round_half_up

Description

Rounds half up raw_tensor to the nearest integer element-wise, away from 0. For example, –1.5 is rounded to –2, and 1.5 is rounded to 2.

Prototype

round_half_up(raw_tensor)

Parameters

raw_tensor: a tvm.tensor for the input tensor

Atlas Training Series Product: supports float16 and float32.

Returns

res_tensor: result Tensor, tvm.tensor type

The default data type is int32.

Restrictions

None

Applicability

Atlas Training Series Product

Example

from tbe import tvm
from tbe import dsl
shape = (1024,1024)
input_dtype = "float16"
data = tvm.placeholder(shape, name="data", dtype=input_dtype)
res = dsl.round_half_up(data)