主要介绍如何自定义Metrics以及如何在`nn.Model`中使用Metrics。\n",
"\n",
"> 详情可参考[评价指标](https://www.mindspore.cn/docs/zh-CN/r1.9/api_python/mindspore.nn.html#评价指标)。\n",
"\n",
"## 自定义Metrics\n",
"\n",
"自定义Metrics函数需要继承`nn.Metric`父类，并重新实现父类中的`clear`方法、`update`方法和`eval`方法。\n",
"\n",
"- `clear`：初始化相关的内部参数。\n",
"- `update`：接收网络预测输出和标签，计算误差，每次step后并更新内部评估结果。\n",
"- `eval`：计算最终评估结果，在没次epoch结束后计算最终的评估结果。\n",
"\n",
"平均绝对误差（MAE）算法如式(1)所示：\n",
"\n",
"$$ MAE=\\frac{1}{n}\\sum_{i=1}^n\\lvert ypred_i - y_i \\rvert \\tag{1}$$\n",
"\n",
"下面以简单的MAE算法为例，介绍`clear`、`update`和`eval`三个函数及其使用方法。"
]
},
{
"cell_type": "code",
"execution_count": 1,
"metadata": {
"ExecuteTime": {
"end_time": "2022-01-04T11:04:40.488396Z",
"start_time": "2022-01-04T11:04:38.090036Z"
}
},
"outputs": [
{
"name": "stdout",
"output_type": "stream",
"text": [
"0.1499999612569809\n"
]
}
],
"source": [
"import numpy as np\n",
"import mindspore as ms\n",
"from mindspore import nn\n",
"\n",
"class MyMAE(nn.Metric):\n",
" def __init__(self):\n",
" super(MyMAE, self).__init__()\n",
" self.clear()\n",
"\n",
" def clear(self):\n",
" \"\"\"初始化变量_abs_error_sum和_samples_num\"\"\"\n",
" self._abs_error_sum = 0 # 保存误差和\n",
" self._samples_num = 0 # 累计数据量\n",
"\n",
" def update(self, *inputs):\n",
" \"\"\"更新_abs_error_sum和_samples_num\"\"\"\n",
" y_pred = inputs[0].asnumpy()\n",
" y = inputs[1].asnumpy()\n",
"\n",
" # 计算预测值与真实值的绝对误差\n",
" abs_error_sum = np.abs(y - y_pred)\n",
" self._abs_error_sum += abs_error_sum.sum()\n",
"\n",
" # 样本的总数\n",
" self._samples_num += y.shape[0]\n",
"\n",
" def eval(self):\n",
" \"\"\"计算最终评估结果\"\"\"\n",
" return self._abs_error_sum / self._samples_num\n",
"\n",
"# 网络有两个输出\n",
"y_pred = ms.Tensor(np.array([[0.1, 0.2, 0.6, 0.9], [0.1, 0.2, 0.6, 0.9]]), ms.float32)\n",
"y = ms.Tensor(np.array([[0.1, 0.25, 0.7, 0.9], [0.1, 0.25, 0.7, 0.9]]), ms.float32)\n",
"\n",
"error = MyMAE()\n",
"error.clear()\n",
"error.update(y_pred, y)\n",
"result = error.eval()\n",
"print(result)"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"## 模型训练中使用Metrics\n",
"\n",
"[mindspore.Model](https://www.mindspore.cn/docs/zh-CN/r1.9/api_python/mindspore/mindspore.Model.html#mindspore.Model)是用于训练和评估的高层API，可以将自定义或MindSpore已有的Metrics作为参数传入，Model能够自动调用传入的Metrics进行评估。\n",
"\n",
"在网络模型训练后，需要使用评价指标，来评估网络模型的训练效果，因此在演示具体代码之前首先简单拟定数据集，对数据集进行加载和定义一个简单的线性回归网络模型：\n",
"\n",
"$$f(x)=w*x+b \\tag{2}$$"
]
},
{
"cell_type": "code",
"execution_count": 2,
"metadata": {},
"outputs": [],
"source": [
"import numpy as np\n",
"from mindspore import dataset as ds\n",
"\n",
"def get_data(num, w=2.0, b=3.0):\n",
" \"\"\"生成数据及对应标签\"\"\"\n",
" for _ in range(num):\n",
" x = np.random.uniform(-10.0, 10.0)\n",
" noise = np.random.normal(0, 1)\n",
" y = x * w + b + noise\n",
" yield np.array([x]).astype(np.float32), np.array([y]).astype(np.float32)\n",
"\n",
"def create_dataset(num_data, batch_size=16):\n",
" \"\"\"加载数据集\"\"\"\n",
" dataset = ds.GeneratorDataset(list(get_data(num_data)), column_names=['data', 'label'])\n",
" dataset = dataset.batch(batch_size)\n",
" return dataset"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"### 使用内置评价指标\n",
"\n",
"使用MindSpore内置的Metrics作为参数传入Model时，Metrics可以定义为一个字典类型，字典的key值为字符串类型，字典的value值为MindSpore内置的[评价指标](https://www.mindspore.cn/docs/zh-CN/r1.9/api_python/mindspore.nn.html#评价指标)，如下示例使用`nn.Accuracy`计算分类的准确率。"
]
},
{
"cell_type": "code",
"execution_count": 3,
"metadata": {},
"outputs": [
{
"name": "stdout",
"output_type": "stream",
"text": [
"epoch: 1 step: 10, loss is 6.0811052322387695\n",
"Eval result: epoch 1, metrics: {'MAE': 5.012505912780762}\n",
"epoch: 2 step: 10, loss is 2.7896716594696045\n",
"Eval result: epoch 2, metrics: {'MAE': 3.380072832107544}\n",
"epoch: 3 step: 10, loss is 3.0297815799713135\n",
"Eval result: epoch 3, metrics: {'MAE': 2.5002413272857664}\n",
"epoch: 4 step: 10, loss is 2.3680481910705566\n",
"Eval result: epoch 4, metrics: {'MAE': 2.4334578275680543}\n",
"epoch: 5 step: 10, loss is 1.8126990795135498\n",
"Eval result: epoch 5, metrics: {'MAE': 1.8317200541496277}\n",
"epoch: 6 step: 10, loss is 1.6006351709365845\n",
"Eval result: epoch 6, metrics: {'MAE': 1.521335732936859}\n",
"epoch: 7 step: 10, loss is 1.1064929962158203\n",
"Eval result: epoch 7, metrics: {'MAE': 1.2528185725212098}\n",
"epoch: 8 step: 10, loss is 0.9595810174942017\n",
"Eval result: epoch 8, metrics: {'MAE': 1.0719563841819764}\n",
"epoch: 9 step: 10, loss is 0.6517931222915649\n",
"Eval result: epoch 9, metrics: {'MAE': 0.9766222715377808}\n",
"epoch: 10 step: 10, loss is 0.9312882423400879\n",
"Eval result: epoch 10, metrics: {'MAE': 0.9238077104091644}\n"
]
}
],
"source": [
"import mindspore.nn as nn\n",
"from mindspore.nn import MAE\n",
"from mindspore import Model, LossMonitor\n",
"\n",
"net = nn.Dense(1, 1)\n",
"loss_fn = nn.L1Loss()\n",
"optimizer = nn.Momentum(net.trainable_params(), learning_rate=0.005, momentum=0.9)\n",
"\n",
"# 定义model，使用内置的Accuracy函数\n",
"model = Model(net, loss_fn, optimizer, metrics={\"MAE\": MAE()})\n",
"\n",
"train_dataset = create_dataset(num_data=160)\n",
"eval_dataset = create_dataset(num_data=160)\n",
"train_dataset_size = train_dataset.get_dataset_size()\n",
"\n",
"model.fit(10, train_dataset, eval_dataset, callbacks=LossMonitor(train_dataset_size))"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"### 使用自定义评价指标\n",
"\n",
"如下示例在`Model`中传入上述自定义的评估指标`MAE()`，将验证数据集传入`model.fit()`接口边训练边验证。\n",
"\n",
"验证结果为一个字典类型，验证结果的key值与`metrics`的key值相同，验证结果的value值为预测值与实际值的平均绝对误差。"
]
},
{
"cell_type": "code",
"execution_count": 4,
"metadata": {},
"outputs": [
{
"name": "stdout",
"output_type": "stream",
"text": [
"epoch: 1 step: 10, loss is 0.5679571628570557\n",
"Eval result: epoch 1, metrics: {'MAE': 0.7907268464565277}\n",
"epoch: 2 step: 10, loss is 0.8198273181915283\n",
"Eval result: epoch 2, metrics: {'MAE': 0.7729107916355134}\n",
"epoch: 3 step: 10, loss is 0.5721814036369324\n",
"Eval result: epoch 3, metrics: {'MAE': 0.7661101937294006}\n",
"epoch: 4 step: 10, loss is 0.6523740291595459\n",
"Eval result: epoch 4, metrics: {'MAE': 0.7704753875732422}\n",
"epoch: 5 step: 10, loss is 0.5641313791275024\n",
"Eval result: epoch 5, metrics: {'MAE': 0.7609358102083206}\n",
"epoch: 6 step: 10, loss is 0.774018406867981\n",
"Eval result: epoch 6, metrics: {'MAE': 0.7739883124828338}\n",
"epoch: 7 step: 10, loss is 0.7306548357009888\n",
"Eval result: epoch 7, metrics: {'MAE': 0.7757290184497834}\n",
"epoch: 8 step: 10, loss is 0.667199969291687\n",
"Eval result: epoch 8, metrics: {'MAE': 0.7627188444137574}\n",
"epoch: 9 step: 10, loss is 0.689708948135376\n",
"Eval result: epoch 9, metrics: {'MAE': 0.7673796474933624}\n",
"epoch: 10 step: 10, loss is 0.7661054134368896\n",
"Eval result: epoch 10, metrics: {'MAE': 0.7654164433479309}\n"
]
}
],
"source": [
"train_dataset = create_dataset(num_data=160)\n",
"eval_dataset = create_dataset(num_data=160)\n",
"\n",
"model = Model(net, loss_fn, optimizer, metrics={\"MAE\": MyMAE()})\n",
"\n",
"# 定义model，将自定义metrics函数MAE传入Model中\n",
"model.fit(10, train_dataset, eval_dataset, callbacks=LossMonitor(train_dataset_size))"
]
}
],
"metadata": {
"kernelspec": {
"display_name": "MindSpore",
"language": "python",
"name": "mindspore"
},
"language_info": {
"codemirror_mode": {
"name": "ipython",
"version": 3
},
"file_extension": ".py",
"mimetype": "text/x-python",
"name": "python",
"nbconvert_exporter": "python",
"pygments_lexer": "ipython3",
"version": "3.7.11"
}
},
"nbformat": 4,
"nbformat_minor": 4
}