搭建CNN模型，dropout层报错

[JJYjackson]

• 标题：搭建CNN模型，dropout层报错
• 版本、环境信息： Paddle version: 2.0.1 Paddle With CUDA: True OS: debian buster/sid Python version: 3.7.10 CUDA version: 10.0.130 cuDNN version: 7.4.2 Nvidia driver version: 440.95.01
• 模型信息    1）模型名称 cnn1 2）使用数据集名称
  RADIOML 2016.04C https://aistudio.baidu.com/aistudio/datasetdetail/80514
• 问题描述：使用SubClass 组网方式搭建cnn1，运行model_CNN1.summary((None, 1, 2, 128))时，dropout层处报错，看不懂原因请大佬指点。报错信息、代码关键片段如下： cnn1模型代码： 运行程序： 整个MR.py代码为：

  !/usr/bin/env python

  coding: utf-8

  In[1]:

  import paddle import paddle.vision.transforms as T from net import senet154, se_resnet50, se_resnet101, se_resnet152, se_resnext50_32x4d, se_resnext101_32x4d, senet103 from net_resnet import resnet18, resnet34, resnet50, resnet101, resnet152, resnext50_32x4d, resnext101_32x8d, wide_resnet50_2, wide_resnet101_2 from cnn1_cnn2 import cnn1, cnn2

  In[4]:

  import pickle import matplotlib.pyplot as plt import numpy as np

  加载数据集

  with open(r'/home/titan-306/workspace/JJY/R2016C-MR_compare/data/data80514/2016.04C.multisnr.pkl','rb') as xd1: Xd = pickle.load(xd1, encoding='latin1') snrs, mods = map(lambda j: sorted(list(set(map(lambda x: x[j], Xd.keys())))), [1, 0]) #set() 函数创建一个无序不重复元素集，可进行关系测试， 删除重复数据， 还可以计算交际、差集、并集等。

X = [] lbl = [] print('信噪比为:',snrs) print('调制模式为:',mods) for mod in mods: for snr in snrs: X.append(Xd[(mod, snr)]) #循环取出数值 for i in range(Xd[(mod, snr)].shape[0]): lbl.append((mod, snr)) X = np.vstack(X) #垂直（按照行顺序）的把数组给堆叠起来 print(X.shape) print(lbl[0:5])

对数据进行分区 数据集划分 60%/20%/20%

n_examples = X.shape[0] n_train = int(n_examples * 0.7) #训练集的数量 train_idx = np.random.choice(range(0, n_examples), size=n_train, replace=False)

测试集=验证集

test_idx = list(set(range(0,n_examples))-set(train_idx)) EsNoArray = [] for idx in test_idx: EsNoArray.append(lbl[idx][1]) EsNoArray = np.reshape(EsNoArray, len(EsNoArray)) print(EsNoArray)

X_train = X[train_idx] X_test = X[test_idx] row = 2 col = 128 X_train = np.asarray(list(map(lambda x: np.reshape(x, (row, col, 1)), X_train))) #将列表转换为数组 ??? 对列表可以进行reshape X_test = np.asarray(list(map(lambda x: np.reshape(x, (row, col, 1)), X_test)))

print('训练集的shape为：{}'.format(X_train.shape)) print('测试集的shape为：{}'.format(X_test.shape))

Y_train = list(map(lambda x: mods.index(lbl[x][0]), train_idx)) Y_train_snrs = list(map(lambda x: snrs.index(lbl[x][1]), train_idx))

Y_test = list(map(lambda x: mods.index(lbl[x][0]), test_idx)) Y_test_snrs = list(map(lambda x: snrs.index(lbl[x][1]), test_idx))

in_shp = list(X_train.shape[1:])

print(Y_train[0:10], Y_train_snrs[0:10])

数据集自定义

In[5]:

数据集自定义

import paddle.vision.transforms as T

class MR_dataset(paddle.io.Dataset): """ 步骤一：继承paddle.io.Dataset类 """ def init(self, mode: str ='train'): """ 步骤二：实现构造函数，定义数据读取方式 """ super(MR_dataset, self).init()

    self.mode = mode
    self.transforms = T.ToTensor()

    if self.mode  == 'train':
        self.data = X_train
        self.label = Y_train

    elif self.mode == 'test':
        self.data = X_test
        self.label = Y_test

    # self.transforms = T.ToTensor()   # 数据的格式转换和标准化 HWC => CHW  

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], dtype='float32')
    data = self.transforms(data)

    label = np.array(self.label[index], dtype='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)

实例化数据集

In[6]:

train_dataset = MR_dataset(mode='train') print(train_dataset[0][0].shape) print(train_dataset[0][1], train_dataset[0][1].shape, len(train_dataset)) print('**') test_dataset = MR_dataset(mode='test') print(test_dataset[0][0].shape) print(test_dataset[0][1], len(test_dataset))

In[7]:

model_CNN1 = cnn1()

In[8]:

model_senet154

model_CNN1 = paddle.Model(model_CNN1) model_CNN1.summary((None, 1, 2, 128))

mods = ['8PSK', 'AM-DSB', 'AM-SSB', 'BPSK', 'CPFSK', 'GFSK', 'PAM4', 'QAM16', 'QAM64', 'QPSK', 'WBFM']

snrs = [-20, -18, -16, -14, -12, -10, -8, -6, -4, -2, 0, 2, 4, 6, 8, 10, 12, 14, 16, 18]

[paddle-bot-old[bot]]

您好，我们已经收到了您的问题，会安排技术人员尽快解答您的问题，请耐心等待。请您再次检查是否提供了清晰的问题描述、复现代码、环境&版本、报错信息等。同时，您也可以通过查看官网API文档、常见问题、历史Issue、AI社区来寻求解答。祝您生活愉快～

Hi! We've received your issue and please be patient to get responded. We will arrange technicians to answer your questions as soon as possible. Please make sure that you have posted enough message to demo your request. You may also check out the API，FAQ，Github Issue and AI community to get the answer.Have a nice day!

[frankwhzhang]

请问这是自己的代码还是官方提供的

[JJYjackson]

请问这是自己的代码还是官方提供的

您好这是我自己的代码，希望大佬能帮我看看是什么问题？