|
- # Copyright 2019 Huawei Technologies Co., Ltd
- #
- # Licensed under the Apache License, Version 2.0 (the "License");
- # you may not use this file except in compliance with the License.
- # You may obtain a copy of the License at
- #
- # http://www.apache.org/licenses/LICENSE-2.0
- #
- # Unless required by applicable law or agreed to in writing, software
- # distributed under the License is distributed on an "AS IS" BASIS,
- # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- # See the License for the specific language governing permissions and
- # limitations under the License.
-
- from akg.utils import kernel_exec as utils
- from tests.common.tensorio import compare_tensor
- import numpy as np
- from akg.ops.nn.mean_ad import mean_ad
- from tests.common.base import get_rtol_atol
-
-
- def mean_ad_run(shape, dtype, axis, keepdims, attrs):
- support_list = {"float16": np.float16, "float32": np.float32}
- if isinstance(axis, int):
- axis = (axis,)
- axis = list(axis)
- for i in range(len(axis)):
- if axis[i] < 0:
- axis[i] += len(shape)
-
- mean_num = 1
- dy_shape = []
- dy_shape_br = []
- for i, dim in enumerate(shape):
- if i in axis:
- if keepdims:
- dy_shape.append(1)
-
- mean_num *= dim
- dy_shape_br.append(1)
- else:
- dy_shape.append(dim)
- dy_shape_br.append(dim)
- if len(dy_shape) == 0:
- dy_shape.append(1)
- dy_shape = tuple(dy_shape)
-
- if 'tuning' in attrs.keys():
- t = attrs.get("tuning", False)
- kernel_name = attrs.get("kernel_name", False)
- mod = utils.op_build_test(mean_ad, [dy_shape], [dtype], kernel_name=kernel_name,
- op_attrs=[shape, axis, keepdims], attrs=attrs, tuning=t)
- if t:
- expect, head_np, output = gen_data(dtype, dy_shape, dy_shape_br, mean_num, shape, support_list)
- return mod, expect, (head_np, output)
- else:
- return mod
- else:
- mod = utils.op_build_test(mean_ad, [dy_shape], [dtype], kernel_name='mean_ad', op_attrs=[shape, axis, keepdims],
- attrs=attrs)
- # if(axis == None):
- # input_np = np.random.uniform(low=-1.0, high=1.0, size=shape).astype(support_list[dtype])
- # head_np = np.random.uniform(low=-1.0, high=1.0, size=(1)).astype(support_list[dtype])
- # expect = np.ones(shape)
- # expect = expect * 1.0 / (shape[0] * shape[1] * shape[2] * shape[3])
- # else:
- expect, head_np, output = gen_data(dtype, dy_shape, dy_shape_br, mean_num, shape, support_list)
- output = utils.mod_launch(mod, (head_np, output), expect=expect)
- rtol, atol = get_rtol_atol("mean_ad", dtype)
- TestCase_Result = compare_tensor(output, expect, rtol=rtol, atol=atol, equal_nan=True)
-
- return (head_np,), output, expect, TestCase_Result
-
-
- def gen_data(dtype, dy_shape, dy_shape_br, mean_num, shape, support_list):
- input_np = np.random.uniform(low=-1.0, high=1.0, size=shape).astype(support_list[dtype])
- head_np = np.random.uniform(low=-1.0, high=1.0, size=dy_shape).astype(support_list[dtype])
- expect = np.ones(shape)
- expect = expect * 1.0 / mean_num
- expect = expect * np.broadcast_to(np.reshape(head_np, dy_shape_br), shape)
- output = np.full(expect.shape, np.nan, dtype)
- return expect, head_np, output
|
