tf-keras, make the average of model weight in same models
Project description
modelaverage
`modelaverage` is a pip package which make the average weight of model weight in `same models`, inspired by [Average weights in keras models](https://stackoverflow.com/questions/48212110/average-weights-in-keras-models). I created this pip package to use distributed computing environment like `kubernetes`.
Usage
pip install modelaverage
average(model, modellist)
- model : Tensorflow-Keras model
- modellist : list of model file names.
- return : averaged weight model
Example
import tensorflow as tf
from modelaverage import average
(x_train,y_train), (x_test,y_test) = tf.keras.datasets.mnist.load_data()
x_train = tf.keras.utils.normalize(x_train,axis=1)
x_test = tf.keras.utils.normalize(x_test,axis=1)
model = tf.keras.models.Sequential()
model.add(tf.keras.layers.Flatten())
model.add(tf.keras.layers.Dense(128, activation=tf.nn.relu))
model.add(tf.keras.layers.Dense(128, activation=tf.nn.relu))
model.add(tf.keras.layers.Dense(10, activation=tf.nn.softmax))
modellist = ['models/mnist1.h5', 'models/mnist2.h5', 'models/mnist3.h5', 'models/mnist4.h5', 'models/mnist5.h5',
'models/mnist6.h5', 'models/mnist7.h5', 'models/mnist8.h5', 'models/mnist9.h5']
averaged_model = average(model, modellist)
for w in averaged_model.get_weights():
print(w.shape)
Author
- Name : Tae Hwan Jung(@graykode)
- Email : nlkey2022@gmail.com
Project details
Release history Release notifications | RSS feed
Download files
Download the file for your platform. If you're not sure which to choose, learn more about installing packages.
Source Distribution
modelaverage-1.0.0.tar.gz
(2.2 kB
view hashes)
Built Distribution
modelaverage-1.0.0-py3.6.egg
(3.8 kB
view hashes)