# coding=utf-8 # Copyright 2020 The Google Research Authors. # # 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). Modified from the original BERT code to allow for having separate learning rates for different layers of the network. """ from __future__ import absolute_import from __future__ import division from __future__ import print_function import collections import re import tensorflow as tf def create_optimizer( loss, learning_rate, num_train_steps, weight_decay_rate=0.0, use_tpu=False, warmup_steps=0, warmup_proportion=0, lr_decay_power=1.0, layerwise_lr_decay_power=-1, n_transformer_layers=None): """Creates an optimizer and training op.""" global_step = tf.train.get_or_create_global_step() learning_rate = tf.train.polynomial_decay( learning_rate, global_step, num_train_steps, end_learning_rate=0.0, power=lr_decay_power, cycle=False) warmup_steps = max(num_train_steps * warmup_proportion, warmup_steps) learning_rate *= tf.minimum( 1.0, tf.cast(global_step, tf.float32) / tf.cast(warmup_steps, tf.float32)) if layerwise_lr_decay_power > 0: learning_rate = _get_layer_lrs(learning_rate, layerwise_lr_decay_power, n_transformer_layers) optimizer = AdamWeightDecayOptimizer( learning_rate=learning_rate, weight_decay_rate=weight_decay_rate, beta_1=0.9, beta_2=0.999, epsilon=1e-6, exclude_from_weight_decay=["LayerNorm", "layer_norm", "bias"]) if use_tpu: optimizer = tf.tpu.CrossShardOptimizer(optimizer) tvars = tf.trainable_variables() grads = tf.gradients(loss, tvars) (grads, _) = tf.clip_by_global_norm(grads, clip_norm=1.0) train_op = optimizer.apply_gradients( zip(grads, tvars), global_step=global_step) new_global_step = global_step + 1 train_op = tf.group(train_op, [global_step.assign(new_global_step)]) return train_op class AdamWeightDecayOptimizer(tf.train.Optimizer): """A basic Adam optimizer that includes "correct" L2 weight decay.""" def __init__(self, learning_rate, weight_decay_rate=0.0, beta_1=0.9, beta_2=0.999, epsilon=1e-6, exclude_from_weight_decay=None, name="AdamWeightDecayOptimizer"): """Constructs a AdamWeightDecayOptimizer.""" super(AdamWeightDecayOptimizer, self).__init__(False, name) self.learning_rate = learning_rate self.weight_decay_rate = weight_decay_rate self.beta_1 = beta_1 self.beta_2 = beta_2 self.epsilon = epsilon self.exclude_from_weight_decay = exclude_from_weight_decay def _apply_gradients(self, grads_and_vars, learning_rate): """See base class.""" assignments = [] for (grad, param) in grads_and_vars: if grad is None or param is None: continue param_name = self._get_variable_name(param.name) m = tf.get_variable( name=param_name + "/adam_m", shape=param.shape.as_list(), dtype=tf.float32, trainable=False, initializer=tf.zeros_initializer()) v = tf.get_variable( name=param_name + "/adam_v", shape=param.shape.as_list(), dtype=tf.float32, trainable=False, initializer=tf.zeros_initializer()) # Standard Adam update. next_m = ( tf.multiply(self.beta_1, m) + tf.multiply(1.0 - self.beta_1, grad)) next_v = ( tf.multiply(self.beta_2, v) + tf.multiply(1.0 - self.beta_2, tf.square(grad))) update = next_m / (tf.sqrt(next_v) + self.epsilon) # 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. if self.weight_decay_rate > 0: if self._do_use_weight_decay(param_name): update += self.weight_decay_rate * param update_with_lr = learning_rate * update next_param = param - update_with_lr assignments.extend( [param.assign(next_param), m.assign(next_m), v.assign(next_v)]) return assignments def apply_gradients(self, grads_and_vars, global_step=None, name=None): if isinstance(self.learning_rate, dict): key_to_grads_and_vars = {} for grad, var in grads_and_vars: update_for_var = False for key in self.learning_rate: if key in var.name: update_for_var = True if key not in key_to_grads_and_vars: key_to_grads_and_vars[key] = [] key_to_grads_and_vars[key].append((grad, var)) if not update_for_var: raise ValueError("No learning rate specified for variable", var) assignments = [] for key, key_grads_and_vars in key_to_grads_and_vars.items(): assignments += self._apply_gradients(key_grads_and_vars, self.learning_rate[key]) else: assignments = self._apply_gradients(grads_and_vars, self.learning_rate) return tf.group(*assignments, name=name) def _do_use_weight_decay(self, param_name): """Whether to use L2 weight decay for `param_name`.""" if not self.weight_decay_rate: return False 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 def _get_variable_name(self, param_name): """Get the variable name from the tensor name.""" m = re.match("^(.*):\\d+$", param_name) if m is not None: param_name = m.group(1) return param_name def _get_layer_lrs(learning_rate, layer_decay, n_layers): """Have lower learning rates for layers closer to the input.""" key_to_depths = collections.OrderedDict({ "/embeddings/": 0, "/embeddings_project/": 0, "task_specific/": n_layers + 2, }) for layer in range(n_layers): key_to_depths["encoder/layer_" + str(layer) + "/"] = layer + 1 return { key: learning_rate * (layer_decay ** (n_layers + 2 - depth)) for key, depth in key_to_depths.items() }