Implementation of Supervised Contrastive Learning with AMP, EMA, SWA, and many other tricks

Overview

SupCon-Framework

The repo is an implementation of Supervised Contrastive Learning. It's based on another implementation, but with several differencies:

  • Fixed bugs (incorrect ResNet implementations, which leads to a very small max batch size),
  • Offers a lot of additional functionality (first of all, rich validation).

To be more precise, in this implementations you will find:

  • Augmentations with albumentations
  • Hyperparameters are moved to .yml configs
  • t-SNE visualizations
  • 2-step validation (for features before and after the projection head) using metrics like AMI, NMI, mAP, precision_at_1, etc with PyTorch Metric Learning.
  • Exponential Moving Average for a more stable training, and Stochastic Moving Average for a better generalization and just overall performance.
  • Automatic Mixed Precision (torch version) training in order to be able to train with a bigger batch size (roughly by a factor of 2).
  • LabelSmoothing loss, and LRFinder for the second stage of the training (FC).
  • TensorBoard logs, checkpoints
  • Support of timm models, and pytorch-optimizer

Install

  1. Clone the repo:
git clone https://github.com/ivanpanshin/SupCon-Framework && cd SupCon-Framework/
  1. Create a clean virtual environment
python3 -m venv venv
source venv/bin/activate
  1. Install dependencies
python -m pip install --upgrade pip
pip install -r requirements.txt

Training

In order to execute Cifar10 training run:

python train.py --config_name configs/train/train_supcon_resnet18_cifar10_stage1.yml
python swa.py --config_name configs/train/swa_supcon_resnet18_cifar10_stage1.yml
python train.py --config_name configs/train/train_supcon_resnet18_cifar10_stage2.yml
python swa.py --config_name configs/train/swa_supcon_resnet18_cifar10_stage2.yml

In order to run LRFinder on the second stage of the training, run:

python learning_rate_finder.py --config_name configs/train/lr_finder_supcon_resnet18_cifar10_stage2.yml

The process of training Cifar100 is exactly the same, just change config names from cifar10 to cifar100.

After that you can check the results of the training either in logs or runs directory. For example, in order to check tensorboard logs for the first stage of Cifar10 training, run:

tensorboard --logdir runs/supcon_first_stage_cifar10

Visualizations

This repo is supplied with t-SNE visualizations so that you can check embeddings you get after the training. Check t-SNE.ipynb for details.

Those are t-SNE visualizations for Cifar10 for validation and train with SupCon (top), and validation and train with CE (bottom).

Those are t-SNE visualizations for Cifar100 for validation and train with SupCon (top), and validation and train with CE (bottom).

Results

Model Stage Dataset Accuracy
ResNet18 Frist CIFAR10 95.9
ResNet18 Second CIFAR10 94.9
ResNet18 Frist CIFAR100 79.0
ResNet18 Second CIFAR100 77.9

Note that even though the accuracy on the second stage is lower, it's not always the case. In my experience, the difference between stages is usually around 1 percent, including the difference that favors the second stage.

Training time for the whole pipeline (without any early stopping) on CIFAR10 or CIFAR100 is around 4 hours (single 2080Ti with AMP). However, with reasonable early stopping that value goes down to around 2.5-3 hours.

Custom datasets

It's fairly easy to adapt this pipeline to custom datasets. First, you need to check tools/datasets.py for that. Second, add a new class for your dataset. The only guideline here is to follow the same augmentation logic, that is

        if self.second_stage:
            image = self.transform(image=image)['image']
        else:
            image = self.transform(image)

Third, add your dataset to DATASETS dict still inside tools/datasets.py, and you're good to go.

FAQ

  • Q: What hyperparameters I should try to change?

    A: First of all, learning rate. Second of all, try to change the augmentation policy. SupCon is build around "cropping + color jittering" scheme, so you can try changing the cropping size or the intensity of jittering. Check tools.utils.build_transforms for that.

  • Q: What backbone and batch size should I use?

    A: This is quite simple. Take the biggest backbone you can, and after that take the highest batch size your GPU can offer. The reason for that: SupCon is more prone (than regular classification training with CE/LabelSmoothing/etc) to improving with stronger backbones. Moverover, it has a property of explicit hard positive and negative mining. It means that the higher the batch size - the more difficult and helpful samples you supply to your model.

  • Q: Do I need the second stage of the training?

    A: Not necessarily. You can do classification based only on embeddings. In order to do that compute embeddings for the train set, and at inference time do the following: take a sample, compute its embedding, take the closest one from the training, take its class. To make this fast and efficient, you something like faiss for similarity search. Note that this is actually how validation is done in this repo. Moveover, during training you will see a metric precision_at_1. This is actually just accuracy based solely on embeddings.

  • Q: Should I use AMP?

    A: If your GPU has tensor cores (like 2080Ti) - yes. If it doesn't (like 1080Ti) - check the speed with AMP and without. If the speed dropped slightly (or even increased by a bit) - use it, since SupCon works better with bigger batch sizes.

  • Q: How should I use EMA?

    A: You only need to choose the ema_decay_per_epoch parameter in the config. The heuristic is fairly simple. If your dataset is big, then something as small as 0.3 will do just fine. And as your dataset gets smaller, you can increase ema_decay_per_epoch. Thanks to bonlime for this idea. I advice you to check his great pytorch tools repo, it's a hidden gem.

  • Q: Is it better than training with Cross Entropy/Label Smoothing/etc?

    A: Unfortunately, in my experience, it's much easier to get better results with something like CE. It's more stable, faster to train, and simply produces better or the same results. For instance, in case on CIFAR10/100 it's trivial to train ResNet18 up tp 96/81 percent respectively. Of cource, I've seen cased where SupCon performs better, but it takes quite a bit of work to make it outperform CE.

  • Q: How long should I train with SupCon?

    A: The answer is tricky. On one hand, authors of the original paper claim that the longer you train with SupCon, the better it gets. However, I did not observe such a behavior in my tests. So the only recommendation I can give is the following: start with 100 epochs for easy datasets (like CIFAR10/100), and 1000 for more industrial ones. Then - monitor the training process. If the validaton metric (such as precision_at_1) doesn't impove for several dozens of epochs - you can stop the training. You might incorporate early stopping for this reason into the pipeline.

Owner
Ivan Panshin
Machine Learning Engineer: CV, NLP, tabular data. Kaggle (top 0.003% worldwide) and Open Source
Ivan Panshin
Foundation Auth Proxy is an abstraction on Foundations' authentication layer and is used to authenticate requests to Atlas's REST API.

foundations-auth-proxy Setup By default the server runs on http://0.0.0.0:5558. This can be changed via the arguments. Arguments: '-H' or '--host': ho

Dessa - Open Source 2 Jul 03, 2020
Corsair_scan is a security tool to test Cross-Origin Resource Sharing (CORS).

Welcome to Corsair_scan Corsair_scan is a security tool to test Cross-Origin Resource Sharing (CORS) misconfigurations. CORS is a mechanism that allow

Santander Security Research 116 Nov 09, 2022
API with high performance to create a simple blog and Auth using OAuth2 ⛏

DogeAPI API with high performance built with FastAPI & SQLAlchemy, help to improve connection with your Backend Side to create a simple blog and Cruds

Yasser Tahiri 111 Jan 05, 2023
API-key based security utilities for FastAPI, focused on simplicity of use

FastAPI simple security API key based security package for FastAPI, focused on simplicity of use: Full functionality out of the box, no configuration

Tolki 154 Jan 03, 2023
Phishing Abusing Microsoft 365 OAuth Authorization Flow

Microsoft365_devicePhish Abusing Microsoft 365 OAuth Authorization Flow for Phishing Attack This is a simple proof-of-concept script that allows an at

bigb0ss 11 Dec 11, 2022
Social auth made simple

Python Social Auth Python Social Auth is an easy-to-setup social authentication/registration mechanism with support for several frameworks and auth pr

MatΓ­as Aguirre 2.8k Dec 24, 2022
JWT Key Confusion PoC (CVE-2015-9235) Written for the Hack the Box challenge - Under Construction

JWT Key Confusion PoC (CVE-2015-9235) Written for the Hack the Box challenge - Under Construction This script performs a Java Web Token Key Confusion

Alex Fronteddu 1 Jan 13, 2022
REST implementation of Django authentication system.

djoser REST implementation of Django authentication system. djoser library provides a set of Django Rest Framework views to handle basic actions such

Sunscrapers 2.2k Jan 01, 2023
A full Rest-API With Oauth2 and JWT for request & response a JSON file Using FastAPI and SQLAlchemy πŸ”‘

Pexon-Rest-API A full Rest-API for request & response a JSON file, Building a Simple WorkFlow that help you to Request a JSON File Format and Handling

Yasser Tahiri 15 Jul 22, 2022
Authentication for Django Rest Framework

Dj-Rest-Auth Drop-in API endpoints for handling authentication securely in Django Rest Framework. Works especially well with SPAs (e.g React, Vue, Ang

Michael 1.1k Jan 03, 2023
Simple yet powerful authorization / authentication client library for Python web applications.

Authomatic Authomatic is a framework agnostic library for Python web applications with a minimalistic but powerful interface which simplifies authenti

1k Dec 28, 2022
CheckList-Api - Created with django rest framework and JWT(Json Web Tokens for Authentication)

CheckList Api created with django rest framework and JWT(Json Web Tokens for Aut

shantanu nimkar 1 Jan 24, 2022
Integrated set of Django applications addressing authentication, registration, account management as well as 3rd party (social) account authentication.

Welcome to django-allauth! Integrated set of Django applications addressing authentication, registration, account management as well as 3rd party (soc

Raymond Penners 7.7k Jan 03, 2023
Easy and secure implementation of Azure AD for your FastAPI APIs πŸ”’ Single- and multi-tenant support.

Easy and secure implementation of Azure AD for your FastAPI APIs πŸ”’ Single- and multi-tenant support.

Intility 220 Jan 05, 2023
JSON Web Token Authentication support for Django REST Framework

REST framework JWT Auth JSON Web Token Authentication support for Django REST Framework Overview This package provides JSON Web Token Authentication s

Styria Digital Development 178 Jan 02, 2023
Easy and secure implementation of Azure AD for your FastAPI APIs πŸ”’ Single- and multi-tenant support.

Easy and secure implementation of Azure AD for your FastAPI APIs πŸ”’ Single- and multi-tenant support.

Intility 220 Jan 05, 2023
A generic, spec-compliant, thorough implementation of the OAuth request-signing logic

OAuthLib - Python Framework for OAuth1 & OAuth2 *A generic, spec-compliant, thorough implementation of the OAuth request-signing logic for Python 3.5+

OAuthlib 2.5k Jan 01, 2023
A secure authentication module to validate user credentials in a Streamlit application.

Streamlit-Authenticator A secure authentication module to validate user credentials in a Streamlit application. Installation Streamlit-Authenticator i

M Khorasani 336 Dec 31, 2022
Simple two factor authemtication system, made by me.

Simple two factor authemtication system, made by me. Honestly, i don't even know How 2FAs work I just used my knowledge and did whatever i could. Send

Refined 5 Jan 04, 2022
Flask user session management.

Flask-Login Flask-Login provides user session management for Flask. It handles the common tasks of logging in, logging out, and remembering your users

Max Countryman 3.2k Dec 28, 2022