Skip to content

Latest commit

 

History

History
53 lines (44 loc) · 2.55 KB

README.md

File metadata and controls

53 lines (44 loc) · 2.55 KB

Addons - Optimizers

Maintainers

Submodule Maintainers Contact Info
conditional_gradient Pengyu Kan, Vishnu Lokhande [email protected], [email protected]
cyclical_learning_rate  Raphael Meudec [email protected]
lamb Jing Li, Junjie Ke [email protected], [email protected]
lazy_adam Saishruthi Swaminathan [email protected]
lookahead Zhao Hanguang [email protected]
moving_average Dheeraj R. Reddy [email protected]
novograd Shreyash Patodia [email protected]
rectified_adam Zhao Hanguang [email protected]
stochastic_weight_averaging Shreyash Patodia [email protected]
weight_decay_optimizers Phil Jund [email protected]
yogi Manzil Zaheer [email protected]

Components

Submodule Optimizer Reference
conditional_gradient ConditionalGradient https://arxiv.org/pdf/1803.06453.pdf
cyclical_learning_rate Cyclical Learning Rate https://arxiv.org/abs/1506.01186
lamb LAMB https://arxiv.org/abs/1904.00962
lazy_adam LazyAdam https://arxiv.org/abs/1412.6980
lookahead Lookahead https://arxiv.org/abs/1907.08610v1
moving_average MovingAverage
novograd NovoGrad https://nvidia.github.io/OpenSeq2Seq/html/optimizers.html
rectified_adam RectifiedAdam https://arxiv.org/pdf/1908.03265v1.pdf
stochastic_weight_averaging SWA https://arxiv.org/abs/1803.05407.pdf
weight_decay_optimizers SGDW, AdamW, extend_with_decoupled_weight_decay https://arxiv.org/pdf/1711.05101.pdf
yogi Yogi https://papers.nips.cc/paper/8186-adaptive-methods-for-nonconvex-optimization.pdf

Contribution Guidelines

Standard API

In order to conform with the current API standard, all optimizers must:

  • Inherit from either keras.optimizer_v2.OptimizerV2 or its subclasses.
  • Register as a keras global object so it can be serialized properly: @tf.keras.utils.register_keras_serializable(package='Addons')
  • Add the addon to the py_library in this sub-package's BUILD file.

Testing Requirements

  • When applicable, run all tests with TensorFlow's @run_in_graph_and_eager_modes (for test method) or run_all_in_graph_and_eager_modes (for TestCase subclass) decorator.
  • Add a py_test to this sub-package's BUILD file.

Documentation Requirements

  • Update the table of contents in this sub-packages's README.