在用LSTM预测轴承剩余寿命时，得到如标题的错误：RuntimeError: Input and parameter tensors are not the same dtype, found input tensor with Double and parameter tensor with Float，可是我检查了输入都是float类型，还请大神解决一下。源代码如下：
#导入库
from numpy import *
import matplotlib.pyplot as plt
import scipy.io as scio
import matplotlib.pyplot as plt
import numpy as np
import torch
from torch import nn
from torch.autograd import Variable
from visdom import Visdom
#读取数据
#导入训练集数据,mat格式
dataFile1 ='D:/！研究生学习生涯/！剩余寿命预测/LSTM相关资料/LSTM预测股票代码/基于LSTM模型的股票预测模型/normal_rms_1-1.mat'
data1 = scio.loadmat(dataFile1) # 读入数据
feature1=data1['D'] # 读取mat里面的RMS数据
print(feature1)
print(len(feature1))
print(feature1.shape)
labels1=[]
#设置训练集标签
for i in range(0,len(feature1)):
if i<12740:
labels1.append(0)
else:
labels1.append(0+(i-12740)/(len(feature1)-12740))
#导入测试集数据
dataFile2 ='D:/！研究生学习生涯/！剩余寿命预测/LSTM相关资料/LSTM预测股票代码/基于LSTM模型的股票预测模型/normal_rms_1-2.mat'
data2 = scio.loadmat(dataFile2) # 读入数据
feature2=data2['D'] # 读取mat里面的RMS数据
print(feature2)
print(len(feature2))
print(feature2.shape)
labels2=[]
#设置测试集标签
for i in range(0,len(feature2)):
if i<9155:
labels2.append(0)
else:
labels2.append(0+(i-9155)/(len(feature2)-9155))
labels1=np.array(labels1)
labels2=np.array(labels2)
dataX=[]
dataY=[]
#数据预处理,用过去的20天数据去预测未来的1天数据。
def create_dataset1(dataset1, look_back=20):
for i in range(len(dataset1) - look_back):
a = dataset1[i:(i + look_back)]
dataX.append(a)
return np.array(dataX)
def create_dataset2(dataset2, look_back=20,pre_steps=1):
for i in range(len(dataset2) - look_back):
b=dataset2[i + look_back:i + look_back + pre_steps]
dataY.append(b)
return np.array(dataY)
train_x=create_dataset1(feature1)
# train_x=torch.tensor(train_x).to('cuda:1',torch.float32)
train_y=create_dataset2(labels1)
train_x=train_x.reshape(-1,20,1)
train_y=train_y.reshape(-1,1)
# train_x=train_x.reshape(-1,1,20)
# train_y=train_y.reshape(-1,1,1)
# train_y= train_y.astype(float)
# train_y=np.array(train_y)
test_x=create_dataset1(feature2)
test_x=test_x.reshape(-1,1,20)
test_y=create_dataset2(labels2)
# test_x=np.array(test_x)
# test_y=np.array(test_y)
#处理数据维度满足模型输入条件
train_x=torch.tensor(train_x).to('cuda',torch.float64).reshape(-1,20,1)
test_x=torch.tensor(test_x).to('cuda',torch.float64).reshape(-1,20,1)
# train_y=torch.tensor(train_y).to('cuda',torch.float32)
train_y=torch.from_numpy(train_y).to('cuda',torch.float64)
test_y=torch.tensor(test_y).to('cuda',torch.float64)
# test_y=torch.tensor(test_y).to('cuda')
#设置模型超参数
input_size=1
hidden_size=20
cidden_size=20
output_size=1
lr=0.01
num_layers=1
#定义模型
class lstm(nn.Module):
def __init__(self,input_size=1,hidden_size=20,num_layers=1):
super(lstm,self).__init__()
self.rnn=nn.LSTM(input_size,hidden_size,num_layers,batch_first=True)
self.fc = nn.Linear(hidden_size, output_size)
def forward(self,x):
# x = self.rnn(x) # x is input, size (seq_len, batch, input_size)
# s, b, h = x[0].shape # x is output, size (seq_len, batch, hidden_size)
# x = x[0].view(s * b, h)
# x = self.fc(x)
# x = x.view(s, b, -1) # 把形状改回来
x = self.rnn(x)[1][0].view(-1,1)
return x
net=lstm().to('cuda')
# input = torch.randn(20124, 20, 1).to('cuda')
# out = net(input)
#
#
criterion=nn.MSELoss()
optimizer=torch.optim.Adam(net.parameters(),lr)
print(net)
for ii in range(1000):
# train_x=torch.Tensor(train_x).to('cuda')
# train_y=torch.Tensor(train_y).to('cuda')
# train_x=train_x.to(torch.float32)
# train_y=train_y.to(torch.float32)
var_x=Variable(train_x)
var_y=Variable(train_y)
#前向传播
out=net(var_x)
loss=criterion(out,var_y).to('cuda')
#反向传播
optimizer.zero_grad()
loss.backward()
optimizer.step()
if (e+1)%100== 0:
print('Epoch: {}, Loss: {:.10f}'.format(e + 1, loss.data))
net=net.eval()#转换为测试模式
# test_x=torch.Tensor(test_x)
# test_x=test_x.to(torch.float32)
varx=Variable(test_x)
pre_test=net(varx)
pre_test = pre_test.view(-1).data.numpy()
pre_test = np.concatenate((np.zeros(), pre_test))
plt.plot(test_y,pre_test)
plt.show()
