46 lines
1.6 KiB
Python
46 lines
1.6 KiB
Python
from keras_core import backend
|
|
from keras_core import operations as ops
|
|
from keras_core import testing
|
|
from keras_core.optimizers.adam import Adam
|
|
import numpy as np
|
|
|
|
class AdamTest(testing.TestCase):
|
|
def test_config(self):
|
|
# TODO
|
|
pass
|
|
|
|
def test_weight_decay(self):
|
|
grads, var1, var2, var3 = (
|
|
np.zeros(()),
|
|
backend.Variable(2.0),
|
|
backend.Variable(2.0, name="exclude"),
|
|
backend.Variable(2.0),
|
|
)
|
|
optimizer_1 = Adam(learning_rate=1, weight_decay=0.004)
|
|
optimizer_1.apply_gradients(zip([grads], [var1]))
|
|
|
|
optimizer_2 = Adam(learning_rate=1, weight_decay=0.004)
|
|
optimizer_2.exclude_from_weight_decay(var_names=["exclude"])
|
|
optimizer_2.apply_gradients(zip([grads, grads], [var1, var2]))
|
|
|
|
optimizer_3 = Adam(learning_rate=1, weight_decay=0.004)
|
|
optimizer_3.exclude_from_weight_decay(var_list=[var3])
|
|
optimizer_3.apply_gradients(zip([grads, grads], [var1, var3]))
|
|
|
|
self.assertAlmostEqual(var1, 1.9760959)
|
|
self.assertAlmostEqual(var2, 2.0)
|
|
self.assertAlmostEqual(var3, 2.0)
|
|
|
|
def test_clip_norm(self):
|
|
optimizer = Adam(clipnorm=1)
|
|
grad = [np.array([100.0, 100.0])]
|
|
clipped_grad = optimizer._clip_gradients(grad)
|
|
self.assertAllClose(clipped_grad[0], [2**0.5 / 2, 2**0.5 / 2])
|
|
|
|
def test_clip_value(self):
|
|
optimizer = Adam(clipvalue=1)
|
|
grad = [np.array([100.0, 100.0])]
|
|
clipped_grad = optimizer._clip_gradients(grad)
|
|
self.assertAllClose(clipped_grad[0], [1.0, 1.0])
|
|
|