本文介绍使用飞桨EasyEdge平台部署蘑菇分类模型的流程。先定义图像分类任务，解压并标注蘑菇数据集，划分训练集和验证集，定义数据集类并做图像增强。选用mobilenet_v2网络，配置优化器训练模型，最后保存为静态图，通过EasyEdge平台完成部署，操作简洁。
☞☞☞AI 智能聊天, 问答助手, AI 智能搜索, 免费无限量使用 DeepSeek R1 模型☜☜☜
项目背景
飞桨最近新上了EasyEdge端与边缘AI服务平台，这对于新手来说非常友好
在原来我们开发了模型之后，没有办法快速部署到手机上
现在有了EasyEdge这个平台，直接在EasyEdge端与边缘AI服务平台部署就可以了，操作十分简洁流畅！下面以一个示例来给大家展示：
① 问题定义
对于一个任务，当你想使用深度学习来解决时，一般流程如下：
①问题定义->②数据准备->③模型选择与开发->④模型训练和调优->⑤模型评估测试->⑥部署上线
本项目中的蘑菇的分类的本质是图像分类任务，采用轻量级卷积神经网络mobilenet_v2进行相关实践。
② 数据准备
2.1 解压缩数据集
我们将网上获取的数据集以压缩包的方式上传到aistudio数据集中，并加载到我们的项目内。
在使用之前我们进行数据集压缩包的一个解压。
DM建站系统幼儿教育机构网站模板1.62
下载
DM建站系统幼儿教育机构网站模板,DM企业建站系统。是由php+mysql开发的一套专门用于中小企业网站建设的开源cms。DM系统的理念就是组装，把模板和区块组装起来，产生不同的网站效果。可以用来快速建设一个响应式的企业网站( PC，手机，微信都可以访问)。后台操作简单，维护方便。DM企业建站系统安装步骤：第一步，先用phpmyadmin导入sql文件。 第二步：把文件放到你的本地服务器，或上传到
In [ ]
# !unzip -oq /home/aistudio/data/data81902/mushrooms_train.zip -d work/
In [ ]
import paddle
paddle.seed(8888)import numpy as npfrom typing import Callable#参数配置config_parameters = { "class_dim": 9, #分类数
"target_path":"/home/aistudio/work/",
'train_image_dir': '/home/aistudio/work/trainImages', 'eval_image_dir': '/home/aistudio/work/evalImages', 'epochs':100, 'batch_size': 128, 'lr': 0.01}
2.2 数据标注
我们先看一下解压缩后的数据集长成什么样子。
In [ ]
import osimport randomfrom matplotlib import pyplot as pltfrom PIL import Image
imgs = []
paths = os.listdir('work/mushrooms_train')for path in paths:
img_path = os.path.join('work/mushrooms_train', path) if os.path.isdir(img_path):
img_paths = os.listdir(img_path)
img = Image.open(os.path.join(img_path, random.choice(img_paths)))
imgs.append((img, path))
f, ax = plt.subplots(3, 3, figsize=(12,12))for i, img in enumerate(imgs[:9]):
ax[i//3, i%3].imshow(img[0])
ax[i//3, i%3].axis('off')
ax[i//3, i%3].set_title('label: %s' % img[1])
plt.show()
2.3 划分数据集与数据集的定义
接下来我们使用标注好的文件进行数据集类的定义，方便后续模型训练使用。
2.3.1 划分数据集
In [ ]
# import os# import shutil# train_dir = config_parameters['train_image_dir']# eval_dir = config_parameters['eval_image_dir']# paths = os.listdir('work/mushrooms_train')# if not os.path.exists(train_dir):# os.mkdir(train_dir)# if not os.path.exists(eval_dir):# os.mkdir(eval_dir)# for path in paths:# imgs_dir = os.listdir(os.path.join('work/mushrooms_train', path))# target_train_dir = os.path.join(train_dir,path)# target_eval_dir = os.path.join(eval_dir,path)# if not os.path.exists(target_train_dir):# os.mkdir(target_train_dir)# if not os.path.exists(target_eval_dir):# os.mkdir(target_eval_dir)# for i in range(len(imgs_dir)):# if ' ' in imgs_dir[i]:# new_name = imgs_dir[i].replace(' ', '_')# else:# new_name = imgs_dir[i]# target_train_path = os.path.join(target_train_dir, new_name)# target_eval_path = os.path.join(target_eval_dir, new_name) # if i % 5 == 0:# shutil.copyfile(os.path.join(os.path.join('work/mushrooms_train', path), imgs_dir[i]), target_eval_path)# else:# shutil.copyfile(os.path.join(os.path.join('work/mushrooms_train', path), imgs_dir[i]), target_train_path)# # print('finished train val split!')
2.3.2 导入数据集的定义实现
In [ ]
#数据集的定义class TowerDataset(paddle.io.Dataset):
"""
步骤一：继承paddle.io.Dataset类
"""
def __init__(self, transforms: Callable, mode: str ='train'):
"""
步骤二：实现构造函数，定义数据读取方式
"""
super(TowerDataset, self).__init__()
self.mode = mode
self.transforms = transforms
train_image_dir = config_parameters['train_image_dir']
eval_image_dir = config_parameters['eval_image_dir']
train_data_folder = paddle.vision.DatasetFolder(train_image_dir)
eval_data_folder = paddle.vision.DatasetFolder(eval_image_dir)
if self.mode == 'train':
self.data = train_data_folder elif self.mode == 'eval':
self.data = eval_data_folder def __getitem__(self, index):
"""
步骤三：实现__getitem__方法，定义指定index时如何获取数据，并返回单条数据（训练数据，对应的标签）
"""
data = np.array(self.data[index][0]).astype('float32')
data = self.transforms(data)
label = np.array([self.data[index][1]]).astype('int64')
return data, label
def __len__(self):
"""
步骤四：实现__len__方法，返回数据集总数目
"""
return len(self.data)
2.3.3 实例化数据集类及图像增强
根据所使用的数据集需求实例化数据集类，并查看总样本量。
In [ ]
from paddle.vision import transforms as T#数据增强transform_train =T.Compose([T.Resize((256,256)),
T.RandomHorizontalFlip(10),
T.RandomRotation(10),
T.Transpose(),
T.Normalize(mean=[0, 0, 0], # 像素值归一化
std =[255, 255, 255]), # transforms.ToTensor(), # transpose操作 + (img / 255),并且数据结构变为PaddleTensor
T.Normalize(mean=[0.50950350, 0.54632660, 0.57409690],# 减均值 除标准差
std= [0.26059777, 0.26041326, 0.29220656])# 计算过程：output[channel] = (input[channel] - mean[channel]) / std[channel]
])
transform_eval =T.Compose([ T.Resize((256,256)),
T.Transpose(),
T.Normalize(mean=[0, 0, 0], # 像素值归一化
std =[255, 255, 255]), # transforms.ToTensor(), # transpose操作 + (img / 255),并且数据结构变为PaddleTensor
T.Normalize(mean=[0.50950350, 0.54632660, 0.57409690],# 减均值 除标准差
std= [0.26059777, 0.26041326, 0.29220656])# 计算过程：output[channel] = (input[channel] - mean[channel]) / std[channel]
])
In [ ]
train_dataset = TowerDataset(mode='train',transforms=transform_train)
eval_dataset = TowerDataset(mode='eval', transforms=transform_eval )#数据异步加载train_loader = paddle.io.DataLoader(train_dataset,
places=paddle.CUDAPlace(0),
batch_size=128,
shuffle=True, #num_workers=2,
#use_shared_memory=True
)
eval_loader = paddle.io.DataLoader (eval_dataset,
places=paddle.CUDAPlace(0),
batch_size=128, #num_workers=2,
#use_shared_memory=True
)
In [ ]
print('训练集样本量: {}，验证集样本量: {}'.format(len(train_loader), len(eval_loader)))
训练集样本量: 42，验证集样本量: 11
③ 模型选择和开发
3.1 网络构建
本次我们使用mobilenet_v2网络来完成我们的案例实践。
In [ ]
network=paddle.vision.models.mobilenet_v2(pretrained=True,num_classes=9) model=paddle.Model(network) model.summary((-1, 3, 256, 256))
2021-06-03 21:51:44,710 - INFO - unique_endpoints {''}
2021-06-03 21:51:44,711 - INFO - Downloading mobilenet_v2_x1.0.pdparams from https://paddle-hapi.bj.bcebos.com/models/mobilenet_v2_x1.0.pdparams
100%|██████████| 20795/20795 [00:00<00:00, 22406.24it/s]
2021-06-03 21:51:46,068 - INFO - File /home/aistudio/.cache/paddle/hapi/weights/mobilenet_v2_x1.0.pdparams md5 checking...
/opt/conda/envs/python35-paddle120-env/lib/python3.7/site-packages/paddle/fluid/dygraph/layers.py:1303: UserWarning: Skip loading for classifier.1.weight. classifier.1.weight receives a shape [1280, 1000], but the expected shape is [1280, 9].
warnings.warn(("Skip loading for {}. ".format(key) + str(err)))
/opt/conda/envs/python35-paddle120-env/lib/python3.7/site-packages/paddle/fluid/dygraph/layers.py:1303: UserWarning: Skip loading for classifier.1.bias. classifier.1.bias receives a shape [1000], but the expected shape is [9].
warnings.warn(("Skip loading for {}. ".format(key) + str(err)))
-------------------------------------------------------------------------------
Layer (type) Input Shape Output Shape Param #
===============================================================================
Conv2D-1 [[1, 3, 256, 256]] [1, 32, 128, 128] 864
BatchNorm2D-1 [[1, 32, 128, 128]] [1, 32, 128, 128] 128
ReLU6-1 [[1, 32, 128, 128]] [1, 32, 128, 128] 0
Conv2D-2 [[1, 32, 128, 128]] [1, 32, 128, 128] 288
BatchNorm2D-2 [[1, 32, 128, 128]] [1, 32, 128, 128] 128
ReLU6-2 [[1, 32, 128, 128]] [1, 32, 128, 128] 0
Conv2D-3 [[1, 32, 128, 128]] [1, 16, 128, 128] 512
BatchNorm2D-3 [[1, 16, 128, 128]] [1, 16, 128, 128] 64
InvertedResidual-1 [[1, 32, 128, 128]] [1, 16, 128, 128] 0
Conv2D-4 [[1, 16, 128, 128]] [1, 96, 128, 128] 1,536
BatchNorm2D-4 [[1, 96, 128, 128]] [1, 96, 128, 128] 384
ReLU6-3 [[1, 96, 128, 128]] [1, 96, 128, 128] 0
Conv2D-5 [[1, 96, 128, 128]] [1, 96, 64, 64] 864
BatchNorm2D-5 [[1, 96, 64, 64]] [1, 96, 64, 64] 384
ReLU6-4 [[1, 96, 64, 64]] [1, 96, 64, 64] 0
Conv2D-6 [[1, 96, 64, 64]] [1, 24, 64, 64] 2,304
BatchNorm2D-6 [[1, 24, 64, 64]] [1, 24, 64, 64] 96
InvertedResidual-2 [[1, 16, 128, 128]] [1, 24, 64, 64] 0
Conv2D-7 [[1, 24, 64, 64]] [1, 144, 64, 64] 3,456
BatchNorm2D-7 [[1, 144, 64, 64]] [1, 144, 64, 64] 576
ReLU6-5 [[1, 144, 64, 64]] [1, 144, 64, 64] 0
Conv2D-8 [[1, 144, 64, 64]] [1, 144, 64, 64] 1,296
BatchNorm2D-8 [[1, 144, 64, 64]] [1, 144, 64, 64] 576
ReLU6-6 [[1, 144, 64, 64]] [1, 144, 64, 64] 0
Conv2D-9 [[1, 144, 64, 64]] [1, 24, 64, 64] 3,456
BatchNorm2D-9 [[1, 24, 64, 64]] [1, 24, 64, 64] 96
InvertedResidual-3 [[1, 24, 64, 64]] [1, 24, 64, 64] 0
Conv2D-10 [[1, 24, 64, 64]] [1, 144, 64, 64] 3,456
BatchNorm2D-10 [[1, 144, 64, 64]] [1, 144, 64, 64] 576
ReLU6-7 [[1, 144, 64, 64]] [1, 144, 64, 64] 0
Conv2D-11 [[1, 144, 64, 64]] [1, 144, 32, 32] 1,296
BatchNorm2D-11 [[1, 144, 32, 32]] [1, 144, 32, 32] 576
ReLU6-8 [[1, 144, 32, 32]] [1, 144, 32, 32] 0
Conv2D-12 [[1, 144, 32, 32]] [1, 32, 32, 32] 4,608
BatchNorm2D-12 [[1, 32, 32, 32]] [1, 32, 32, 32] 128
InvertedResidual-4 [[1, 24, 64, 64]] [1, 32, 32, 32] 0
Conv2D-13 [[1, 32, 32, 32]] [1, 192, 32, 32] 6,144
BatchNorm2D-13 [[1, 192, 32, 32]] [1, 192, 32, 32] 768
ReLU6-9 [[1, 192, 32, 32]] [1, 192, 32, 32] 0
Conv2D-14 [[1, 192, 32, 32]] [1, 192, 32, 32] 1,728
BatchNorm2D-14 [[1, 192, 32, 32]] [1, 192, 32, 32] 768
ReLU6-10 [[1, 192, 32, 32]] [1, 192, 32, 32] 0
Conv2D-15 [[1, 192, 32, 32]] [1, 32, 32, 32] 6,144
BatchNorm2D-15 [[1, 32, 32, 32]] [1, 32, 32, 32] 128
InvertedResidual-5 [[1, 32, 32, 32]] [1, 32, 32, 32] 0
Conv2D-16 [[1, 32, 32, 32]] [1, 192, 32, 32] 6,144
BatchNorm2D-16 [[1, 192, 32, 32]] [1, 192, 32, 32] 768
ReLU6-11 [[1, 192, 32, 32]] [1, 192, 32, 32] 0
Conv2D-17 [[1, 192, 32, 32]] [1, 192, 32, 32] 1,728
BatchNorm2D-17 [[1, 192, 32, 32]] [1, 192, 32, 32] 768
ReLU6-12 [[1, 192, 32, 32]] [1, 192, 32, 32] 0
Conv2D-18 [[1, 192, 32, 32]] [1, 32, 32, 32] 6,144
BatchNorm2D-18 [[1, 32, 32, 32]] [1, 32, 32, 32] 128
InvertedResidual-6 [[1, 32, 32, 32]] [1, 32, 32, 32] 0
Conv2D-19 [[1, 32, 32, 32]] [1, 192, 32, 32] 6,144
BatchNorm2D-19 [[1, 192, 32, 32]] [1, 192, 32, 32] 768
ReLU6-13 [[1, 192, 32, 32]] [1, 192, 32, 32] 0
Conv2D-20 [[1, 192, 32, 32]] [1, 192, 16, 16] 1,728
BatchNorm2D-20 [[1, 192, 16, 16]] [1, 192, 16, 16] 768
ReLU6-14 [[1, 192, 16, 16]] [1, 192, 16, 16] 0
Conv2D-21 [[1, 192, 16, 16]] [1, 64, 16, 16] 12,288
BatchNorm2D-21 [[1, 64, 16, 16]] [1, 64, 16, 16] 256
InvertedResidual-7 [[1, 32, 32, 32]] [1, 64, 16, 16] 0
Conv2D-22 [[1, 64, 16, 16]] [1, 384, 16, 16] 24,576
BatchNorm2D-22 [[1, 384, 16, 16]] [1, 384, 16, 16] 1,536
ReLU6-15 [[1, 384, 16, 16]] [1, 384, 16, 16] 0
Conv2D-23 [[1, 384, 16, 16]] [1, 384, 16, 16] 3,456
BatchNorm2D-23 [[1, 384, 16, 16]] [1, 384, 16, 16] 1,536
ReLU6-16 [[1, 384, 16, 16]] [1, 384, 16, 16] 0
Conv2D-24 [[1, 384, 16, 16]] [1, 64, 16, 16] 24,576
BatchNorm2D-24 [[1, 64, 16, 16]] [1, 64, 16, 16] 256
InvertedResidual-8 [[1, 64, 16, 16]] [1, 64, 16, 16] 0
Conv2D-25 [[1, 64, 16, 16]] [1, 384, 16, 16] 24,576
BatchNorm2D-25 [[1, 384, 16, 16]] [1, 384, 16, 16] 1,536
ReLU6-17 [[1, 384, 16, 16]] [1, 384, 16, 16] 0
Conv2D-26 [[1, 384, 16, 16]] [1, 384, 16, 16] 3,456
BatchNorm2D-26 [[1, 384, 16, 16]] [1, 384, 16, 16] 1,536
ReLU6-18 [[1, 384, 16, 16]] [1, 384, 16, 16] 0
Conv2D-27 [[1, 384, 16, 16]] [1, 64, 16, 16] 24,576
BatchNorm2D-27 [[1, 64, 16, 16]] [1, 64, 16, 16] 256
InvertedResidual-9 [[1, 64, 16, 16]] [1, 64, 16, 16] 0
Conv2D-28 [[1, 64, 16, 16]] [1, 384, 16, 16] 24,576
BatchNorm2D-28 [[1, 384, 16, 16]] [1, 384, 16, 16] 1,536
ReLU6-19 [[1, 384, 16, 16]] [1, 384, 16, 16] 0
Conv2D-29 [[1, 384, 16, 16]] [1, 384, 16, 16] 3,456
BatchNorm2D-29 [[1, 384, 16, 16]] [1, 384, 16, 16] 1,536
ReLU6-20 [[1, 384, 16, 16]] [1, 384, 16, 16] 0
Conv2D-30 [[1, 384, 16, 16]] [1, 64, 16, 16] 24,576
BatchNorm2D-30 [[1, 64, 16, 16]] [1, 64, 16, 16] 256
InvertedResidual-10 [[1, 64, 16, 16]] [1, 64, 16, 16] 0
Conv2D-31 [[1, 64, 16, 16]] [1, 384, 16, 16] 24,576
BatchNorm2D-31 [[1, 384, 16, 16]] [1, 384, 16, 16] 1,536
ReLU6-21 [[1, 384, 16, 16]] [1, 384, 16, 16] 0
Conv2D-32 [[1, 384, 16, 16]] [1, 384, 16, 16] 3,456
BatchNorm2D-32 [[1, 384, 16, 16]] [1, 384, 16, 16] 1,536
ReLU6-22 [[1, 384, 16, 16]] [1, 384, 16, 16] 0
Conv2D-33 [[1, 384, 16, 16]] [1, 96, 16, 16] 36,864
BatchNorm2D-33 [[1, 96, 16, 16]] [1, 96, 16, 16] 384
InvertedResidual-11 [[1, 64, 16, 16]] [1, 96, 16, 16] 0
Conv2D-34 [[1, 96, 16, 16]] [1, 576, 16, 16] 55,296
BatchNorm2D-34 [[1, 576, 16, 16]] [1, 576, 16, 16] 2,304
ReLU6-23 [[1, 576, 16, 16]] [1, 576, 16, 16] 0
Conv2D-35 [[1, 576, 16, 16]] [1, 576, 16, 16] 5,184
BatchNorm2D-35 [[1, 576, 16, 16]] [1, 576, 16, 16] 2,304
ReLU6-24 [[1, 576, 16, 16]] [1, 576, 16, 16] 0
Conv2D-36 [[1, 576, 16, 16]] [1, 96, 16, 16] 55,296
BatchNorm2D-36 [[1, 96, 16, 16]] [1, 96, 16, 16] 384
InvertedResidual-12 [[1, 96, 16, 16]] [1, 96, 16, 16] 0
Conv2D-37 [[1, 96, 16, 16]] [1, 576, 16, 16] 55,296
BatchNorm2D-37 [[1, 576, 16, 16]] [1, 576, 16, 16] 2,304
ReLU6-25 [[1, 576, 16, 16]] [1, 576, 16, 16] 0
Conv2D-38 [[1, 576, 16, 16]] [1, 576, 16, 16] 5,184
BatchNorm2D-38 [[1, 576, 16, 16]] [1, 576, 16, 16] 2,304
ReLU6-26 [[1, 576, 16, 16]] [1, 576, 16, 16] 0
Conv2D-39 [[1, 576, 16, 16]] [1, 96, 16, 16] 55,296
BatchNorm2D-39 [[1, 96, 16, 16]] [1, 96, 16, 16] 384
InvertedResidual-13 [[1, 96, 16, 16]] [1, 96, 16, 16] 0
Conv2D-40 [[1, 96, 16, 16]] [1, 576, 16, 16] 55,296
BatchNorm2D-40 [[1, 576, 16, 16]] [1, 576, 16, 16] 2,304
ReLU6-27 [[1, 576, 16, 16]] [1, 576, 16, 16] 0
Conv2D-41 [[1, 576, 16, 16]] [1, 576, 8, 8] 5,184
BatchNorm2D-41 [[1, 576, 8, 8]] [1, 576, 8, 8] 2,304
ReLU6-28 [[1, 576, 8, 8]] [1, 576, 8, 8] 0
Conv2D-42 [[1, 576, 8, 8]] [1, 160, 8, 8] 92,160
BatchNorm2D-42 [[1, 160, 8, 8]] [1, 160, 8, 8] 640
InvertedResidual-14 [[1, 96, 16, 16]] [1, 160, 8, 8] 0
Conv2D-43 [[1, 160, 8, 8]] [1, 960, 8, 8] 153,600
BatchNorm2D-43 [[1, 960, 8, 8]] [1, 960, 8, 8] 3,840
ReLU6-29 [[1, 960, 8, 8]] [1, 960, 8, 8] 0
Conv2D-44 [[1, 960, 8, 8]] [1, 960, 8, 8] 8,640
BatchNorm2D-44 [[1, 960, 8, 8]] [1, 960, 8, 8] 3,840
ReLU6-30 [[1, 960, 8, 8]] [1, 960, 8, 8] 0
Conv2D-45 [[1, 960, 8, 8]] [1, 160, 8, 8] 153,600
BatchNorm2D-45 [[1, 160, 8, 8]] [1, 160, 8, 8] 640
InvertedResidual-15 [[1, 160, 8, 8]] [1, 160, 8, 8] 0
Conv2D-46 [[1, 160, 8, 8]] [1, 960, 8, 8] 153,600
BatchNorm2D-46 [[1, 960, 8, 8]] [1, 960, 8, 8] 3,840
ReLU6-31 [[1, 960, 8, 8]] [1, 960, 8, 8] 0
Conv2D-47 [[1, 960, 8, 8]] [1, 960, 8, 8] 8,640
BatchNorm2D-47 [[1, 960, 8, 8]] [1, 960, 8, 8] 3,840
ReLU6-32 [[1, 960, 8, 8]] [1, 960, 8, 8] 0
Conv2D-48 [[1, 960, 8, 8]] [1, 160, 8, 8] 153,600
BatchNorm2D-48 [[1, 160, 8, 8]] [1, 160, 8, 8] 640
InvertedResidual-16 [[1, 160, 8, 8]] [1, 160, 8, 8] 0
Conv2D-49 [[1, 160, 8, 8]] [1, 960, 8, 8] 153,600
BatchNorm2D-49 [[1, 960, 8, 8]] [1, 960, 8, 8] 3,840
ReLU6-33 [[1, 960, 8, 8]] [1, 960, 8, 8] 0
Conv2D-50 [[1, 960, 8, 8]] [1, 960, 8, 8] 8,640
BatchNorm2D-50 [[1, 960, 8, 8]] [1, 960, 8, 8] 3,840
ReLU6-34 [[1, 960, 8, 8]] [1, 960, 8, 8] 0
Conv2D-51 [[1, 960, 8, 8]] [1, 320, 8, 8] 307,200
BatchNorm2D-51 [[1, 320, 8, 8]] [1, 320, 8, 8] 1,280
InvertedResidual-17 [[1, 160, 8, 8]] [1, 320, 8, 8] 0
Conv2D-52 [[1, 320, 8, 8]] [1, 1280, 8, 8] 409,600
BatchNorm2D-52 [[1, 1280, 8, 8]] [1, 1280, 8, 8] 5,120
ReLU6-35 [[1, 1280, 8, 8]] [1, 1280, 8, 8] 0
AdaptiveAvgPool2D-1 [[1, 1280, 8, 8]] [1, 1280, 1, 1] 0
Dropout-1 [[1, 1280]] [1, 1280] 0
Linear-1 [[1, 1280]] [1, 9] 11,529
===============================================================================
Total params: 2,269,513
Trainable params: 2,201,289
Non-trainable params: 68,224
-------------------------------------------------------------------------------
Input size (MB): 0.75
Forward/backward pass size (MB): 199.66
Params size (MB): 8.66
Estimated Total Size (MB): 209.07
-------------------------------------------------------------------------------
{'total_params': 2269513, 'trainable_params': 2201289}
④ 模型训练和优化器的选择
In [ ]
#优化器选择class SaveBestModel(paddle.callbacks.Callback):
def __init__(self, target=0.5, path='work/best_model', verbose=0):
self.target = target
self.epoch = None
self.path = path def on_epoch_end(self, epoch, logs=None):
self.epoch = epoch def on_eval_end(self, logs=None):
if logs.get('acc') > self.target:
self.target = logs.get('acc')
self.model.save(self.path) print('best acc is {} at epoch {}'.format(self.target, self.epoch))
callback_visualdl = paddle.callbacks.VisualDL(log_dir='work/mushroom')
callback_savebestmodel = SaveBestModel(target=0.5, path='work/best_model')
callbacks = [callback_visualdl, callback_savebestmodel]
base_lr = config_parameters['lr']
epochs = config_parameters['epochs']def make_optimizer(parameters=None):
momentum = 0.9
learning_rate= paddle.optimizer.lr.CosineAnnealingDecay(learning_rate=base_lr, T_max=epochs, verbose=False)
weight_decay=paddle.regularizer.L2Decay(0.0001)
optimizer = paddle.optimizer.Momentum(
learning_rate=learning_rate,
momentum=momentum,
weight_decay=weight_decay,
parameters=parameters) return optimizer
optimizer = make_optimizer(model.parameters())
In [ ]
model.prepare(optimizer,
paddle.nn.CrossEntropyLoss(),
paddle.metric.Accuracy())
In [14]
model.fit(train_loader,
eval_loader,
epochs=100,
batch_size=128, # 是否打乱样本集
callbacks=callbacks,
verbose=1) # 日志展示格式
⑤模型评估测试
⑥部署上线
将我们训练得到的模型进行保存为静态图，得到mushroom.pdmodel和mushroom.pdiparams两个文件，准备一个label_list.txt文件
In [15]
model.save('mushroom',training=False)
6.1上传原模型
根据图示进行上传文件验证，生成Demo
6.2生成端模型
6.3Demo下载体验
