An error occurred when executing torch.nn.functional.softmax(), "softmax_lastdim_kernel_impl" not implemented for 'Long'

This error usually occurs when the nn.functional.softmax() function is executed, and the input data type is an integer type (Long).

The softmax() function requires the input data type to be a floating-point number (Float) or a double-precision floating-point number (Double).

To resolve this error, the input data type needs to be converted to a float type. Type conversion can be done using the tensor.float() or tensor.double() functions. For example:

import torch.nn.functional as F
import torch

# 创建一个整数类型的tensor
x = torch.LongTensor([[1, 2, 3], [4, 5, 6]])

# 将数据类型转换为float类型
x = x.float()

# 执行softmax函数
y = F.softmax(x, dim=1)

print(y)

The output is:

tensor([[0.0900, 0.2447, 0.6652],
        [0.0900, 0.2447, 0.6652]])