|
- # 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.
-
- import numpy as np
- from tests.common.tensorio import compare_tensor
- from akg.utils import kernel_exec as utils
- from tests.common.test_op import focal_loss
- from tests.common.gen_random import random_gaussian
-
- def softmax(x):
- mv = np.max(x, axis=-1, keepdims=True)
- v = x - mv
- s = np.exp(v) / np.sum(np.exp(v), axis=-1, keepdims=True)
- return s
-
-
- def logsoftmax(x):
- mv = np.max(x, axis=-1, keepdims=True)
- v = x - mv
- exp_x = np.exp(v)
- Z = np.sum(exp_x, axis=-1, keepdims=True)
- return v - np.log(Z)
-
-
- def benchmark(x, y, gamma):
- y_pred = softmax(x)
- expect = -y * ((1 - y_pred) ** gamma) * logsoftmax(x)
- res = np.sum(expect, axis=-1)
- return res
-
-
- def focal_loss_run(shape, p_dtype, t_dtype, gamma, kernel_name, attrs):
- if 'tuning' in attrs.keys():
- t = attrs.get("tuning", False)
- kernel_name = attrs.get("kernel_name", False)
- mod = utils.op_build_test(focal_loss.focal_loss, [shape, shape], [p_dtype, t_dtype], op_attrs=[gamma],
- kernel_name=kernel_name, attrs=attrs, tuning=t)
- if t:
- expect, pred, targ = gen_data(attrs, gamma, p_dtype, shape, t_dtype)
- output = np.full(expect.shape, 0.0, p_dtype)
- return mod, expect, (pred, targ, output)
- else:
- return mod
- else:
- mod = utils.op_build_test(focal_loss.focal_loss, [shape, shape], [p_dtype, t_dtype], op_attrs=[gamma],
- kernel_name=kernel_name, attrs=attrs)
- expect, pred, targ = gen_data(attrs, gamma, p_dtype, shape, t_dtype)
- output = np.full(expect.shape, 0.0, p_dtype)
- output = utils.mod_launch(mod, (pred, targ, output), expect=expect)
-
- return (pred, targ), output, expect, compare_tensor(output, expect, rtol=5e-2, atol=1e-4)
-
-
- def gen_data(attrs, gamma, p_dtype, shape, t_dtype):
- # if mode=='rpc_cloud': # 4\6\7\12\13 fail in multi core
- if not utils.product_is_mini():
- # if (mode=='rpc_cloud' or mode=='aic_cloud') and p_dtype=='float16':
- attrs['enable_multicore'] = True
- pred = abs(random_gaussian(shape, miu=1, sigma=0.1).astype(p_dtype) * 10.0).astype(p_dtype)
- targ = (np.eye(shape[-1])[np.random.randint(0, shape[-1], size=shape[:-1])]).astype(t_dtype)
- expect = benchmark(pred, targ, gamma)
- return expect, pred, targ
|
