63 lines
2.5 KiB
Python
63 lines
2.5 KiB
Python
#!/usr/bin/env python
|
|
# -*- coding: utf-8 -*-
|
|
|
|
# ==============================================================================
|
|
# Copyright 2018 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.
|
|
# ==============================================================================
|
|
|
|
import tensorflow as tf
|
|
from dllogger.logger import LOGGER
|
|
|
|
__all__ = ['learning_rate_scheduler']
|
|
|
|
|
|
def learning_rate_scheduler(lr_init, lr_warmup_epochs, global_step, batch_size,
|
|
num_batches_per_epoch, num_decay_steps, num_gpus, use_cosine_lr):
|
|
|
|
def get_scaled_base_learning_rate():
|
|
"""Calculates base learning rate for creating lr schedule.
|
|
In replicated mode, gradients are summed rather than averaged which, with
|
|
the sgd and momentum optimizers, increases the effective learning rate by
|
|
lr * num_gpus. Dividing the base lr by num_gpus negates the increase.
|
|
Args:
|
|
batch_size: Total batch-size.
|
|
Returns:
|
|
Base learning rate to use to create lr schedule.
|
|
"""
|
|
|
|
base_lr = lr_init * num_gpus
|
|
|
|
# Starting LR = 0.1 with BS = 256, else linearly scale
|
|
return base_lr * (batch_size / 256.0)
|
|
|
|
rescaled_lr = get_scaled_base_learning_rate()
|
|
|
|
if use_cosine_lr:
|
|
LOGGER.log("Using cosine learning rate schedule")
|
|
lr = tf.train.cosine_decay(rescaled_lr, global_step, num_decay_steps)
|
|
|
|
else:
|
|
LOGGER.log("Using step learning rate schedule")
|
|
boundaries = [int(num_batches_per_epoch * x) for x in [30, 60, 80, 90]]
|
|
|
|
values = [1e0, 1e-1, 1e-2, 1e-3, 1e-4]
|
|
values = [rescaled_lr * v for v in values]
|
|
|
|
lr = tf.train.piecewise_constant(global_step, boundaries, values)
|
|
|
|
warmup_steps = int(num_batches_per_epoch * lr_warmup_epochs)
|
|
warmup_lr = (rescaled_lr * tf.cast(global_step, tf.float32) / tf.cast(warmup_steps, tf.float32))
|
|
|
|
return tf.cond(global_step < warmup_steps, lambda: warmup_lr, lambda: lr)
|