assert_allclose
Description
Checks the tolerances between the benchmark tensor value and the expected value.
Prototype
def assert_allclose(tensor, desired, tol=[1e-7, 1e-7]):
Parameters
|
Parameter |
Type |
Description |
|---|---|---|
|
tensor |
Tensor |
Tensor to be compared. |
|
desired |
numpy.ndarray |
Expected data. |
|
tol |
List[float, float] |
tol=[rtol, atol] That is, a list of relative tolerances (rtol) and absolute tolerances (atol). Default value: [1e-7, 1e-7] |
Returns
None
Restrictions
The assert_allclose API supports only the scenario where the data type to be compared is Tensor and the expected data type is numpy.ndarray.
If the data type to be compared is numpy.ndarray, call the tvm.testing.assert_allclose API.
Parent topic: TBE DSL Testing API