tensorflow_batchnorm_folding-1.0.4.tar.gz

# Batch-Normalization Folding In this repository, we propose an implementation of the batch-normalization folding algorithm from [IJCAI 2022](https://arxiv.org/pdf/2203.14646.pdf). Batch-Normalization Folding consists in emoving batch-normalization layers without changing the predictive function defiend by the neural network. The simpliest scenario is an application for a fully-connected layer followed by a batch-normalization layer, we get ```math x \mapsto \gamma \frac{Ax + b - \mu}{\sigma + \epsilon} + \beta = \gamma \frac{A}{\sigma +\epsilon} x + \frac{b - \mu}{\sigma + \epsilon} + \beta ``` Thus the two layers can be expressed as a single fully-connected layer at inference without any change in the predictive function. ## use This repository is available as a pip package (use `pip install tensorflow-batchnorm-folding`). This implementation is compatible with tf.keras.Model instances. It was tested with the following models - [x] ResNet 50 - [x] MobileNet V2 - [x] MobileNet V3 - [x] EfficentNet B0 To run a simple test: ```python from batch_normalization_folding.folder import fold_batchnormalization_layers import tensorflow as tf mod=tf.keras.applications.efficientnet.EfficientNetB0() folded_model,output_str=fold_batchnormalization_layers(mod,True) ``` The `output_str` is either the ratio num_layers_folded/num_layers_not_folded or 'failed' to state a failure in the process. ## To Do - [x] unit test on all keras applciations models - [x] check package installement - [ ] deal with Concatenate layers ## cite ``` @inproceedings{yvinec2022fold, title={To Fold or Not to Fold: a Necessary and Sufficient Condition on Batch-Normalization Layers Folding}, author={Yvinec, Edouard and Dapogny, Arnaud and Bailly, Kevin}, journal={IJCAI}, year={2022} } ``` ## Performance on Base Models ``` +------------------------------------+ | ResNet 50 | +------------------------------------+ | BN layers folded | 53 | | BN layers not folded | 0 | +------------------------------------+ | EfficientNet B0 | +------------------------------------+ | BN layers folded | 49 | | BN layers not folded | 0 | +------------------------------------+ | MobileNet V2 | +------------------------------------+ | BN layers folded | 52 | | BN layers not folded | 0 | +------------------------------------+ | MobileNet V3 | +------------------------------------+ | BN layers folded | 34 | | BN layers not folded | 0 | +------------------------------------+ | Inception ResNet V2 | +------------------------------------+ | BN layers folded | 204 | | BN layers not folded | 0 | +------------------------------------+ | Inception V3 | +------------------------------------+ | BN layers folded | 94 | | BN layers not folded | 0 | +------------------------------------+ | NASNet | +------------------------------------+ | BN layers folded | 28 | | BN layers not folded | 164 | +------------------------------------+ | DenseNet 121 | +------------------------------------+ | BN layers folded | 59 | | BN layers not folded | 62 | +------------------------------------+ ```