TensorLight - A high-level framework for TensorFlow

Overview
TensorLight

TensorLight is a high-level framework for TensorFlow-based machine intelligence applications. It reduces boilerplate code and enables advanced features that are not yet provided out-of-the-box.

Setup

After cloning the repository, we can install the package locally (for use on our system), with:

$ cd /path/to/tensorlight
$ sudo pip install .

We can also install the package with a symlink, so that changes to the source files will be immediately available to other users of the package on our system:

$ sudo pip install -e .

Guiding Principles

The TensorLight framework is developed under its four core principles:

  • Simplicity: Straight-forward to use for anybody who has already worked with TensorFlow. Especially, no further learning is required regarding how to define a model's graph definition.
  • Compactness: Reduce boilerplate code, while keeping the transparency and flexibility of TensorFlow.
  • Standardization: Provide a standard way in respect to the implementation of models and datasets in order to save time. Further, it automates the whole training and validation process, but also provides hooks to maintain customizability.
  • Superiority: Enable advanced features that are not included in the TensorFlow API, as well as retain its full functionality.

Key Features

To highlight the advanced features of TensorLight, an incomplete list of some main functionalities is provided that are not shipped with TensorFlow by default, or might even be missing in other high-level APIs. These include:

  • Transparent lifecycle management of the session and graph definition.
  • Abstraction of models and datasets to provide a reusable plug-and-play support.
  • Effortless support to train a model symmetrically on multiple GPUs, as well as prevent TensorFlow to allocate memory on other GPU devices of the cluster.
  • Train or evaluate a model with a single line of code.
  • Abstracted, runtime-exchangeable input pipelines which either use the simple feeding mechanism with NumPy arrays, or even multi-threaded input queues.
  • Automatic saving and loading of hyperparameters as JSON to simplify the evaluation management of numerous trainings.
  • Ready-to-use loss functions and metrics, even with latest advances for perceptual motivated image similarity assessment.
  • Extended recurrent functions to enable scheduled sampling, as well as an implementation of a ConvLSTM cell.
  • Automatic creation of periodic checkpoints and TensorBoard summaries.
  • Ability to work with other higher-level libraries hand in hand, such as tf.contrib or TF-slim.

Architecture

From an architectural perspective, the framework can be split into three main components. First, a collection of utility function that are unrelated to machine learning. Examples are functions to download and extract datasets, to process images and videos, or to generate animated GIFs and videos from a data array, to name just a few. Second, the high-level library which builds on top of TensorFlow. It includes several modules that either provide a simple access to functionally that it repeatedly required when developing deep learning applications, or features that are not included in TensorFlow yet. For instance, it handles the creation of weight and bias variables internally, offers a bunch of ready-to-use loss and initialization functions, or comes with some advanced visualization features to display feature maps or output images directly in an IPython Notebook. Third, an abstraction layer to simplify the overall lifecycle, to generalize the definition of a model graphs, as well as to enable a reusable and consistent access to datasets.

TensorLight Architecture

The user program can either exploit the high-level library and the provided utility functions for his existing projects, or take advantage from TensorLight's abstraction layes while creating new deep learning applications. The latter enables to radically reduce the amount of code that has to be written for training or evaluating the model. This is realized by encapsulating the lifecycle of TensorFlow's session, graph, summary-writer or checkpoint-saver, as well as the entire training or evaluation loop within a runtime module.

Examples

You want to learn more? Check out the tutorial and code examples.

Owner
Benjamin Kan
Passionate coder with focus on machine learning, mobile apps and game development
Benjamin Kan
A graph neural network (GNN) model to predict protein-protein interactions (PPI) with no sample features

A graph neural network (GNN) model to predict protein-protein interactions (PPI) with no sample features

2 Jul 25, 2022
RANZCR-CLiP 7th Place Solution

RANZCR-CLiP 7th Place Solution This repository is WIP. (18 Mar 2021) Installation git clone https://github.com/analokmaus/kaggle-ranzcr-clip-public.gi

Hiroshechka Y 21 Oct 22, 2022
A library of scripts that interact with the PythonTurtle module to create games, drawings, and more

TurtleLib TurtleLib is a library of scripts that interact with the PythonTurtle module to create games, drawings, and more! Using the Scripts Copy or

1 Jan 15, 2022
Python package to generate image embeddings with CLIP without PyTorch/TensorFlow

imgbeddings A Python package to generate embedding vectors from images, using OpenAI's robust CLIP model via Hugging Face transformers. These image em

Max Woolf 81 Jan 04, 2023
Learning to Reach Goals via Iterated Supervised Learning

Vanilla GCSL This repository contains a vanilla implementation of "Learning to Reach Goals via Iterated Supervised Learning" proposed by Dibya Gosh et

Christoph Heindl 4 Aug 10, 2022
Deep learning with TensorFlow and earth observation data.

Deep Learning with TensorFlow and EO Data Complete file set for Jupyter Book Autor: Development Seed Date: 04 October 2021 ISBN: (to come) Notebook tu

Development Seed 20 Nov 16, 2022
SMD-Nets: Stereo Mixture Density Networks

SMD-Nets: Stereo Mixture Density Networks This repository contains a Pytorch implementation of "SMD-Nets: Stereo Mixture Density Networks" (CVPR 2021)

Fabio Tosi 115 Dec 26, 2022
OMNIVORE is a single vision model for many different visual modalities

Omnivore: A Single Model for Many Visual Modalities [paper][website] OMNIVORE is a single vision model for many different visual modalities. It learns

Meta Research 451 Dec 27, 2022
I-BERT: Integer-only BERT Quantization

I-BERT: Integer-only BERT Quantization HuggingFace Implementation I-BERT is also available in the master branch of HuggingFace! Visit the following li

Sehoon Kim 139 Dec 27, 2022
Yolo object detection - Yolo object detection with python

How to run download required files make build_image make download Docker versio

3 Jan 26, 2022
Implementation of DocFormer: End-to-End Transformer for Document Understanding, a multi-modal transformer based architecture for the task of Visual Document Understanding (VDU)

DocFormer - PyTorch Implementation of DocFormer: End-to-End Transformer for Document Understanding, a multi-modal transformer based architecture for t

171 Jan 06, 2023
FuseDream: Training-Free Text-to-Image Generationwith Improved CLIP+GAN Space OptimizationFuseDream: Training-Free Text-to-Image Generationwith Improved CLIP+GAN Space Optimization

FuseDream This repo contains code for our paper (paper link): FuseDream: Training-Free Text-to-Image Generation with Improved CLIP+GAN Space Optimizat

XCL 191 Dec 31, 2022
N-HiTS: Neural Hierarchical Interpolation for Time Series Forecasting

N-HiTS: Neural Hierarchical Interpolation for Time Series Forecasting Recent progress in neural forecasting instigated significant improvements in the

Cristian Challu 82 Jan 04, 2023
LLVIP: A Visible-infrared Paired Dataset for Low-light Vision

LLVIP: A Visible-infrared Paired Dataset for Low-light Vision Project | Arxiv | Abstract It is very challenging for various visual tasks such as image

CVSM Group - email: <a href=[email protected]"> 377 Jan 07, 2023
Unsupervised Feature Ranking via Attribute Networks.

FRANe Unsupervised Feature Ranking via Attribute Networks (FRANe) converts a dataset into a network (graph) with nodes that correspond to the features

7 Sep 29, 2022
A project which aims to protect your privacy using inexpensive hardware and easily modifiable software

Protecting your privacy using an ESP32, an IR sensor and a python script This project, which I personally call the "never-gonna-catch-me-in-the-act-ev

8 Oct 10, 2022
UniFormer - official implementation of UniFormer

UniFormer This repo is the official implementation of "Uniformer: Unified Transformer for Efficient Spatiotemporal Representation Learning". It curren

SenseTime X-Lab 573 Jan 04, 2023
Pytorch Implementation of "Desigining Network Design Spaces", Radosavovic et al. CVPR 2020.

RegNet Pytorch Implementation of "Desigining Network Design Spaces", Radosavovic et al. CVPR 2020. Paper | Official Implementation RegNet offer a very

Vishal R 2 Feb 11, 2022
Unsupervised Real-World Super-Resolution: A Domain Adaptation Perspective

Unofficial pytorch implementation of the paper "Unsupervised Real-World Super-Resolution: A Domain Adaptation Perspective"

16 Nov 21, 2022
Deep Learning GPU Training System

DIGITS DIGITS (the Deep Learning GPU Training System) is a webapp for training deep learning models. The currently supported frameworks are: Caffe, To

NVIDIA Corporation 4.1k Jan 03, 2023