def test_Malloc2(self):
model = keras.models.Sequential()
model.add(
keras.layers.Conv2D(8, (3, 3),
padding='same',
input_shape=(32, 32, 3)))
model.add(keras.layers.Activation('tanh'))
model.add(keras.layers.Conv2D(8, (3, 3)))
model.add(keras.layers.Activation('relu'))
model.add(keras.layers.MaxPooling2D(pool_size=(2, 2)))
model.add(keras.layers.Dropout(0.25))
model.add(keras.layers.Conv2D(8, (3, 3), padding='same'))
model.add(keras.layers.Activation('relu'))
model.add(keras.layers.Conv2D(8, (3, 3)))
model.add(keras.layers.Activation('relu'))
model.add(keras.layers.MaxPooling2D(pool_size=(2, 2)))
model.add(keras.layers.Dropout(0.25))
model.add(keras.layers.Flatten())
model.add(keras.layers.Dense(20))
model.add(keras.layers.Activation('relu'))
model.add(keras.layers.Dropout(0.5))
model.add(keras.layers.Dense(10))
model.add(keras.layers.Activation('softmax'))
name = 'test___Malloc2' + str(int(time.time()))
keras2c_main.k2c(model, name, malloc=True)
rcode = build_and_run(name)
self.assertEqual(rcode, 0)
def test_UpSampling3D(self):
inshp = (4, 5, 10, 3)
a = keras.layers.Input(inshp)
b = keras.layers.UpSampling3D((3, 4, 2))(a)
model = keras.models.Model(a, b)
name = 'test___UpSampling3D' + str(int(time.time()))
keras2c_main.k2c(model, name)
rcode = build_and_run(name)
self.assertEqual(rcode, 0)
def test_GlobalMaxPooling3D(self):
inshp = (31, 21, 6, 7)
a = keras.layers.Input(inshp)
b = keras.layers.GlobalMaxPooling3D()(a)
model = keras.models.Model(inputs=a, outputs=b)
name = 'test___GlobalMaxPooling3D' + str(int(time.time()))
keras2c_main.k2c(model, name)
rcode = build_and_run(name)
self.assertEqual(rcode, 0)
def test_PReLU(self):
inshp = (3, 6, 9, 3)
a = keras.layers.Input(inshp)
b = keras.layers.PReLU(alpha_initializer='glorot_uniform')(a)
model = keras.models.Model(inputs=a, outputs=b)
name = 'test___PReLU' + str(int(time.time()))
keras2c_main.k2c(model, name)
rcode = build_and_run(name)
self.assertEqual(rcode, 0)
def test_ThresholdedReLU(self):
inshp = (3, 6, 19, 11)
theta = 0.3
a = keras.layers.Input(inshp)
b = keras.layers.ThresholdedReLU(theta=theta)(a)
model = keras.models.Model(inputs=a, outputs=b)
name = 'test___ThresholdedReLU' + str(int(time.time()))
keras2c_main.k2c(model, name)
rcode = build_and_run(name)
self.assertEqual(rcode, 0)
def test_Dense1(self):
inshp = (21, 4, 9)
units = 45
a = keras.layers.Input(inshp)
b = keras.layers.Dense(units, activation='relu')(a)
model = keras.models.Model(inputs=a, outputs=b)
name = 'test___Dense1' + str(int(time.time()))
keras2c_main.k2c(model, name)
rcode = build_and_run(name)
self.assertEqual(rcode, 0)
def test_TimeDistributed3(self):
model = keras.models.Sequential()
model.add(keras.layers.Dense(10, input_shape=(8, 5)))
model.add(keras.layers.TimeDistributed(
keras.layers.Dense(7, use_bias=True)))
model.build()
name = 'test___TimeDistributed3' + str(int(time.time()))
keras2c_main.k2c(model, name)
rcode = build_and_run(name)
self.assertEqual(rcode, 0)
def test_Permute(self):
inshp = (6, 12, 9)
a = keras.layers.Input(inshp)
b = keras.layers.Permute((3, 1, 2))(a)
# c = keras.layers.Dense(20)(b)
model = keras.models.Model(inputs=a, outputs=b)
name = 'test___permute' + str(int(time.time()))
keras2c_main.k2c(model, name)
rcode = build_and_run(name)
self.assertEqual(rcode, 0)
def test_ELU(self):
inshp = (13, 6, 9, 13)
alpha = 1.3
a = keras.layers.Input(inshp)
b = keras.layers.ELU(alpha=alpha)(a)
model = keras.models.Model(inputs=a, outputs=b)
name = 'test___ELU' + str(int(time.time()))
keras2c_main.k2c(model, name)
rcode = build_and_run(name)
self.assertEqual(rcode, 0)
def test_Dense2_Activation(self):
inshp = (40, 30)
units = 500
a = keras.layers.Input(inshp)
b = keras.layers.Dense(units, activation='tanh', use_bias=False)(a)
c = keras.layers.Activation('exponential')(b)
model = keras.models.Model(inputs=a, outputs=c)
name = 'test___Dense2' + str(int(time.time()))
keras2c_main.k2c(model, name)
rcode = build_and_run(name)
self.assertEqual(rcode, 0)
def test_repeat_vector(self):
inshp = (13,)
a = keras.layers.Input(inshp)
b = keras.layers.RepeatVector(23)(a)
c = keras.layers.ActivityRegularization(l1=.5, l2=.3)(b)
d = keras.layers.Dense(20)(c)
model = keras.models.Model(inputs=a, outputs=d)
name = 'test___repeat_vector' + str(int(time.time()))
keras2c_main.k2c(model, name)
rcode = build_and_run(name)
self.assertEqual(rcode, 0)
def test_TimeDistributed1(self):
inputs = keras.layers.Input(shape=(8, 5))
outputs = keras.layers.TimeDistributed(
keras.layers.Dense(7, use_bias=True,))(inputs)
model = keras.models.Model(inputs, outputs)
model.build((None, 8, 5))
name = 'test___TimeDistributed1' + str(int(time.time()))
keras2c_main.k2c(model, name)
rcode = build_and_run(name)
self.assertEqual(rcode, 0)
def test_TimeDistributed6(self):
model = keras.models.Sequential()
model.add(keras.layers.Dense(10, input_shape=(4, 7, 8, 5)))
model.add(keras.layers.TimeDistributed(
keras.layers.Conv2D(7, kernel_size=2)))
model.add(keras.layers.Dense(10))
model.build()
name = 'test___TimeDistributed6' + str(int(time.time()))
keras2c_main.k2c(model, name)
rcode = build_and_run(name)
self.assertEqual(rcode, 0)
def test_Dropout_Reshape_Flatten(self):
inshp = (10, 40, 30)
a = keras.layers.Input(inshp)
b = keras.layers.Flatten()(a)
c = keras.layers.Dropout(.4)(b)
d = keras.layers.Reshape((20, 30, 20))(c)
model = keras.models.Model(inputs=a, outputs=d)
name = 'test___flatten_dropout_reshape' + str(int(time.time()))
keras2c_main.k2c(model, name)
rcode = build_and_run(name)
self.assertEqual(rcode, 0)
def test_Subtract1(self):
inshp1 = (10, 8, 2, 3)
inshp2 = (10, 8, 2, 3)
a = keras.layers.Input(inshp1)
b = keras.layers.Input(inshp2)
c = keras.layers.Subtract()([a, b])
model = keras.models.Model(inputs=[a, b], outputs=c)
name = 'test___Subtract1' + str(int(time.time()))
keras2c_main.k2c(model, name)
rcode = build_and_run(name)
self.assertEqual(rcode, 0)
def test_ZeroPad1D(self):
inshp = (10, 12)
pad_top = 3
pad_bottom = 1
a = keras.layers.Input(inshp)
b = keras.layers.ZeroPadding1D(padding=(pad_top, pad_bottom))(a)
model = keras.models.Model(inputs=a, outputs=b)
name = 'test___ZeroPad1D' + str(int(time.time()))
keras2c_main.k2c(model, name)
rcode = build_and_run(name)
self.assertEqual(rcode, 0)
def test_Cropping1D(self):
inshp = (10, 12)
crop_top = 3
crop_bottom = 1
a = keras.layers.Input(inshp)
b = keras.layers.Cropping1D(cropping=(crop_top, crop_bottom))(a)
model = keras.models.Model(inputs=a, outputs=b)
name = 'test___Cropping1D' + str(int(time.time()))
keras2c_main.k2c(model, name)
rcode = build_and_run(name)
self.assertEqual(rcode, 0)
def test_Bidirectional1(self):
model = keras.models.Sequential()
model.add(keras.layers.Bidirectional(keras.layers.LSTM(10, return_sequences=True),
input_shape=(5, 10), merge_mode='concat'))
model.add(keras.layers.Bidirectional(keras.layers.LSTM(10, return_sequences=True),
merge_mode='mul'))
model.add(keras.layers.Dense(5))
model.build()
name = 'test___Bidirectional1' + str(int(time.time()))
keras2c_main.k2c(model, name)
rcode = build_and_run(name)
self.assertEqual(rcode, 0)
def test_Dot2(self):
dotaxes = (2, 2)
inshape1 = (5, 9)
inshape2 = (9, 9)
i1 = keras.layers.Input(inshape1)
i2 = keras.layers.Input(inshape2)
d = keras.layers.Dot(axes=dotaxes, normalize=True)([i1, i2])
model = keras.models.Model(inputs=[i1, i2], outputs=d)
name = 'test___Dot2' + str(int(time.time()))
keras2c_main.k2c(model, name)
rcode = build_and_run(name)
self.assertEqual(rcode, 0)
def test_ProfilePredictor(self):
inshp = (8, 32)
inp = keras.layers.Input(inshp)
a = keras.layers.Dense(20, activation='relu')(inp)
a = keras.layers.Dense(20, activation='relu')(a)
a = keras.layers.LSTM(20, activation='relu')(a)
outp = keras.layers.Dense(30)(a)
model = keras.models.Model(inp, outp)
name = 'test___ProfilePredictor' + str(int(time.time()))
keras2c_main.k2c(model, name)
rcode = build_and_run(name)
self.assertEqual(rcode, 0)
def test_Embedding1(self):
inshp = (10, 20)
input_dim = 20
output_dim = 30
a = keras.layers.Input(inshp)
b = keras.layers.Activation('relu')(a)
c = keras.layers.Embedding(
input_dim=input_dim, output_dim=output_dim)(b)
model = keras.models.Model(inputs=a, outputs=c)
name = 'test___Embedding1' + str(int(time.time()))
keras2c_main.k2c(model, name)
rcode = build_and_run(name)
self.assertEqual(rcode, 0)
def test_LSTM1(self):
inshp = (23, 32)
units = 19
a = keras.layers.Input(inshp)
b = keras.layers.LSTM(units,
activation='relu',
return_sequences=False,
recurrent_activation='hard_sigmoid')(a)
model = keras.models.Model(inputs=a, outputs=b)
name = 'test___LSTM1' + str(int(time.time()))
keras2c_main.k2c(model, name)
rcode = build_and_run(name)
self.assertEqual(rcode, 0)
def test_SimpleRNN2(self):
inshp = (34, 17)
units = 40
a = keras.layers.Input(inshp)
b = keras.layers.SimpleRNN(units,
go_backwards=True,
return_sequences=True,
activation='tanh')(a)
model = keras.models.Model(inputs=a, outputs=b)
name = 'test___SimpleRNN2' + str(int(time.time()))
keras2c_main.k2c(model, name)
rcode = build_and_run(name)
self.assertEqual(rcode, 0)
def test_GRU1(self):
inshp = (12, 46)
units = 17
a = keras.layers.Input(inshp)
b = keras.layers.GRU(units,
activation='softmax',
recurrent_activation='softsign',
return_sequences=False)(a)
model = keras.models.Model(inputs=a, outputs=b)
name = 'test___GRU1' + str(int(time.time()))
keras2c_main.k2c(model, name)
rcode = build_and_run(name)
self.assertEqual(rcode, 0)
def test_SimpleRNN1(self):
inshp = (4, 4, 46)
units = 17
a = keras.layers.Input(batch_shape=inshp)
b = keras.layers.SimpleRNN(units,
activation='relu',
return_sequences=False,
stateful=True)(a)
model = keras.models.Model(inputs=a, outputs=b)
name = 'test___SimpleRNN1' + str(int(time.time()))
keras2c_main.k2c(model, name)
rcode = build_and_run(name)
self.assertEqual(rcode, 0)
def main(args=sys.argv[1:]):
args = parse_args(args)
if args.malloc:
malloc = True
else:
malloc = False
if args.num_tests:
num_tests = args.num_tests
else:
num_tests = 10
k2c(args.model_path, args.function_name, malloc, num_tests)
def test_Add1(self):
inshp1 = (10, 8, 12)
inshp2 = (10, 8, 12)
inshp3 = (10, 8, 12)
a = keras.layers.Input(inshp1)
b = keras.layers.Input(inshp2)
c = keras.layers.Input(inshp3)
d = keras.layers.Add()([a, b, c])
model = keras.models.Model(inputs=[a, b, c], outputs=d)
name = 'test___Add1' + str(int(time.time()))
keras2c_main.k2c(model, name)
rcode = build_and_run(name)
self.assertEqual(rcode, 0)
def test_Concatenate3(self):
inshp1 = (2, 3, 6, 3)
inshp2 = (3, 3, 6, 3)
inshp3 = (1, 3, 6, 3)
axis = 1
a = keras.layers.Input(inshp1)
b = keras.layers.Input(inshp2)
c = keras.layers.Input(inshp3)
d = keras.layers.Concatenate(axis=axis)([a, b, c])
model = keras.models.Model([a, b, c], d)
name = 'test___Concatenate3' + str(int(time.time()))
keras2c_main.k2c(model, name)
rcode = build_and_run(name)
self.assertEqual(rcode, 0)
def test_LSTM2(self):
inshp = (4, 80)
units = 23
a = keras.layers.Input(inshp)
b = keras.layers.LSTM(units,
go_backwards=True,
return_sequences=True,
activation='sigmoid',
recurrent_activation='tanh')(a)
model = keras.models.Model(inputs=a, outputs=b)
name = 'test___LSTM2' + str(int(time.time()))
keras2c_main.k2c(model, name)
rcode = build_and_run(name)
self.assertEqual(rcode, 0)
def test_ReLU(self):
inshp = (12, 7, 9, 21)
max_value = 1.0
negative_slope = 1.0
threshold = 0.3
a = keras.layers.Input(inshp)
b = keras.layers.ReLU(max_value=max_value,
negative_slope=negative_slope,
threshold=threshold)(a)
model = keras.models.Model(inputs=a, outputs=b)
name = 'test___ReLU' + str(int(time.time()))
keras2c_main.k2c(model, name)
rcode = build_and_run(name)
self.assertEqual(rcode, 0)