Solution: runtimeerror: expected object of scalar type int but got scalar type double

Problem description ：

Recently, I am writing code using neural network prediction , After replacing the imported data , The original loss function （ As shown below ） Report errors RuntimeError: Expected object of scalar type Int but got scalar type Double. Later according to CSDN Many blog posts on try to use many methods , for example ：loss Function is preceded by X=torch.tensor(np.int(X)) perhaps y=torch.IntTensor(y), But then again “ValueError: only one element tensors can be converted to Python scalar Waiting for a wrong report , So according to other blog posts, it is changed to X= torch.tensor([item.cpu().detach().numpy() for item in np.int(X)]).cuda(), So I went back to the above error report , It's a headache .

    for X,y in data_iter(batch_size,train_feature,train_label):# Internal circulation training one batch
        l=loss(X,y).sum()  # Calculate the gap between the output of the model and the real data 

Solution ：

In fact, solving this problem is very simple , Because the suffix of the imported table is xlsx（ namely ： natural Excel file ）, Just save the file as csv The format is just （ It should be noted that , The number of reads must be utf-8, Otherwise, when reading, it will be due to csv The encoding format in is ‘ANSI’, and Python 3 The default read format is ‘utf-8’, Lead to decode abnormal . Therefore, an error will be reported ’utf-8’ codec can’t decode bytes in position 16: invalid continuation byte）.
After adding the suffix from xlsx Change it to csv after , Read the file smoothly and train , There is no other error report in the middle （ヾ(◍°∇°◍)ﾉﾞ Kaisensen ~ Kaisensen ~）

Reason for modification ：

1、CSV It's a plain text file ,excel It's not plain text ,excel Contains a lot of format information in it . namely ：CSV It's a plain text file ,CSV The file in windows The default opening mode of the platform is excel, But its essence is a text file .excel It's not plain text ,Excel It's a binary file , It saves information about all worksheets in the workbook . therefore ,csv The document is more pure .
Other reasons ：
1、CSV The file size will be smaller , Creating a distribution is easier to read , Suitable for storing structured information , For example, export of records , Traffic statistics, etc .
2、 Because there is no clear distinction or distinction between numerical value and text ,Excel You can use the automatic formatting function to confuse the values .
CSV It's safer , Value and text can be clearly distinguished .CSV Do not process data and store as is .
3、Excel More memory is consumed when importing data . Import CSV Files can be faster , And consume less memory .
4、Excel Files can only be used Microsoft Excel documents opening .CSV It can be used Windows Open any text editor in , Like Notepad ,MS Excel,Microsoft Works 9 etc. .

reference ：

https://www.cnblogs.com/Dongmy/p/16434120.html
https://blog.csdn.net/qq_38703529/article/details/120216078
https://blog.csdn.net/qq_18649781/article/details/90771682
https://blog.csdn.net/qq_38703529/article/details/120216078
https://blog.csdn.net/yzf0011/article/details/93321576