optimization.py

# Copyright 2019 The TensorFlow Authors. All Rights Reserved.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
#     http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
# ==============================================================================
"""Functions and classes related to optimization (weight updates)."""
from __future__ import absolute_import
from __future__ import division
from __future__ import print_function

import re

from absl import logging
import gin
import tensorflow as tf
import tensorflow_addons.optimizers as tfa_optimizers


class WarmUp(tf.keras.optimizers.schedules.LearningRateSchedule):
  """Applies a warmup schedule on a given learning rate decay schedule."""

  def __init__(self,
               initial_learning_rate,
               decay_schedule_fn,
               warmup_steps,
               power=1.0,
               name=None):
    super(WarmUp, self).__init__()
    self.initial_learning_rate = initial_learning_rate
    self.warmup_steps = warmup_steps
    self.power = power
    self.decay_schedule_fn = decay_schedule_fn
    self.name = name

  def __call__(self, step):
    with tf.name_scope(self.name or 'WarmUp') as name:
      # Implements polynomial warmup. i.e., if global_step < warmup_steps, the
      # learning rate will be `global_step/num_warmup_steps * init_lr`.
      global_step_float = tf.cast(step, tf.float32)
      warmup_steps_float = tf.cast(self.warmup_steps, tf.float32)
      warmup_percent_done = global_step_float / warmup_steps_float
      warmup_learning_rate = (
          self.initial_learning_rate *
          tf.math.pow(warmup_percent_done, self.power))
      return tf.cond(
          global_step_float < warmup_steps_float,
          lambda: warmup_learning_rate,
          lambda: self.decay_schedule_fn(step),
          name=name)

  def get_config(self):
    return {
        'initial_learning_rate': self.initial_learning_rate,
        'decay_schedule_fn': self.decay_schedule_fn,
        'warmup_steps': self.warmup_steps,
        'power': self.power,
        'name': self.name
    }


@gin.configurable
def create_optimizer(init_lr,
                     num_train_steps,
                     num_warmup_steps,
                     end_lr=0.0,
                     optimizer_type='adamw'):
  """Creates an optimizer with learning rate schedule."""
  # Implements linear decay of the learning rate.
  lr_schedule = tf.keras.optimizers.schedules.PolynomialDecay(
      initial_learning_rate=init_lr,
      decay_steps=num_train_steps,
      end_learning_rate=end_lr)
  if num_warmup_steps:
    lr_schedule = WarmUp(
        initial_learning_rate=init_lr,
        decay_schedule_fn=lr_schedule,
        warmup_steps=num_warmup_steps)

  if optimizer_type == 'adamw':
    logging.info('using Adamw optimizer')
    optimizer = AdamWeightDecay(
        learning_rate=lr_schedule,
        weight_decay_rate=0.01,
        beta_1=0.9,
        beta_2=0.999,
        epsilon=1e-6,
        exclude_from_weight_decay=['LayerNorm', 'layer_norm', 'bias'])
  elif optimizer_type == 'lamb':
    logging.info('using Lamb optimizer')
    optimizer = tfa_optimizers.LAMB(
        learning_rate=lr_schedule,
        weight_decay_rate=0.01,
        beta_1=0.9,
        beta_2=0.999,
        epsilon=1e-6,
        exclude_from_weight_decay=['LayerNorm', 'layer_norm', 'bias'])
  else:
    raise ValueError('Unsupported optimizer type: ', optimizer_type)

  return optimizer


class AdamWeightDecay(tf.keras.optimizers.Adam):
  """Adam enables L2 weight decay and clip_by_global_norm on gradients.

  Just adding the square of the weights to the loss function is *not* the
  correct way of using L2 regularization/weight decay with Adam, since that will
  interact with the m and v parameters in strange ways.

  Instead we want ot decay the weights in a manner that doesn't interact with
  the m/v parameters. This is equivalent to adding the square of the weights to
  the loss with plain (non-momentum) SGD.
  """

  def __init__(self,
               learning_rate=0.001,
               beta_1=0.9,
               beta_2=0.999,
               epsilon=1e-7,
               amsgrad=False,
               weight_decay_rate=0.0,
               include_in_weight_decay=None,
               exclude_from_weight_decay=None,
               gradient_clip_norm=1.0,
               name='AdamWeightDecay',
               **kwargs):
    super(AdamWeightDecay, self).__init__(learning_rate, beta_1, beta_2,
                                          epsilon, amsgrad, name, **kwargs)
    self.weight_decay_rate = weight_decay_rate
    self.gradient_clip_norm = gradient_clip_norm
    self._include_in_weight_decay = include_in_weight_decay
    self._exclude_from_weight_decay = exclude_from_weight_decay
    logging.info('gradient_clip_norm=%f', gradient_clip_norm)

  @classmethod
  def from_config(cls, config):
    """Creates an optimizer from its config with WarmUp custom object."""
    custom_objects = {'WarmUp': WarmUp}
    return super(AdamWeightDecay, cls).from_config(
        config, custom_objects=custom_objects)

  def _prepare_local(self, var_device, var_dtype, apply_state):
    super(AdamWeightDecay, self)._prepare_local(var_device, var_dtype,
                                                apply_state)
    apply_state[(var_device, var_dtype)]['weight_decay_rate'] = tf.constant(
        self.weight_decay_rate, name='adam_weight_decay_rate')

  def _decay_weights_op(self, var, learning_rate, apply_state):
    do_decay = self._do_use_weight_decay(var.name)
    if do_decay:
      return var.assign_sub(
          learning_rate * var *
          apply_state[(var.device, var.dtype.base_dtype)]['weight_decay_rate'],
          use_locking=self._use_locking)
    return tf.no_op()

  def apply_gradients(self,
                      grads_and_vars,
                      name=None,
                      experimental_aggregate_gradients=True):
    grads, tvars = list(zip(*grads_and_vars))
    if experimental_aggregate_gradients and self.gradient_clip_norm > 0.0:
      # when experimental_aggregate_gradients = False, apply_gradients() no
      # longer implicitly allreduce gradients, users manually allreduce gradient
      # and passed the allreduced grads_and_vars. For now, the
      # clip_by_global_norm will be moved to before the explicit allreduce to
      # keep the math the same as TF 1 and pre TF 2.2 implementation.
      (grads, _) = tf.clip_by_global_norm(grads, clip_norm=1.0)
    return super(AdamWeightDecay, self).apply_gradients(
        zip(grads, tvars),
        name=name,
        experimental_aggregate_gradients=experimental_aggregate_gradients)

  def _get_lr(self, var_device, var_dtype, apply_state):
    """Retrieves the learning rate with the given state."""
    if apply_state is None:
      return self._decayed_lr_t[var_dtype], {}

    apply_state = apply_state or {}
    coefficients = apply_state.get((var_device, var_dtype))
    if coefficients is None:
      coefficients = self._fallback_apply_state(var_device, var_dtype)
      apply_state[(var_device, var_dtype)] = coefficients

    return coefficients['lr_t'], dict(apply_state=apply_state)

  def _resource_apply_dense(self, grad, var, apply_state=None):
    # As the weight decay doesn't take any tensors from forward pass as inputs,
    # add a control dependency here to make sure it happens strictly in the
    # backward pass.
    # TODO(b/171088214): Remove it after the control dependency in
    # nested function is fixed.
    with tf.control_dependencies([tf.identity(grad)]):
      lr_t, kwargs = self._get_lr(var.device, var.dtype.base_dtype, apply_state)
      decay = self._decay_weights_op(var, lr_t, apply_state)
    with tf.control_dependencies([decay]):
      return super(AdamWeightDecay,
                   self)._resource_apply_dense(grad, var, **kwargs)

  def _resource_apply_sparse(self, grad, var, indices, apply_state=None):
    # As the weight decay doesn't take any tensors from forward pass as inputs,
    # add a control dependency here to make sure it happens strictly in the
    # backward pass.
    # TODO(b/171088214): Remove it after the control dependency in
    # nested function is fixed.
    with tf.control_dependencies([tf.identity(grad)]):
      lr_t, kwargs = self._get_lr(var.device, var.dtype.base_dtype, apply_state)
      decay = self._decay_weights_op(var, lr_t, apply_state)
    with tf.control_dependencies([decay]):
      return super(AdamWeightDecay,
                   self)._resource_apply_sparse(grad, var, indices, **kwargs)

  def get_config(self):
    config = super(AdamWeightDecay, self).get_config()
    config.update({
        'weight_decay_rate': self.weight_decay_rate,
    })
    return config

  def _do_use_weight_decay(self, param_name):
    """Whether to use L2 weight decay for `param_name`."""
    if self.weight_decay_rate == 0:
      return False

    if self._include_in_weight_decay:
      for r in self._include_in_weight_decay:
        if re.search(r, param_name) is not None:
          return True

    if self._exclude_from_weight_decay:
      for r in self._exclude_from_weight_decay:
        if re.search(r, param_name) is not None:
          return False
    return True