def gen_data(begin, begin_mask, dtype, ellipsis_mask, end, end_mask,
new_axis_mask, shape, shrink_axis_mask, strides):
""" Generate data for testing the op """
input = np.random.uniform(low=-1.0, high=1.0,
size=tuple(shape)).astype(dtype)
# get numpy result
slices = strided_slice.args_to_slices(begin, end, strides, begin_mask,
end_mask, ellipsis_mask,
new_axis_mask, shrink_axis_mask)
expect = input[tuple(slices)]
out_shape = expect.shape if expect.shape != (0, ) else (1, )
output = np.full(out_shape, np.nan, dtype)
return expect, input, output
def strided_slice_python(input_shape,
begin,
end,
strides,
begin_mask,
end_mask,
ellipsis_mask,
new_axis_mask,
shrink_axis_mask,
grad,
dtype=np.float16):
slices = strided_slice.args_to_slices(begin, end, strides, begin_mask,
end_mask, ellipsis_mask,
new_axis_mask, shrink_axis_mask)
dx = np.zeros(input_shape).astype(dtype)
dx[tuple(slices)] = grad
return dx
def check_grad_shape(input_shape,
begin,
end,
strides,
begin_mask,
end_mask,
ellipsis_mask,
new_axis_mask,
shrink_axis_mask,
grad_shape_given,
dtype=np.float16):
slices = strided_slice.args_to_slices(begin, end, strides, begin_mask,
end_mask, ellipsis_mask,
new_axis_mask, shrink_axis_mask)
dx = np.zeros(input_shape)
grad_shape = dx[tuple(slices)].shape
assert list(grad_shape) == list(grad_shape_given), \
("parameters invalid: grad shape should be ", list(grad_shape),
"but given is", list(grad_shape_given))
return slices