tensorflow-radam-0.15.0.tar.gz

# Keras RAdam [](https://travis-ci.org/CyberZHG/keras-radam) [](https://coveralls.io/github/CyberZHG/keras-radam) [](https://pypi.org/project/keras-rectified-adam/)        \[[中文](https://github.com/CyberZHG/keras-radam/blob/master/README.zh-CN.md)|[English](https://github.com/CyberZHG/keras-radam/blob/master/README.md)\] Unofficial implementation of [RAdam](https://arxiv.org/pdf/1908.03265v1.pdf) in Keras and TensorFlow. ## Install ```bash pip install keras-rectified-adam ``` ## External Link - [tensorflow/addons:RectifiedAdam](https://github.com/tensorflow/addons/blob/master/tensorflow_addons/optimizers/rectified_adam.py) ## Usage ```python import keras import numpy as np from keras_radam import RAdam # Build toy model with RAdam optimizer model = keras.models.Sequential() model.add(keras.layers.Dense(input_shape=(17,), units=3)) model.compile(RAdam(), loss='mse') # Generate toy data x = np.random.standard_normal((4096 * 30, 17)) w = np.random.standard_normal((17, 3)) y = np.dot(x, w) # Fit model.fit(x, y, epochs=5) ``` ### TensorFlow without Keras ```python from keras_radam.training import RAdamOptimizer RAdamOptimizer(learning_rate=1e-3) ``` ### Use Warmup ```python from keras_radam import RAdam RAdam(total_steps=10000, warmup_proportion=0.1, min_lr=1e-5) ``` ## Q & A ### About Correctness The optimizer produces similar losses and weights to the official optimizer after 500 steps. ### Use `tf.keras` or `tf-2.0` Add `TF_KERAS=1` to environment variables to use `tensorflow.python.keras`. ### Use `theano` Backend Add `KERAS_BACKEND=theano` to environment variables to enable `theano` backend.