def _get_net(W_std, b_std, filter_shape, is_conv, use_pooling, is_res, padding,
phi, strides, width, is_ntk, proj_into_2d, layer_norm,
parameterization, use_dropout):
fc = partial(stax.Dense,
W_std=W_std,
b_std=b_std,
parameterization=parameterization)
conv = partial(stax.Conv,
filter_shape=filter_shape,
strides=strides,
padding=padding,
W_std=W_std,
b_std=b_std,
parameterization=parameterization)
affine = conv(width) if is_conv else fc(width)
rate = np.onp.random.uniform(0.5, 0.9)
dropout = stax.Dropout(rate, mode='train')
ave_pool = stax.AvgPool((2, 3), None,
'SAME' if padding == 'SAME' else 'CIRCULAR')
ave_pool_or_identity = ave_pool if use_pooling else stax.Identity()
dropout_or_identity = dropout if use_dropout else stax.Identity()
layer_norm_or_identity = (stax.Identity() if layer_norm is None else
stax.LayerNorm(axis=layer_norm))
res_unit = stax.serial(ave_pool_or_identity, phi, dropout_or_identity,
affine)
if is_res:
block = stax.serial(affine, stax.FanOut(2),
stax.parallel(stax.Identity(), res_unit),
stax.FanInSum(), layer_norm_or_identity)
else:
block = stax.serial(affine, res_unit, layer_norm_or_identity)
if proj_into_2d == 'FLAT':
proj_layer = stax.Flatten()
elif proj_into_2d == 'POOL':
proj_layer = stax.GlobalAvgPool()
elif proj_into_2d.startswith('ATTN'):
n_heads = int(np.sqrt(width))
n_chan_val = int(np.round(float(width) / n_heads))
fixed = proj_into_2d == 'ATTN_FIXED'
proj_layer = stax.serial(
stax.GlobalSelfAttention(width,
n_chan_key=width,
n_chan_val=n_chan_val,
n_heads=n_heads,
fixed=fixed,
W_key_std=W_std,
W_value_std=W_std,
W_query_std=W_std,
W_out_std=1.0,
b_std=b_std), stax.Flatten())
else:
raise ValueError(proj_into_2d)
readout = stax.serial(proj_layer, fc(1 if is_ntk else width))
return stax.serial(block, readout)
def WideResnetnt(
block_size,
k,
num_classes,
batchnorm='std'): #, batch_norm=None,layer_norm=None,freezelast=None):
"""Based off of WideResnet from paper, with or without BatchNorm.
(Set config.wrn_block_size=3, config.wrn_widening_f=10 in that case).
Uses default weight and bias init."""
parameterization = 'standard'
layers_lst = [
stax_nt.Conv(16, (3, 3),
padding='SAME',
parameterization=parameterization),
WideResnetGroupnt(block_size,
16 * k,
parameterization=parameterization,
batchnorm=batchnorm),
WideResnetGroupnt(block_size,
32 * k, (2, 2),
parameterization=parameterization,
batchnorm=batchnorm),
WideResnetGroupnt(block_size,
64 * k, (2, 2),
parameterization=parameterization,
batchnorm=batchnorm)
]
layers_lst += [_batch_norm_internal(batchnorm), stax_nt.Relu()]
layers_lst += [
stax_nt.AvgPool((8, 8)),
stax_nt.Flatten(),
stax_nt.Dense(num_classes, parameterization=parameterization)
]
return stax_nt.serial(*layers_lst)
def testPredictOnCPU(self):
x_train = random.normal(random.PRNGKey(1), (10, 4, 5, 3))
x_test = random.normal(random.PRNGKey(1), (8, 4, 5, 3))
y_train = random.uniform(random.PRNGKey(1), (10, 7))
_, _, kernel_fn = stax.serial(stax.Conv(1, (3, 3)), stax.Relu(),
stax.Flatten(), stax.Dense(1))
for store_on_device in [False, True]:
for device_count in [0, 1]:
for get in ['ntk', 'nngp', ('nngp', 'ntk'), ('ntk', 'nngp')]:
with self.subTest(store_on_device=store_on_device,
device_count=device_count,
get=get):
kernel_fn_batched = batch.batch(
kernel_fn, 2, device_count, store_on_device)
predictor = predict.gradient_descent_mse_gp(
kernel_fn_batched, x_train, y_train, x_test, get,
0., True)
gp_inference = predict.gp_inference(
kernel_fn_batched, x_train, y_train, x_test, get,
0., True)
self.assertAllClose(predictor(None), predictor(np.inf),
True)
self.assertAllClose(predictor(None), gp_inference,
True)
def wide_resnet(block_size, k, num_classes):
return stax.serial(stax.Conv(16, (3, 3), padding='SAME'),
wide_resnet_group(block_size, int(16 * k)),
wide_resnet_group(block_size, int(32 * k), (2, 2)),
wide_resnet_group(block_size, int(64 * k), (2, 2)),
stax.AvgPool((8, 8)), stax.Flatten(),
stax.Dense(num_classes, 1., 0.))
def test_composition_conv(self, avg_pool, same_inputs):
rng = random.PRNGKey(0)
x1 = random.normal(rng, (3, 5, 5, 3))
x2 = None if same_inputs else random.normal(rng, (4, 5, 5, 3))
Block = stax.serial(stax.Conv(256, (3, 3)), stax.Relu())
if avg_pool:
Readout = stax.serial(stax.Conv(256, (3, 3)),
stax.GlobalAvgPool(),
stax.Dense(10))
else:
Readout = stax.serial(stax.Flatten(), stax.Dense(10))
block_ker_fn, readout_ker_fn = Block[2], Readout[2]
_, _, composed_ker_fn = stax.serial(Block, Readout)
composed_ker_out = composed_ker_fn(x1, x2)
ker_out_no_marg = readout_ker_fn(block_ker_fn(x1, x2,
diagonal_spatial=False))
ker_out_default = readout_ker_fn(block_ker_fn(x1, x2))
self.assertAllClose(composed_ker_out, ker_out_no_marg)
self.assertAllClose(composed_ker_out, ker_out_default)
if avg_pool:
with self.assertRaises(ValueError):
ker_out = readout_ker_fn(block_ker_fn(x1, x2, diagonal_spatial=True))
else:
ker_out_marg = readout_ker_fn(block_ker_fn(x1, x2,
diagonal_spatial=True))
self.assertAllClose(composed_ker_out, ker_out_marg)
def main(unused_argv):
key1, key2, key3 = random.split(random.PRNGKey(1), 3)
x1 = random.normal(key1, (2, 8, 8, 3))
x2 = random.normal(key2, (3, 8, 8, 3))
# A vanilla CNN.
init_fn, f, _ = stax.serial(
stax.Conv(8, (3, 3)),
stax.Relu(),
stax.Conv(8, (3, 3)),
stax.Relu(),
stax.Conv(8, (3, 3)),
stax.Flatten(),
stax.Dense(10)
)
_, params = init_fn(key3, x1.shape)
kwargs = dict(
f=f,
trace_axes=(),
vmap_axes=0,
)
# Default, baseline Jacobian contraction.
jacobian_contraction = nt.empirical_ntk_fn(
**kwargs,
implementation=nt.NtkImplementation.JACOBIAN_CONTRACTION)
# (6, 3, 10, 10) full `np.ndarray` test-train NTK
ntk_jc = jacobian_contraction(x2, x1, params)
# NTK-vector products-based implementation.
ntk_vector_products = nt.empirical_ntk_fn(
**kwargs,
implementation=nt.NtkImplementation.NTK_VECTOR_PRODUCTS)
ntk_vp = ntk_vector_products(x2, x1, params)
# Structured derivatives-based implementation.
structured_derivatives = nt.empirical_ntk_fn(
**kwargs,
implementation=nt.NtkImplementation.STRUCTURED_DERIVATIVES)
ntk_sd = structured_derivatives(x2, x1, params)
# Auto-FLOPs-selecting implementation. Doesn't work correctly on CPU/GPU.
auto = nt.empirical_ntk_fn(
**kwargs,
implementation=nt.NtkImplementation.AUTO)
ntk_auto = auto(x2, x1, params)
# Check that implementations match
for ntk1 in [ntk_jc, ntk_vp, ntk_sd, ntk_auto]:
for ntk2 in [ntk_jc, ntk_vp, ntk_sd, ntk_auto]:
diff = np.max(np.abs(ntk1 - ntk2))
print(f'NTK implementation diff {diff}.')
assert diff < (1e-4 if jax.default_backend() != 'tpu' else 0.1), diff
print('All NTK implementations match.')
def test_composition_conv(self, avg_pool):
rng = random.PRNGKey(0)
x1 = random.normal(rng, (5, 10, 10, 3))
x2 = random.normal(rng, (5, 10, 10, 3))
Block = stax.serial(stax.Conv(256, (3, 3)), stax.Relu())
if avg_pool:
Readout = stax.serial(stax.GlobalAvgPool(), stax.Dense(10))
marginalization = 'none'
else:
Readout = stax.serial(stax.Flatten(), stax.Dense(10))
marginalization = 'auto'
block_ker_fn, readout_ker_fn = Block[2], Readout[2]
_, _, composed_ker_fn = stax.serial(Block, Readout)
ker_out = readout_ker_fn(
block_ker_fn(x1, marginalization=marginalization))
composed_ker_out = composed_ker_fn(x1)
self.assertAllClose(ker_out, composed_ker_out, True)
if avg_pool:
with self.assertRaises(ValueError):
ker_out = readout_ker_fn(block_ker_fn(x1))
ker_out = readout_ker_fn(
block_ker_fn(x1, x2, marginalization=marginalization))
composed_ker_out = composed_ker_fn(x1, x2)
self.assertAllClose(ker_out, composed_ker_out, True)
def WideResnet(block_size, k, num_classes):
return stax.serial(
stax.Conv(16, (3, 3), padding='SAME'),
ntk_generator.ResnetGroup(block_size, int(16 * k)),
ntk_generator.ResnetGroup(block_size, int(32 * k), (2, 2)),
ntk_generator.ResnetGroup(block_size, int(64 * k), (2, 2)),
stax.Flatten(), stax.Dense(num_classes, 1., 0.))
def _get_net(W_std, b_std, filter_shape, is_conv, use_pooling, is_res,
padding, phi, strides, width, is_ntk):
fc = partial(stax.Dense, W_std=W_std, b_std=b_std)
conv = partial(
stax.Conv,
filter_shape=filter_shape,
strides=strides,
padding=padding,
W_std=W_std,
b_std=b_std)
affine = conv(width) if is_conv else fc(width)
res_unit = stax.serial((stax.AvgPool(
(2, 3), None, 'SAME' if padding == 'SAME' else 'CIRCULAR')
if use_pooling else stax.Identity()), phi, affine)
if is_res:
block = stax.serial(affine, stax.FanOut(2),
stax.parallel(stax.Identity(), res_unit),
stax.FanInSum())
else:
block = stax.serial(affine, res_unit)
readout = stax.serial(stax.GlobalAvgPool() if use_pooling else stax.Flatten(),
fc(1 if is_ntk else width))
net = stax.serial(block, readout)
return net
def test_nested_parallel(self, same_inputs, kernel_type):
platform = default_backend()
rtol = RTOL if platform != 'tpu' else 0.05
rng = random.PRNGKey(0)
(input_key1,
input_key2,
input_key3,
input_key4,
mask_key,
mc_key) = random.split(rng, 6)
x1_1, x2_1 = _get_inputs(input_key1, same_inputs, (BATCH_SIZE, 5))
x1_2, x2_2 = _get_inputs(input_key2, same_inputs, (BATCH_SIZE, 2, 2, 2))
x1_3, x2_3 = _get_inputs(input_key3, same_inputs, (BATCH_SIZE, 2, 2, 3))
x1_4, x2_4 = _get_inputs(input_key4, same_inputs, (BATCH_SIZE, 3, 4))
m1_key, m2_key, m3_key, m4_key = random.split(mask_key, 4)
x1_1 = test_utils.mask(
x1_1, mask_constant=-1, mask_axis=(1,), key=m1_key, p=0.5)
x1_2 = test_utils.mask(
x1_2, mask_constant=-1, mask_axis=(2, 3,), key=m2_key, p=0.5)
if not same_inputs:
x2_3 = test_utils.mask(
x2_3, mask_constant=-1, mask_axis=(1, 3,), key=m3_key, p=0.5)
x2_4 = test_utils.mask(
x2_4, mask_constant=-1, mask_axis=(2,), key=m4_key, p=0.5)
x1 = (((x1_1, x1_2), x1_3), x1_4)
x2 = (((x2_1, x2_2), x2_3), x2_4) if not same_inputs else None
N_in = 2 ** 7
# We only include dropout on non-TPU backends, because it takes large N to
# converge on TPU.
dropout_or_id = stax.Dropout(0.9) if platform != 'tpu' else stax.Identity()
init_fn, apply_fn, kernel_fn = stax.parallel(
stax.parallel(
stax.parallel(stax.Dense(N_in),
stax.serial(stax.Conv(N_in + 1, (2, 2)),
stax.Flatten())),
stax.serial(stax.Conv(N_in + 2, (2, 2)),
dropout_or_id,
stax.GlobalAvgPool())),
stax.Conv(N_in + 3, (2,)))
kernel_fn_empirical = nt.monte_carlo_kernel_fn(
init_fn, apply_fn, mc_key, N_SAMPLES,
implementation=_DEFAULT_TESTING_NTK_IMPLEMENTATION,
vmap_axes=(((((0, 0), 0), 0), (((0, 0), 0), 0), {})
if platform == 'tpu' else None)
)
test_utils.assert_close_matrices(
self,
kernel_fn(x1, x2, get=kernel_type, mask_constant=-1),
kernel_fn_empirical(x1, x2, get=kernel_type, mask_constant=-1),
rtol)
def _build_network(input_shape, network, out_logits, use_dropout):
dropout = stax.Dropout(0.9,
mode='train') if use_dropout else stax.Identity()
if len(input_shape) == 1:
assert network == 'FLAT'
return stax.serial(stax.Dense(WIDTH, W_std=2.0, b_std=0.5), dropout,
stax.Dense(out_logits, W_std=2.0, b_std=0.5))
elif len(input_shape) == 3:
if network == POOLING:
return stax.serial(
stax.Conv(CONVOLUTION_CHANNELS, (2, 2), W_std=2.0, b_std=0.05),
stax.GlobalAvgPool(), dropout,
stax.Dense(out_logits, W_std=2.0, b_std=0.5))
elif network == FLAT:
return stax.serial(
stax.Conv(CONVOLUTION_CHANNELS, (2, 2), W_std=2.0, b_std=0.05),
stax.Flatten(), dropout,
stax.Dense(out_logits, W_std=2.0, b_std=0.5))
elif network == INTERMEDIATE_CONV:
return stax.Conv(CONVOLUTION_CHANNELS, (2, 2),
W_std=2.0,
b_std=0.05)
else:
raise ValueError(
'Unexpected network type found: {}'.format(network))
else:
raise ValueError('Expected flat or image test input.')
def test_hermite(self, same_inputs, degree, get, readout):
key = random.PRNGKey(1)
key1, key2, key = random.split(key, 3)
if degree > 2:
width = 10000
n_samples = 5000
test_utils.skip_test(self)
else:
width = 10000
n_samples = 100
x1 = np.cos(random.normal(key1, [2, 6, 6, 3]))
x2 = x1 if same_inputs else np.cos(random.normal(key2, [3, 6, 6, 3]))
conv_layers = [
stax.Conv(width, (3, 3), W_std=2., b_std=0.5),
stax.LayerNorm(),
stax.Hermite(degree),
stax.GlobalAvgPool() if readout == 'pool' else stax.Flatten(),
stax.Dense(1) if get == 'ntk' else stax.Identity()]
init_fn, apply_fn, kernel_fn = stax.serial(*conv_layers)
analytic_kernel = kernel_fn(x1, x2, get)
mc_kernel_fn = nt.monte_carlo_kernel_fn(init_fn, apply_fn, key, n_samples)
mc_kernel = mc_kernel_fn(x1, x2, get)
rot = degree / 2. * 1e-2
test_utils.assert_close_matrices(self, mc_kernel, analytic_kernel, rot)
def _build_network(input_shape, network, out_logits):
if len(input_shape) == 1:
assert network == FLAT
return stax.Dense(out_logits, W_std=2.0, b_std=0.5)
elif len(input_shape) == 3:
if network == POOLING:
return stax.serial(
stax.Conv(CONVOLUTION_CHANNELS, (3, 3), W_std=2.0, b_std=0.05),
stax.GlobalAvgPool(),
stax.Dense(out_logits, W_std=2.0, b_std=0.5))
elif network == CONV:
return stax.serial(
stax.Conv(CONVOLUTION_CHANNELS, (1, 2), W_std=1.5, b_std=0.1),
stax.Relu(),
stax.Conv(CONVOLUTION_CHANNELS, (3, 2), W_std=2.0, b_std=0.05),
)
elif network == FLAT:
return stax.serial(
stax.Conv(CONVOLUTION_CHANNELS, (3, 3), W_std=2.0, b_std=0.05),
stax.Flatten(), stax.Dense(out_logits, W_std=2.0, b_std=0.5))
else:
raise ValueError(
'Unexpected network type found: {}'.format(network))
else:
raise ValueError('Expected flat or image test input.')
def testPredictOnCPU(self):
x_train = random.normal(random.PRNGKey(1), (4, 4, 4, 2))
x_test = random.normal(random.PRNGKey(1), (8, 4, 4, 2))
y_train = random.uniform(random.PRNGKey(1), (4, 2))
_, _, kernel_fn = stax.serial(
stax.Conv(1, (3, 3)), stax.Relu(), stax.Flatten(), stax.Dense(1))
for store_on_device in [False, True]:
for device_count in [0, 1]:
for get in ['ntk', 'nngp', ('nngp', 'ntk'), ('ntk', 'nngp')]:
for x in [None, 'x_test']:
with self.subTest(
store_on_device=store_on_device,
device_count=device_count,
get=get,
x=x):
kernel_fn_batched = batch.batch(kernel_fn, 2, device_count,
store_on_device)
predictor = predict.gradient_descent_mse_ensemble(
kernel_fn_batched, x_train, y_train)
x = x if x is None else x_test
predict_none = predictor(None, x, get, compute_cov=True)
predict_inf = predictor(np.inf, x, get, compute_cov=True)
self.assertAllClose(predict_none, predict_inf)
if x is not None:
on_cpu = (not store_on_device or
xla_bridge.get_backend().platform == 'cpu')
self.assertEqual(on_cpu, utils.is_on_cpu(predict_inf))
self.assertEqual(on_cpu, utils.is_on_cpu(predict_none))
def Resnet(block_size, num_classes):
return stax.serial(stax.Conv(64, (3, 3), padding='SAME'),
ResnetGroup(block_size, 64),
ResnetGroup(block_size, 128, (2, 2)),
ResnetGroup(block_size, 256, (2, 2)),
ResnetGroup(block_size, 512, (2, 2)), stax.Flatten(),
stax.Dense(num_classes, 1., 0.05))
def _test_activation(self, activation_fn, same_inputs, model, get,
rbf_gamma=None):
if 'conv' in model:
test_utils.skip_test(self)
key = random.PRNGKey(1)
key, split = random.split(key)
output_dim = 1024 if get == 'nngp' else 1
b_std = 0.5
W_std = 2.0
if activation_fn[2].__name__ == 'Sin':
W_std = 0.9
if activation_fn[2].__name__ == 'Rbf':
W_std = 1.0
b_std = 0.0
if model == 'fc':
rtol = 0.04
X0_1 = random.normal(key, (4, 2))
X0_2 = None if same_inputs else random.normal(split, (2, 2))
affine = stax.Dense(1024, W_std, b_std)
readout = stax.Dense(output_dim)
depth = 1
else:
rtol = 0.05
X0_1 = random.normal(key, (2, 4, 4, 3))
X0_2 = None if same_inputs else random.normal(split, (4, 4, 4, 3))
affine = stax.Conv(512, (3, 2), W_std=W_std, b_std=b_std, padding='SAME')
readout = stax.serial(stax.GlobalAvgPool() if 'pool' in model else
stax.Flatten(),
stax.Dense(output_dim))
depth = 2
if default_backend() == 'cpu':
num_samplings = 200
rtol *= 2
else:
num_samplings = (500 if activation_fn[2].__name__ in ('Sin', 'Rbf')
else 300)
init_fn, apply_fn, kernel_fn = stax.serial(
*[affine, activation_fn]*depth, readout)
analytic_kernel = kernel_fn(X0_1, X0_2, get)
mc_kernel_fn = nt.monte_carlo_kernel_fn(
init_fn, apply_fn, split, num_samplings, implementation=2,
vmap_axes=0
)
empirical_kernel = mc_kernel_fn(X0_1, X0_2, get)
test_utils.assert_close_matrices(self, analytic_kernel,
empirical_kernel, rtol)
# Check match with explicit RBF
if rbf_gamma is not None and get == 'nngp' and model == 'fc':
input_dim = X0_1.shape[1]
_, _, kernel_fn = self._RBF(rbf_gamma / input_dim)
direct_rbf_kernel = kernel_fn(X0_1, X0_2, get)
test_utils.assert_close_matrices(self, analytic_kernel,
direct_rbf_kernel, rtol)
def _get_inputs_and_model(width=1, n_classes=2):
key = random.PRNGKey(1)
key, split = random.split(key)
x1 = random.normal(key, (8, 4, 3, 2))
x2 = random.normal(split, (4, 4, 3, 2))
init_fun, apply_fun, ker_fun = stax.serial(stax.Conv(width, (3, 3)),
stax.Relu(), stax.Flatten(),
stax.Dense(n_classes, 2., 0.5))
return x1, x2, init_fun, apply_fun, ker_fun, key
def testPredictOnCPU(self):
key1 = stateless_uniform(shape=[2],
seed=[1, 1],
minval=None,
maxval=None,
dtype=tf.int32)
key2 = stateless_uniform(shape=[2],
seed=[1, 1],
minval=None,
maxval=None,
dtype=tf.int32)
key3 = stateless_uniform(shape=[2],
seed=[1, 1],
minval=None,
maxval=None,
dtype=tf.int32)
x_train = np.asarray(normal((4, 4, 4, 2), seed=key1))
x_test = np.asarray(normal((8, 4, 4, 2), seed=key2))
y_train = np.asarray(stateless_uniform(shape=(4, 2), seed=key3))
_, _, kernel_fn = stax.serial(stax.Conv(1, (3, 3)), stax.Relu(),
stax.Flatten(), stax.Dense(1))
for store_on_device in [False, True]:
for device_count in [0, 1]:
for get in ['ntk', 'nngp', ('nngp', 'ntk'), ('ntk', 'nngp')]:
for x in [None, 'x_test']:
with self.subTest(store_on_device=store_on_device,
device_count=device_count,
get=get,
x=x):
kernel_fn_batched = batch.batch(
kernel_fn, 2, device_count, store_on_device)
predictor = predict.gradient_descent_mse_ensemble(
kernel_fn_batched, x_train, y_train)
x = x if x is None else x_test
predict_none = predictor(None,
x,
get,
compute_cov=True)
predict_inf = predictor(np.inf,
x,
get,
compute_cov=True)
self.assertAllClose(predict_none, predict_inf)
if x is not None:
on_cpu = (not store_on_device
or xla_bridge.get_backend().platform
== 'cpu')
self.assertEqual(on_cpu,
utils.is_on_cpu(predict_inf))
self.assertEqual(on_cpu,
utils.is_on_cpu(predict_none))
def WideResnet(block_size, k, num_classes):
return stax.serial(
stax.Conv(16, (3, 3), padding='SAME', parameterization='standard'),
WideResnetGroup(block_size, int(16 * k)),
WideResnetGroup(block_size, int(32 * k), (2, 2)),
WideResnetGroup(block_size, int(64 * k), (2, 2)),
stax.AvgPool((8, 8), padding='SAME'),
stax.Flatten(),
stax.Dense(num_classes, 1.0, 0.0, parameterization='standard'),
)
def _get_net(W_std, b_std, filter_shape, is_conv, use_pooling, is_res,
padding, phi, strides, width, is_ntk, proj_into_2d):
fc = partial(stax.Dense, W_std=W_std, b_std=b_std)
conv = partial(
stax.Conv,
filter_shape=filter_shape,
strides=strides,
padding=padding,
W_std=W_std,
b_std=b_std)
affine = conv(width) if is_conv else fc(width)
res_unit = stax.serial((stax.AvgPool(
(2, 3), None, 'SAME' if padding == 'SAME' else 'CIRCULAR')
if use_pooling else stax.Identity()), phi, affine)
if is_res:
block = stax.serial(affine, stax.FanOut(2),
stax.parallel(stax.Identity(), res_unit),
stax.FanInSum())
else:
block = stax.serial(affine, res_unit)
if proj_into_2d == 'FLAT':
proj_layer = stax.Flatten()
elif proj_into_2d == 'POOL':
proj_layer = stax.GlobalAvgPool()
elif proj_into_2d.startswith('ATTN'):
n_heads = int(np.sqrt(width))
n_chan_val = int(np.round(float(width) / n_heads))
fixed = proj_into_2d == 'ATTN_FIXED'
proj_layer = stax.serial(
stax.GlobalSelfAttention(
width, n_chan_key=width, n_chan_val=n_chan_val, n_heads=n_heads,
fixed=fixed, W_key_std=W_std, W_value_std=W_std, W_query_std=W_std,
W_out_std=1.0, b_std=b_std),
stax.Flatten())
else:
raise ValueError(proj_into_2d)
readout = stax.serial(proj_layer, fc(1 if is_ntk else width))
return stax.serial(block, readout)
def WideResnet(block_size, k, num_classes, W_std=1., b_std=0.):
return stax.serial(
stax.Conv(16, (3, 3), W_std=W_std, b_std=b_std, padding='SAME'),
WideResnetGroup(block_size, int(16 * k), W_std=W_std, b_std=b_std),
WideResnetGroup(block_size,
int(32 * k), (2, 2),
W_std=W_std,
b_std=b_std),
WideResnetGroup(block_size,
int(64 * k), (2, 2),
W_std=W_std,
b_std=b_std), stax.AvgPool((7, 7)), stax.Flatten(),
stax.Dense(num_classes, W_std=W_std, b_std=b_std))
def _get_inputs_and_model(width=1, n_classes=2, use_conv=True):
key = random.PRNGKey(1)
key, split = random.split(key)
x1 = random.normal(key, (8, 4, 3, 2))
x2 = random.normal(split, (4, 4, 3, 2))
if not use_conv:
x1 = np.reshape(x1, (x1.shape[0], -1))
x2 = np.reshape(x2, (x2.shape[0], -1))
init_fn, apply_fn, kernel_fn = stax.serial(
stax.Conv(width, (3, 3)) if use_conv else stax.Dense(width),
stax.Relu(), stax.Flatten(), stax.Dense(n_classes, 2., 0.5))
return x1, x2, init_fn, apply_fn, kernel_fn, key
def build_le_net(network_width):
""" Construct the LeNet of width network_width with average pooling using neural tangent's stax."""
return stax.serial(
stax.Conv(out_chan=6 * network_width,
filter_shape=(3, 3),
strides=(1, 1),
padding='VALID'), stax.Relu(),
stax.AvgPool(window_shape=(2, 2), strides=(2, 2)),
stax.Conv(out_chan=16 * network_width,
filter_shape=(3, 3),
strides=(1, 1),
padding='VALID'), stax.Relu(),
stax.AvgPool(window_shape=(2, 2), strides=(2, 2)), stax.Flatten(),
stax.Dense(120 * network_width), stax.Relu(),
stax.Dense(84 * network_width), stax.Relu(), stax.Dense(10))
def test_flatten_first(self, same_inputs):
key = random.PRNGKey(1)
X0_1 = random.normal(key, (5, 4, 3, 2))
X0_2 = None if same_inputs else random.normal(key, (3, 4, 3, 2))
X0_1_flat = np.reshape(X0_1, (X0_1.shape[0], -1))
X0_2_flat = None if same_inputs else np.reshape(
X0_2, (X0_2.shape[0], -1))
_, _, fc_flat = stax.serial(stax.Dense(10, 2., 0.5), stax.Erf())
_, _, fc = stax.serial(stax.Flatten(), stax.Dense(10, 2., 0.5),
stax.Erf())
K_flat = fc_flat(X0_1_flat, X0_2_flat)
K = fc(X0_1, X0_2)
self.assertAllClose(K_flat, K, True)
def _MyrtleNetwork(width, depth, W_std=jnp.sqrt(2.0), b_std=0.0):
layer_factor = {5: [2, 1, 1], 7: [2, 2, 2], 10: [3, 3, 3]}
activation_fn = stax.Relu()
layers = []
conv = functools.partial(stax.Conv,
W_std=W_std,
b_std=b_std,
padding="SAME")
layers += [conv(width, (3, 3)), activation_fn] * layer_factor[depth][0]
layers += [stax.AvgPool((2, 2), strides=(2, 2))]
layers += [conv(width, (3, 3)), activation_fn] * layer_factor[depth][1]
layers += [stax.AvgPool((2, 2), strides=(2, 2))]
layers += [conv(width, (3, 3)), activation_fn] * layer_factor[depth][2]
layers += [stax.AvgPool((2, 2), strides=(2, 2))] * 3
layers += [stax.Flatten(), stax.Dense(10, W_std, b_std)]
return stax.serial(*layers)
def _get_inputs_and_model(width=1, n_classes=2, use_conv=True):
key = stateless_uniform(shape=[2],
seed=[1, 1],
minval=None,
maxval=None,
dtype=tf.int32)
keys = tf_random_split(key)
key = keys[0]
split = keys[1]
x1 = np.asarray(normal((8, 4, 3, 2), seed=key))
x2 = np.asarray(normal((4, 4, 3, 2), seed=split))
if not use_conv:
x1 = np.reshape(x1, (x1.shape[0], -1))
x2 = np.reshape(x2, (x2.shape[0], -1))
init_fn, apply_fn, kernel_fn = stax.serial(
stax.Conv(width, (3, 3)) if use_conv else stax.Dense(width),
stax.Relu(), stax.Flatten(), stax.Dense(n_classes, 2., 0.5))
return x1, x2, init_fn, apply_fn, kernel_fn, key
def test_elementwise_numerical(self, same_inputs, model, phi, get):
if 'conv' in model:
test_utils.skip_test(self)
key, split = random.split(random.PRNGKey(1))
output_dim = 1
b_std = 0.01
W_std = 1.0
rtol = 2e-3
deg = 25
if get == 'ntk':
rtol *= 2
if default_backend() == 'tpu':
rtol *= 2
if model == 'fc':
X0_1 = random.normal(key, (3, 7))
X0_2 = None if same_inputs else random.normal(split, (5, 7))
affine = stax.Dense(1024, W_std, b_std)
readout = stax.Dense(output_dim)
depth = 1
else:
X0_1 = random.normal(key, (2, 8, 8, 3))
X0_2 = None if same_inputs else random.normal(split, (3, 8, 8, 3))
affine = stax.Conv(1024, (3, 2), W_std=W_std, b_std=b_std, padding='SAME')
readout = stax.serial(stax.GlobalAvgPool() if 'pool' in model else
stax.Flatten(),
stax.Dense(output_dim))
depth = 2
_, _, kernel_fn = stax.serial(*[affine, phi] * depth, readout)
analytic_kernel = kernel_fn(X0_1, X0_2, get)
fn = lambda x: phi[1]((), x)
_, _, kernel_fn = stax.serial(
*[affine, stax.ElementwiseNumerical(fn, deg=deg)] * depth, readout)
numerical_activation_kernel = kernel_fn(X0_1, X0_2, get)
test_utils.assert_close_matrices(self, analytic_kernel,
numerical_activation_kernel, rtol)
def CNNStandard(n_channels,
L,
filter=(3, 3),
data='cifar10',
gap=True,
nonlinearity='relu',
parameterization='standard',
order=None):
if data == 'cifar10':
num_classes = 10
if data == 'cifar100':
num_classes = 100
if nonlinearity == 'relu':
nonlin = Relu
elif nonlinearity == 'swish':
nonlin = Swish
init_fn, f = jax_stax.serial(*[
jax_stax.serial(
MyConv(n_channels,
filter,
parameterization=parameterization,
order=order),
nonlin,
) for _ in range(L)
])
if gap:
init_fn, f = jax_stax.serial((init_fn, f),
stax.GlobalAvgPool()[:2],
MyDense(num_classes,
parameterization=parameterization,
order=order))
else:
init_fn, f = jax_stax.serial((init_fn, f),
stax.Flatten()[:2],
MyDense(num_classes,
parameterization=parameterization,
order=order))
return init_fn, f
def cnn1d(hidden_widths, nonlin, args):
layers = []
layers_ker = []
window = (3, )
stride = (1, )
if args != None:
window = args
for i, ch in enumerate(hidden_widths):
layers += [
stax.Conv(ch,
window,
stride,
'CIRCULAR',
b_std=0,
parameterization='ntk')
]
if nonlin != 'relu':
layers_ker += [i]
else:
layers += [stax.Relu()]
layers_ker += [2 * i + 1]
layers += [stax.Flatten()]
if nonlin != 'relu':
layers_ker += [i + 1]
else:
layers += [stax.Relu()]
layers_ker += [2 * i + 3]
layers += [stax.Dense(10, parameterization='ntk')]
if nonlin != 'relu':
layers_ker += [i + 2]
else:
layers_ker += [2 * i + 4]
return layers, layers_ker
def test_fan_in_conv(self, same_inputs, axis, n_branches, get, branch_in,
readout, fan_in_mode):
test_utils.skip_test(self)
if fan_in_mode in ['FanInSum', 'FanInProd']:
if axis != 0:
raise absltest.SkipTest(
'`FanInSum` and `FanInProd()` are skipped when '
'axis != 0.')
axis = None
if (fan_in_mode == 'FanInSum'
or axis in [0, 1, 2]) and branch_in == 'dense_after_branch_in':
raise absltest.SkipTest('`FanInSum` and `FanInConcat(0/1/2)` '
'require `is_gaussian`.')
if ((axis == 3 or fan_in_mode == 'FanInProd')
and branch_in == 'dense_before_branch_in'):
raise absltest.SkipTest(
'`FanInConcat` or `FanInProd` on feature axis '
'requires a dense layer after concatenation '
'or Hadamard product.')
if fan_in_mode == 'FanInSum':
fan_in_layer = stax.FanInSum()
elif fan_in_mode == 'FanInProd':
fan_in_layer = stax.FanInProd()
else:
fan_in_layer = stax.FanInConcat(axis)
key = random.PRNGKey(1)
X0_1 = random.normal(key, (2, 5, 6, 3))
X0_2 = None if same_inputs else random.normal(key, (3, 5, 6, 3))
if default_backend() == 'tpu':
width = 2048
n_samples = 1024
tol = 0.02
else:
width = 1024
n_samples = 512
tol = 0.01
conv = stax.Conv(out_chan=width,
filter_shape=(3, 3),
padding='SAME',
W_std=1.25,
b_std=0.1)
input_layers = [conv, stax.FanOut(n_branches)]
branches = []
for b in range(n_branches):
branch_layers = [FanInTest._get_phi(b)]
for i in range(b):
multiplier = 1 if axis not in (3, -1) else (1 + 0.25 * i)
branch_layers += [
stax.Conv(out_chan=int(width * multiplier),
filter_shape=(i + 1, 4 - i),
padding='SAME',
W_std=1.25 + i,
b_std=0.1 + i),
FanInTest._get_phi(i)
]
if branch_in == 'dense_before_branch_in':
branch_layers += [conv]
branches += [stax.serial(*branch_layers)]
output_layers = [
fan_in_layer,
stax.Relu(),
stax.GlobalAvgPool() if readout == 'pool' else stax.Flatten()
]
if branch_in == 'dense_after_branch_in':
output_layers.insert(1, conv)
nn = stax.serial(*(input_layers + [stax.parallel(*branches)] +
output_layers))
init_fn, apply_fn, kernel_fn = stax.serial(
nn, stax.Dense(1 if get == 'ntk' else width, 1.25, 0.5))
kernel_fn_mc = nt.monte_carlo_kernel_fn(
init_fn,
apply_fn,
key,
n_samples,
device_count=0 if axis in (0, -4) else -1,
implementation=_DEFAULT_TESTING_NTK_IMPLEMENTATION,
vmap_axes=None if axis in (0, -4) else 0,
)
exact = kernel_fn(X0_1, X0_2, get=get)
empirical = kernel_fn_mc(X0_1, X0_2, get=get)
test_utils.assert_close_matrices(self, empirical, exact, tol)
