def get_flops(a):
session = tf.compat.v1.Session()
graph = tf.compat.v1.get_default_graph()
with graph.as_default():
with session.as_default():
model = mobilenet_block.MobileNet(
input_shape=None,
alpha=1.0,
depth_multiplier=1,
dropout=0.001,
include_top=True,
weights=None,
input_tensor=tf.compat.v1.placeholder('float32',
shape=(1, 224, 224, 3)),
pooling=None,
classes=1000,
classifier_activation='softmax',
#i_64=1, i_128=1, i_256=1, i_512=1, i_1024=1
i_64=a[0],
i_128=a[1],
i_256=a[2],
i_512=a[3],
i_1024=a[4])
run_meta = tf.compat.v1.RunMetadata()
opts = tf.compat.v1.profiler.ProfileOptionBuilder.float_operation()
flops = tf.compat.v1.profiler.profile(graph=graph,
run_meta=run_meta,
cmd='op',
options=opts)
tf.compat.v1.reset_default_graph()
return flops.total_float_ops
b = ["i64", "i128", "i256", "i512", "i1024"]
FLOPS = []
for i in range(1, 11):
for j in range(5):
a = [0, 0, 0, 0, 0]
a[j] = i
model = mobilenet_block.MobileNet(input_shape=None,
alpha=1.0,
depth_multiplier=1,
dropout=0.001,
include_top=True,
weights=None,
input_tensor=None,
pooling=None,
classes=1000,
classifier_activation='softmax',
i_64=a[0],
i_128=a[1],
i_256=a[2],
i_512=a[3],
i_1024=a[4])
model.compile(optimizer='adam',
loss='mean_squared_error',
metrics=['accuracy'])
history = model.fit(x_train,
y_train,
epochs=TRAIN_EPOCHS,