def testKLDivergenceLossGrad(self):
np.random.seed(1701)
layer = core_layers.KLDivergenceLossLayer(name='loss')
checker = gradcheck.GradChecker(1e-6)
shape = (4, 5)
# For the input, we make sure it is not too close to 0 (which would
# create numerical issues).
input_blob = base.Blob(shape,
filler=fillers.RandFiller(min=0.1, max=0.9))
# normalize input blob
input_data = input_blob.data()
input_data /= input_data.sum(1)[:, np.newaxis]
# check index input
target_blob = base.Blob(shape[:1],
dtype=np.int,
filler=fillers.RandIntFiller(high=shape[1]))
result = checker.check(layer, [input_blob, target_blob], [],
check_indices=[0])
print(result)
self.assertTrue(result[0])
# also, check if weight works.
self._testWeight(layer, [input_blob, target_blob])
# check full input
target_blob = base.Blob(shape, filler=fillers.RandFiller())
# normalize target
target_data = target_blob.data()
target_data /= target_data.sum(1)[:, np.newaxis]
result = checker.check(layer, [input_blob, target_blob], [],
check_indices=[0])
print(result)
self.assertTrue(result[0])
# also, check if weight works.
self._testWeight(layer, [input_blob, target_blob])
def testIdentityLayer(self):
layer = identity.IdentityLayer(name='identity')
np.random.seed(1701)
filler = fillers.RandFiller()
bottom = base.Blob((100, 4), filler=filler)
top = base.Blob()
# run the dropout layer
layer.forward([bottom], [top])
# simulate a diff
fillers.RandFiller().fill(top.init_diff())
layer.backward([bottom], [top], True)
np.testing.assert_array_equal(top.data(), bottom.data())
np.testing.assert_array_equal(top.diff(), bottom.diff())
def testMultinomialLogisticLossGrad(self):
np.random.seed(1701)
layer = core_layers.MultinomialLogisticLossLayer(name='loss')
checker = gradcheck.GradChecker(1e-6)
shape = (10, 5)
# check index input
input_blob = base.Blob(shape, filler=fillers.GaussianRandFiller())
target_blob = base.Blob(shape[:1],
dtype=np.int,
filler=fillers.RandIntFiller(high=shape[1]))
result = checker.check(layer, [input_blob, target_blob], [],
check_indices=[0])
print(result)
self.assertTrue(result[0])
# also, check if weight works.
self._testWeight(layer, [input_blob, target_blob])
# check full input
target_blob = base.Blob(shape, filler=fillers.RandFiller())
# normalize target
target_data = target_blob.data()
target_data /= target_data.sum(1)[:, np.newaxis]
result = checker.check(layer, [input_blob, target_blob], [],
check_indices=[0])
print(result)
self.assertTrue(result[0])
# also, check if weight works.
self._testWeight(layer, [input_blob, target_blob])
def testLocalResponseNormalizeLayer(self):
np.random.seed(1701)
output_blob = base.Blob()
checker = gradcheck.GradChecker(1e-6)
shapes = [(1,10), (5,10)]
alphas = [1.0, 2.0]
betas = [0.75, 1.0]
for shape in shapes:
for alpha in alphas:
for beta in betas:
input_blob = base.Blob(shape, filler=fillers.RandFiller())
# odd size
layer = core_layers.LocalResponseNormalizeLayer(
name='normalize', k = 1., alpha=alpha, beta=beta, size=5)
result = checker.check(layer, [input_blob], [output_blob])
print(result)
self.assertTrue(result[0])
layer = core_layers.LocalResponseNormalizeLayer(
name='normalize', k = 2., alpha=alpha, beta=beta, size=5)
result = checker.check(layer, [input_blob], [output_blob])
print(result)
self.assertTrue(result[0])
# even size
layer = core_layers.LocalResponseNormalizeLayer(
name='normalize', k = 1., alpha=alpha, beta=beta, size=6)
result = checker.check(layer, [input_blob], [output_blob])
print(result)
self.assertTrue(result[0])
def testResponseNormalizeLayer(self):
np.random.seed(1701)
output_blob = base.Blob()
checker = gradcheck.GradChecker(1e-5)
shapes = [(1,5,5,1), (1,5,5,3), (5,5), (1,5)]
for shape in shapes:
input_blob = base.Blob(shape,
filler=fillers.RandFiller(min=0.1, max=1.))
layer = core_layers.ResponseNormalizeLayer(
name='normalize')
result = checker.check(layer, [input_blob], [output_blob])
print(result)
self.assertTrue(result[0])
def testAutoencoderLossGrad(self):
np.random.seed(1701)
shapes = [(4, 3), (1, 10), (4, 3, 2)]
layer = core_layers.AutoencoderLossLayer(name='loss', ratio=0.5)
checker = gradcheck.GradChecker(1e-5)
for shape in shapes:
input_blob = base.Blob(shape,
filler=fillers.RandFiller(min=0.05,
max=0.95))
result = checker.check(layer, [input_blob], [])
print(result)
self.assertTrue(result[0])
# also, check if weight works.
self._testWeight(layer, [input_blob])
def testdropoutlayer(self):
layer = dropout.DropoutLayer(name='dropout', ratio=0.5)
np.random.seed(1701)
filler = fillers.RandFiller(min=1, max=2)
bottom = base.Blob((100, 4), filler=filler)
top = base.Blob()
# run the dropout layer
layer.forward([bottom], [top])
# simulate a diff
fillers.RandFiller().fill(top.init_diff())
layer.backward([bottom], [top], True)
np.testing.assert_array_equal(top.data()[top.data() != 0] * 0.5,
bottom.data()[top.data() != 0])
np.testing.assert_array_equal(bottom.diff()[top.data() == 0], 0)
np.testing.assert_array_equal(bottom.diff()[top.data() != 0],
top.diff()[top.data() != 0] * 2.)
# test if debug_freeze works
layer = dropout.DropoutLayer(name='dropout',
ratio=0.5,
debug_freeze=True)
layer.forward([bottom], [top])
snapshot = top.data().copy()
layer.forward([bottom], [top])
np.testing.assert_array_equal(snapshot, top.data())