A tensor is a multidimensional array that is used to store the data. So for using a Tensor, we have to import the torch module.
To create a tensor, the method used is tensor()”
Syntax:
Where data is a multi-dimensional array.
tensor.view()
view() in PyTorch is used to change the tensor object view by converting it into a specified number of rows and columns.
Syntax:
It takes two parameters.
- r specifies the number of rows to be formed from the tensor_object.
- c specifies the number of columns to be formed from the tensor_object.
Be sure that the actual tensor object contains an even count of elements.
Example 1
Here, we will create a tensor that holds six elements with Float type and change its view that has 3 rows and 2 columns.
import torch
#create 1D tensor with Float data type that hold 6 elements
data1 = torch.FloatTensor([23,45,54,32,23,78])
#display
print("Actual Tensor: ",data1)
#change the data1 view to 3 rows and 2 columns.
print("Tensor with 3 rows and 2 columns: ",data1.view(3,2))
Output:
Tensor with 3 rows and 2 columns: tensor([[23., 45.],
[54., 32.],
[23., 78.]])
We can see that the view of the tensor is changed to 3 rows and 2 columns.
Example 2
Here, we will create a tensor that holds six elements with Float type and change its view that has 2 rows and 3 columns.
import torch
#create 1D tensor with Float data type that hold 6 elements
data1 = torch.FloatTensor([23,45,54,32,23,78])
#display
print("Actual Tensor: ",data1)
#change the data1 view to 2 rows and 3 columns.
print("Tensor with 2 rows and 3 columns: ",data1.view(2,3))
Output:
Tensor with 2 rows and 3 columns: tensor([[23., 45., 54.],
[32., 23., 78.]])
We can see that the view of the tensor is changed to 2 rows and 3 columns.
Change the datatype
It can be possible to change the datatype of the tensor using view().
We need to specify the datatype inside the view method.
Syntax:
Parameter:
It takes datatype as a parameter like int8,int16, etc.
Example 1
In this example, we will create a tensor with Float type and convert it to int data types.
dtype is used to return the datatype of a tensor.
import torch
#create 1D tensor with Float data type that hold 6 elements
data1 = torch.FloatTensor([23,45,54,32,23,78])
#display
print("Actual Tensor data type: ",data1.dtype)
#change the data1 data type to int8
print("Converting to int8: ",data1.view(torch.int8).dtype)
#change the data1 data type to int16
print("Converting to int16: ",data1.view(torch.int16).dtype)
#change the data1 data type to int32
print("Converting to int32: ",data1.view(torch.int32).dtype)
#change the data1 data type to int64
print("Converting to int64: ",data1.view(torch.int64).dtype)
Output:
Converting to int8: torch.int8
Converting to int16: torch.int16
Converting to int32: torch.int32
Converting to int64: torch.int64
Example 2
In this example, we will create a tensor with Float type and convert it to int data types and get the size.
import torch
#create 1D tensor with Float data type that hold 6 elements
data1 = torch.FloatTensor([23,45,54,32,23,78])
#display
print("Actual Tensor datatype: ",data1.size())
#change the data1 datatype to int8
print("Converting to int8: ",data1.view(torch.int8).size())
#change the data1 datatype to int16
print("Converting to int16: ",data1.view(torch.int16).size())
#change the data1 datatype to int32
print("Converting to int32: ",data1.view(torch.int32).size())
#change the data1 datatype to int64
print("Converting to int64: ",data1.view(torch.int64).size())
Output:
Converting to int8: torch.Size([24])
Converting to int16: torch.Size([12])
Converting to int32: torch.Size([6])
Converting to int64: torch.Size([3])
Conclusion
In this PyTorch lesson, we discussed how to change the view of a tensor in pytorch using view() and also modify the datatypes of an existing tensor by specifying data types inside the view() method.
