Pytorch：如果是张量，则从张量中提取值

Question

I have a variable initialized at eg var = 0, the algorithm changes this variable to a tensor type Int if some conditions are met, otherwise it will stay at zero. I want to just print the variable without the tensor wrapping, however, if I use:

var = 0
if random.uniform(0, 1) < 0.5:
   var = torch.IntTensor(1)
print (f' var: {var.item()}')

It will throw an error that 'var' doesn't have function item() when the if condition is > 0.5 as it's just a zero. Is there a way to print the 'var' variable without any checking conditions?

I'm looking to do this without an else statement and with the print function outside the condition.

Does a PyTorch function, similar to item(), that can automatically detect whether its a tensor or not and deal with the condition exist?

Answer 1

A very obvious solution would be to use try

as in

var = 0
if random.uniform(0, 1) < 0.5:
   try:
       var = torch.IntTensor(1)
   except:
       pass # Or Throw and exception or whatever your use case migth be

print (f' var: {var.item()}')

This will try to run the IntTensor function, which if it gives an exception, it will handle.

You can also get the type of a variable by running type( variable ) which in case of 0 will return <class 'int'>, so you can add that to the if as well.

Python has a builtin method to get the type of a variable type()

type( 0 ) == int

Look into Try-catch And Get-type