Skip to content
A PyTorch implementation for exploring deep and shallow knowledge distillation (KD) experiments with flexibility
Branch: master
Clone or download
peterliht Merge pull request #3 from akarle/master
Adding note about Issue #2 in KD Loss
Latest commit e4c4013 May 8, 2018
Permalink
Type Name Latest commit message Commit time
Failed to load latest commit information.
experiments polish code comments; cleanup ./experiments folder; modify readme; up… Mar 20, 2018
mnist earlier simple experiments with mnist Mar 20, 2018
model Adding note about #2 in KD Loss Apr 29, 2018
.gitignore
LICENSE Initial commit Mar 9, 2018
README.md update readme Apr 20, 2018
count_model_size.py dev cleanup and update, not final version yet Mar 19, 2018
distillation_analysis.py distill analysis Mar 17, 2018
evaluate.py fix redundancy Mar 20, 2018
requirements.txt update README and requirements Mar 27, 2018
search_hyperparams.py polish code comments; cleanup ./experiments folder; modify readme; up… Mar 20, 2018
synthesize_results.py first commit Mar 10, 2018
train.py polish code comments; cleanup ./experiments folder; modify readme; up… Mar 20, 2018
utils.py polish code comments; cleanup ./experiments folder; modify readme; up… Mar 20, 2018

README.md

knowledge-distillation-pytorch

  • Exploring knowledge distillation of DNNs for efficient hardware solutions
  • Author: Haitong Li
  • Framework: PyTorch
  • Dataset: CIFAR-10

Features

  • A framework for exploring "shallow" and "deep" knowledge distillation (KD) experiments
  • Hyperparameters defined by "params.json" universally (avoiding long argparser commands)
  • Hyperparameter searching and result synthesizing (as a table)
  • Progress bar, tensorboard support, and checkpoint saving/loading (utils.py)
  • Pretrained teacher models available for download

Install

  • Clone the repo

    git clone https://github.com/peterliht/knowledge-distillation-pytorch.git
    
  • Install the dependencies (including Pytorch)

    pip install -r requirements.txt
    

Organizatoin:

  • ./train.py: main entrance for train/eval with or without KD on CIFAR-10
  • ./experiments/: json files for each experiment; dir for hypersearch
  • ./model/: teacher and student DNNs, knowledge distillation (KD) loss defination, dataloader

Key notes about usage for your experiments:

  • Download the zip file for pretrained teacher model checkpoints from this Box folder
  • Simply move the unzipped subfolders into 'knowledge-distillation-pytorch/experiments/' (replacing the existing ones if necessary; follow the default path naming)
  • Call train.py to start training 5-layer CNN with ResNet-18's dark knowledge, or training ResNet-18 with state-of-the-art deeper models distilled
  • Use search_hyperparams.py for hypersearch
  • Hyperparameters are defined in params.json files universally. Refer to the header of search_hyperparams.py for details

Train (dataset: CIFAR-10)

Note: all the hyperparameters can be found and modified in 'params.json' under 'model_dir'

-- Train a 5-layer CNN with knowledge distilled from a pre-trained ResNet-18 model

python train.py --model_dir experiments/cnn_distill

-- Train a ResNet-18 model with knowledge distilled from a pre-trained ResNext-29 teacher

python train.py --model_dir experiments/resnet18_distill/resnext_teacher

-- Hyperparameter search for a specified experiment ('parent_dir/params.json')

python search_hyperparams.py --parent_dir experiments/cnn_distill_alpha_temp

--Synthesize results of the recent hypersearch experiments

python synthesize_results.py --parent_dir experiments/cnn_distill_alpha_temp

Results: "Shallow" and "Deep" Distillation

Quick takeaways (more details to be added):

  • Knowledge distillation provides regularization for both shallow DNNs and state-of-the-art DNNs
  • Having unlabeled or partial dataset can benefit from dark knowledge of teacher models

-Knowledge distillation from ResNet-18 to 5-layer CNN

Model Dropout = 0.5 No Dropout
5-layer CNN 83.51% 84.74%
5-layer CNN w/ ResNet18 84.49% 85.69%

-Knowledge distillation from deeper models to ResNet-18

Model Test Accuracy
Baseline ResNet-18 94.175%
+ KD WideResNet-28-10 94.333%
+ KD PreResNet-110 94.531%
+ KD DenseNet-100 94.729%
+ KD ResNext-29-8 94.788%

References

H. Li, "Exploring knowledge distillation of Deep neural nets for efficient hardware solutions," CS230 Report, 2018

Hinton, Geoffrey, Oriol Vinyals, and Jeff Dean. "Distilling the knowledge in a neural network." arXiv preprint arXiv:1503.02531 (2015).

Romero, A., Ballas, N., Kahou, S. E., Chassang, A., Gatta, C., & Bengio, Y. (2014). Fitnets: Hints for thin deep nets. arXiv preprint arXiv:1412.6550.

https://github.com/cs230-stanford/cs230-stanford.github.io

https://github.com/bearpaw/pytorch-classification

You can’t perform that action at this time.