Tensor.half()

Changes the dtype of the tensor to float16.

Usage

from tinygrad.tensor import Tensor
 
a1 = Tensor([[1, 2], [3, 4]])
a1 = a1.half()
 
print(a1)

Return value

<Tensor <LB (2, 2) dtypes.half op=UnaryOps.CAST st=ShapeTracker(shape=(2, 2), views=[View(shape=(2, 2), strides=(2, 1), offset=0, mask=None, contiguous=True, shape_strides=((4, 1),))], var_vals={})> on GPU with grad None>

glorot_uniform hardswish

;