def build_optimizer(self,
clip=15.0,
lr=5e-4,
warmup=2000,
cosine_decay_steps=None,
optimizer_name="adabelief") -> GradientTransformation:
chain = []
if optimizer_name == "adabelief":
chain.append(util.scale_by_belief())
elif optimizer_name == "adam":
chain.append(optax.scale_by_adam())
else:
assert 0
# Make sure to use the negative learning rate so that we minimize
if warmup and warmup > 0:
warmup_schedule = partial(util.linear_warmup_lr_schedule,
warmup=warmup,
lr_decay=1.0,
lr=-lr)
chain.append(optax.scale_by_schedule(warmup_schedule))
else:
chain.append(optax.scale(-lr))
if cosine_decay_steps and cosine_decay_steps > 0:
cosine_lr = optax.cosine_decay_schedule(
init_value=1.0, decay_steps=cosine_decay_steps, alpha=1e-1)
chain.append(optax.scale_by_schedule(cosine_lr))
if clip and clip > 0:
chain.append(optax.clip(clip))
return optax.chain(*chain)
def test_optimizer_chain(self):
optimizer = elegy.Optimizer(
optax.sgd(0.1),
optax.clip(0.5),
)
params = np.zeros(shape=(3, 4))
grads = np.ones(shape=(3, 4)) * 100_000
rng = elegy.RNGSeq(42)
optimizer_states = optimizer.init(
rng=rng,
net_params=params,
)
params, optimizer_states = optimizer.apply(params, grads,
optimizer_states, rng)
assert np.all(-0.5 <= params) and np.all(params <= 0.5)
def create_train_state(config, rng, learning_rate_fn, example_batch):
"""Create and initialize the model.
Args:
config: Configuration for model.
rng: JAX PRNG Key.
learning_rate_fn: learning rate function
example_batch: for model intialization
Returns:
The initialized TrainState with the optimizer.
"""
model, variables = create_model(config, rng, example_batch)
params = variables['params']
parameter_overview.log_parameter_overview(params)
optimizer = optax.adamw(learning_rate=learning_rate_fn,
b1=0.9,
b2=.98,
eps=1e-9,
weight_decay=config.train.weight_decay)
if config.train.grad_max_norm > 0:
tx = optax.chain(optax.clip_by_global_norm(config.train.grad_max_norm),
optimizer)
elif config.train.grad_max_val > 1:
tx = optax.chain(optax.clip(config.train.grad_max_val), optimizer)
else:
tx = optimizer
state = train_state.TrainState.create(
apply_fn=model.apply,
params=variables['params'],
tx=tx,
)
return model, state
import numpyro.distributions as dist
from numpyro.distributions import constraints
from numpyro.infer import SVI, RenyiELBO, Trace_ELBO
try:
import optax
from numpyro.contrib.optim import optax_to_numpyro
# the optimizer test is parameterized by different optax optimizers, but we have
# to define them here to ensure that `optax` is defined. pytest.mark.parameterize
# decorators are run even if tests are skipped at the top of the file.
optimizers = [
(optax.adam, (1e-2, ), {}),
# clipped adam
(optax.chain, (optax.clip(10.0), optax.adam(1e-2)), {}),
(optax.adagrad, (1e-1, ), {}),
# SGD with momentum
(optax.sgd, (1e-2, ), {
"momentum": 0.9
}),
(optax.rmsprop, (1e-2, ), {
"decay": 0.95
}),
# RMSProp with momentum
(optax.rmsprop, (1e-4, ), {
"decay": 0.9,
"momentum": 0.9
}),
(optax.sgd, (1e-2, ), {}),
]
accuracy), grads = jax.value_and_grad(loss,
has_aux=True)(state.params,
batch, labels)
state = state.apply_gradients(grads=grads)
return state, loss_val, accuracy
for i, (batch, labels) in enumerate(zip(train_data, train_labels)):
state, loss_val, accuracy = step(state, batch, labels)
if i % 100 == 0:
print(
f"step {i}/{nb_steps} | loss: {loss_val:.5f} | accuracy: {accuracy*100:.2f}%"
)
return params
params = net.init(jax.random.PRNGKey(0), train_data[0])
schedule = optax.warmup_cosine_decay_schedule(
init_value=0.0,
peak_value=1.0,
warmup_steps=50,
decay_steps=5_000,
end_value=0.0,
)
opt = optax.chain(
optax.clip(1.0),
# optax.adamw(learning_rate=1e-3)
optax.adamw(learning_rate=schedule))
params = fit(params, opt)
def create_optimizer(config):
"""Creates the optimizer associated to a config."""
ops = []
# Gradient clipping either by norm `gradient_norm_clip` or by absolute value
# `gradient_value_clip`.
if "gradient_clip" in config:
raise ValueError("'gradient_clip' is deprecated, please use "
"'gradient_norm_clip'.")
assert not ("gradient_norm_clip" in config
and "gradient_value_clip" in config), (
"Gradient clipping by norm and by value are exclusive.")
if "gradient_norm_clip" in config:
ops.append(optax.clip_by_global_norm(config.gradient_norm_clip))
if "gradient_value_clip" in config:
ops.append(optax.clip(config.gradient_value_clip))
# Define the learning rate schedule.
schedule_fn = utils.get_optax_schedule_fn(
warmup_ratio=config.get("warmup_ratio", 0.),
num_train_steps=config.num_train_steps,
decay=config.get("learning_rate_step_decay", 1.0),
decay_at_steps=config.get("learning_rate_decay_at_steps", []),
cosine_decay_schedule=config.get("cosine_decay", False))
schedule_ops = [optax.scale_by_schedule(schedule_fn)]
# Scale some parameters matching a regex by a multiplier. Config field
# `scaling_by_regex` is a list of pairs (regex: str, multiplier: float).
scaling_by_regex = config.get("scaling_learning_rate_by_regex", [])
for regex, multiplier in scaling_by_regex:
logging.info(
"Learning rate is scaled by %f for parameters matching '%s'",
multiplier, regex)
schedule_ops.append(utils.scale_selected_parameters(regex, multiplier))
schedule_optimizer = optax.chain(*schedule_ops)
if config.optimizer.lower() == "adam":
optimizer = optax.adam(config.learning_rate)
ops.append(optimizer)
ops.append(schedule_optimizer)
elif config.optimizer.lower() == "sgd":
ops.append(schedule_optimizer)
optimizer = optax.sgd(config.learning_rate, momentum=config.momentum)
ops.append(optimizer)
else:
raise NotImplementedError("Invalid optimizer: {}".format(
config.optimizer))
if "weight_decay" in config and config.weight_decay > 0.:
ops.append(
utils.decoupled_weight_decay(decay=config.weight_decay,
step_size_fn=schedule_fn))
# Freeze parameters that match the given regexes (if any).
freeze_weights_regexes = config.get("freeze_weights_regex", []) or []
if isinstance(freeze_weights_regexes, str):
freeze_weights_regexes = [freeze_weights_regexes]
for reg in freeze_weights_regexes:
ops.append(utils.freeze(reg))
return optax.chain(*ops)
state, loss_val, accuracy = train_step(state, batch, labels)
train_loss += loss_val
train_accuracy += accuracy
test_loss, test_accuracy = 0.0, 0.0
for batch, labels in test_loader:
batch, labels = jnp.array(batch), jnp.array(labels)
loss_val, accuracy = eval_step(state.params, batch, labels)
test_loss += loss_val
test_accuracy += accuracy
train_loss /= len(train_loader)
train_accuracy /= len(train_loader)
test_loss /= len(test_loader)
test_accuracy /= len(test_loader)
print(
f"epoch {i+1}/{nb_epochs} | train: {train_loss:.5f} [{train_accuracy*100:.2f}%] | eval: {test_loss:.5f} [{test_accuracy*100:.2f}%]"
)
return params
params = net.init(jax.random.PRNGKey(0),
jnp.ones((1, *image_shape)).astype(jnp.float32))
opt = optax.chain(optax.clip(1.0), optax.adamw(learning_rate=1e-4))
params = fit(params, opt)
(loss_val, accuracy), grads = jax.value_and_grad(loss,
has_aux=True)(params,
batch,
labels)
updates, opt_state = opt.update(grads, opt_state, params)
params = optax.apply_updates(params, updates)
return params, opt_state, loss_val, accuracy
for i, (batch, labels) in enumerate(zip(train_data, train_labels)):
params, opt_state, loss_val, accuracy = step(params, opt_state, batch,
labels)
if i % 100 == 0:
print(
f"step {i}/{nb_steps} | loss: {loss_val:.5f} | accuracy: {accuracy*100:.2f}%"
)
return params
# opt = optax.adam(learning_rate=1e-2)
schedule = optax.warmup_cosine_decay_schedule(
init_value=0.0,
peak_value=1.0,
warmup_steps=50,
decay_steps=5_000,
end_value=0.0,
)
opt = optax.chain(optax.clip(1.0), optax.adamw(learning_rate=schedule))
params = fit(params, opt)
