PyTorch .bitwise_left_shift()

In PyTorch, the .bitwise_left_shift() function shifts each element of the input tensor to the left by a specified number of bits. Both operands must be of an integral type (e.g., int32, int64). Shifting by a negative number of bits returns 0.

Syntax

torch.bitwise_left_shift(input, other, *, out=None)

Parameters:

• input: The first input tensor; must be of an integral type.
• other: The second input (can be a tensor or integer).
• out (Optional): A tensor where the output will be stored.

Return value:

Returns a tensor of the same shape as the broadcasted inputs, with each element being the result of left-shifting the corresponding element in input by the amount in other.

Example

This example shows how to use .bitwise_left_shift() on a tensor:

import torch

# Create two tensors
a = torch.tensor([-1, 4, 3], dtype=torch.int8)
b = torch.tensor([1, 0, 3], dtype=torch.int8)

# Apply bitwise_left_shift on two tensors
c = torch.bitwise_left_shift(a,b)
print("c =",c)

# Apply bitwise_left_shift on a tensor and number
d = torch.bitwise_left_shift(a, 2)
print("d =",d)

Copy to clipboard
Copy to clipboard

The code produces this output:

c = tensor([-2,  4, 24], dtype=torch.int8)
d = tensor([-4, 16, 12], dtype=torch.int8)

Copy to clipboard
Copy to clipboard