src/test/scripts/applications/nn/component/selu.dml

#-------------------------------------------------------------
#
# Licensed to the Apache Software Foundation (ASF) under one
# or more contributor license agreements.  See the NOTICE file
# distributed with this work for additional information
# regarding copyright ownership.  The ASF licenses this file
# to you 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.
#
#-------------------------------------------------------------

source("nn/layers/selu.dml") as selu
source("src/test/scripts/applications/nn/util.dml") as test_util

selu_test1 = function() {
    print("Testing SeLU, test 1")

    X = matrix("1.  -0.5
                0.   2.", rows=2, cols=2)
    dout = matrix("1 1
                   1 1", rows=2, cols=2)

    # Reference from PyTorch nn.SELU
    out_expected = matrix("1.050701   -0.69175816
                           0.         2.101402", rows=2, cols=2)
    gradient_expected = matrix("1.050701   1.0663412
                                1.7580993  1.050701", rows=2, cols=2)

    out = selu::forward(X)
    test_util::check_all_close(out, out_expected, 0.00001)

    gradient = selu::backward(dout, X)
    test_util::check_all_close(gradient, gradient_expected, 0.00001)
}

selu_test2 = function() {
    print("Testing SeLU, test 2")

    X = matrix("0.5  -1.5
                1.   -2.", rows=2, cols=2)
    dout = matrix("1 1
                   1 1", rows=2, cols=2)

    # Precomputed reference from PyTorch nn.SELU
    out_expected = matrix("0.5253505   -1.3658143
                           1.050701    -1.5201665", rows=2, cols=2)
    gradient_expected = matrix("1.050701   0.392285
                                1.050701   0.23793286", rows=2, cols=2)

    out = selu::forward(X)
    test_util::check_all_close(out, out_expected, 0.00001)

    gradient = selu::backward(dout, X)
    test_util::check_all_close(gradient, gradient_expected, 0.00001)
}

selu_test1()
selu_test2()