Cast¶
Description¶
This module performs the function of changing the tensor data type.
Initializing¶
def __init__(self, intype, outtype, name=None):
Parameters
| Parameter | Allowed types | Description | Default |
|---|---|---|---|
| intype | dtype | Input data type | - |
| outtype | dtype | Output data type | None |
| name | str | Layer name | None |
Explanations
-
Examples¶
Necessary imports.
import numpy as np
from PuzzleLib.Backend import gpuarray
from PuzzleLib.Modules import Cast
Info
gpuarray is required to properly place the tensor in the GPU.
batchsize, size, maps = 2, 3, 2
data1 = gpuarray.to_gpu(np.arange(batchsize * size * maps).reshape((batchsize, size, maps)).astype(np.float32))
data2 = gpuarray.to_gpu(np.arange(batchsize * size * maps).reshape((batchsize, size, maps)).astype(np.float16))
print(data1.dtype)
dtype('float32')
print(data2.dtype)
dtype('float16')
cast = Cast(data1.dtype, data2.dtype)
data1 = cast(data1)
print(data1.dtype)
dtype('float16')