TensorFlow Hub: A Library for Reusable Machine Learning Modules in TensorFlow

Introducing TensorFlow Hub: Library for Reusable Machine Learning Modules at TensorFlow

One of the most fundamental things in software development is the idea of ​​a store of shared code that is easy to ignore. As programmers, libraries instantly make us more effective. In a sense, they change the process of problem-solving programming. When using the library, we often think of programming in terms of building blocks - or modules - that can be glued together.


How can a library be considered a machine education developer? Of course, in addition to the share to code, we also want to share pre-trend models. Sharing pre-trained models make it possible for developers to optimize for their domain, without access to computer resources or data used to train the model in the original hands. For example, the NASNet train took thousands of GPU-hours. By sharing the weights learned, a model developer can make it easier for others to reuse and build their work.





It's a library idea for machine education developers that inspired the TensorFlow Hub, and we're happy to share it with the community today. TensorFlow Hub TensorFlow is a platform for publishing, locating, and reusing portions of machine learning modules. By a module, we mean a self-contained fraction of a TensorFlow graph, with its weight, which can be reused in other, similar functions. By reusing the module, a developer can train the model using a small dataset, improve generalizations, or simply speed up the training. Let's look at some examples of how to make this concrete.

Image retraining


As a first example, let's look at a technique that you can use to train an image classifier, starting with a small amount of training data. Modern image recognition models have millions of parameters, and of course, training from scratch requires a large amount of labeled data and computing power. Using a technique called image retraining, you can train the model using very little data, and very little computing time. Here's what it looks like in a TensorFlow hub.

The basic idea is to reuse existing image recognition modules to extract features from your images, and then train new categorizers on these. As you can see above, TensorFlow Hub modules can be instant from the URL (or, from the file system path) when the TensorFlow graph is being created. From there, TensorFlow Hub has a variety of modules that you can choose from, including NASNet, MobileNet (including its latest V2), Inception, ResNet, and more. To use the module you import the TensorFlow Hub, then copy/paste the module's URL into your code.


Some image modules are available at TensorFlow Hub.


Each module has a defined interface that allows it to be used in a modifiable way, with little or no knowledge of its interiors. In this case, this module has a method that you can use to retrieve the desired image size. As a developer, you only need to provide the correct size of the group of images, and call modules in them to get feature representation. This module takes care of the pre-processing of images for you so that you can represent the feature in a single step directly from the group of images. From here, you can classify linear models or other types, above all.

In this case, notify that the module we are using is hosted by Google, and has been edited (so that you cannot change the module that you use in your experiments). Modules can be used as a simple Python type to construct a portion of a graphic. Once exported to disk, the modules contain themselves, and others can use it to create code and data, create it and use it for training (of course you can also publish).



Text classification


Let's take a look at the second example. Imagine that you want to train a model to categorize movie reviews as positive or negative, starting with just a small amount of training data (say, several hundred positive and negative movie reviews). Since you have a limited number of examples you decide to pick up a dataset of word embedding, first trained in a very large corpus. Here's how to use a tensor flow hub.


As before, we start by selecting the module. TensorFlow Hub has a variety of text modules for you to explore, including neural network language models in different languages ​​(N, JP, DE, and ES), as well as Word 2vec trained on Wikipedia, and NNLM embedding trained on Google News.


Some text modules shared in TensorFlow Hub.


In this case, we will use the module for word embedding. The code above downloads the module, uses it to preprocess the sentence, then retrieves the embedding for each token. This means you can go directly to your dataset from a sentence that can go to the appropriate format for the classifier in one step. The module works to tokenize sentences, and to handle other logic-like words outside of vocabulary. Both preprocessing logic and embedding are encapsulated in a single module, making it easy to use different datasets of simple word embedding, or different pre-processing strategies, without changing your code.


Text classification tutorial using TensorFlow Hub.


If you want to try it, use this tutorial to take it for a spin and to learn how the TensorFlow Hub module works with TensorFlow Estimators.


Universal Sentence Encoder


We've also shared the new TensorFlow Hub module! Below is an example of the Universal Sentence Encoder used. It is a sentence-level embedding module that is trained in a wide variety of datasets (in other words, "universal"). Some of the good things about it are semantic similarity, custom text classification, and cluster.


This notebook shows how to train a simple binary text classifier above any TF-Hub module.


As with image retraining, the module requires relatively little label data to adapt to your own work. Let's try this in a restaurant review for example.



Other modules

Tensorflow hub is more than image and text classification. On the website, you will also find some modules for progressive GAN and Google Landmark Deep local features.

Ideas

There are some important considerations when using the TensorFlow Hub modules. First, note that the modules contain code that can be turned on. Always use modules from trusted sources. Second, as in all machine learning, fairness is an important consideration. Both examples are large pre-trained datasets we have shown above. When replicating such a dataset, it is important to consider what data it contains (and whether there are any existing biases), and how this may affect the product you are building and its users.

Get started with TensorFlow.js → https://js.tensorflow.org

Comments

Popular posts from this blog

Artificial intelligence (AI) - the ability of a digital computer.

Facebook's name has been changed to 'rebranding'

What is SEO and how to do search engine optimization?

Labels

Social media Facebook of What a and phone on mobile This you are Do smartphone internet IT Android workforce Nepal app your from robot iPhone use Machine Learning for Python will company computer with account can data does password these twitter Apple digital feature Instagram Whatsapp YouTube like machine media not why Tiktok new ChatGPT China an be by free out people search that website without work Future India ML corona features find information make online or public video Elon Musk Microsoft One apps has market million social user users way year Intelligence Laptop US billion education history home protect service videos want Bitcoin Have Machine Learning Future Nepali Now Operators Scientists Wi-Fi Windows chrome code cyber download hacking money network photos tips world Amazon Artificial Intelligence Future Avoid Cryptocurrency Here If Know Learning TV Things artificial battery being browser human malware many need netflix photo security smart software study system there update which 10 15 Beginners Buy Deep Learning Did Privacy Who about business career chat cloud digital marketing down hacker marketing millions number phones sent virus when work force Agriculture Bug Deep Earth GPS Gmail Google Maps Kaggle Keep NASA RAM Than Top Windows 11 World Cup Xiaomi address after also as at available camera change dangerous difference drive earn easy email going its job jobs language life look may message news old open price really search engine settings storage store such two used version watch windows 10 working 14 2020 2022 4 5 6 7 Cambridge Content Dark Web GB GPT Global Health-care Lite Maps Messages More Oppo Pakistan PayPal Print Pro QR Risk SEE SEO Samsung So Some Telegram TensorFlow Tutorial Type Types Vision Ways WiFi Zoom advertising attack been best better biggest blue brain chip comments country created cyber attacks don't electricity engine eyes fake files first football function game get go government hacked hackers hidden hours image install lost medical mind misused monitor moon once pay percent play problem processing program quantum robots scan science send share signal smartphones space stay story take their them thousands time topics tricks up using was water web where while wireless workers 000 2024 5G AI Education Alan Musk America Analytica Applications Army Blockchain Bounty CCTV COVID-19 Chat GPT Choose Clean Close Clubhouse Computer Vision Crypto DL Developer Development Docs Electric Explain Factory Finally Gemini Google chrome Google drive Healthcare Help Here's I IBM Japan Keras Kernels Large Lifestyle Looking MDMS Mac Models Musk Natural Ncell Net Notebooks PC Preparing Reasons Russia SIM SMS Save Scikit-Learn Skills SpaceX Stephen Hawking Sun Tesla Theme Therefore Thinking VPN Variables Word WorldLink ability accounts ads age airplane all any aware background bandwidth bank become beneficial between blocked bring bully cable call cameras captions capture care cause charge chatbots check come coming companies complete computers consumption copyright corona-virus courses create currency cyber security dataset datasets days delete deleted deleting details developed device dislike doctor documents doing domain due during dynamic energy engineer engineering exactly forever found fraud full gadgets games getting given good got guest handle his humans iOS iPhone 14 iPhones important including increase industry keyboard known launch law learn listen live manager map meaning megapixel memory messenger mode model month months most movies much name nonsense nuclear opening over own phishing physics porn post posts prevent problems product production programming protection quickly real-world reduce reward robotics run safe same saving say scandal searched selfie show site sold someone speaking speed spyware stuck students subscription systems target techology television tick today torrent traffic trillion universe upload verification voice war weakest women worldwide years & 'Buy the Dip' 'HDR' 'I' 'Mr. Beast' 'Professional Mode' 'football intelligence' 'hidden' 'refill station' (IoT) (LLM) (NLP) 1 100 10:10 10th 12 145 16 17 19 2 200 2007 25 35 3D 40 4000 48 4K 5 P's 60 7 C's 8 @everyone on A17 AI Tool AI ethics API AR Adjust Adobe Adopt Adsense Adsense Supports Africa Alexa Ali Baba Altman Amazon Jungle Amazon Prime Ambani American Anaconda Android 11 Android TV Android phone Annoyed Appoints Arithmetic Art Art through NFTs Artficial Intelligence Artificial neural Artuficial Intellegence Ashika Tamang Assignment Assistant Astronauts Astronomy Atrificial Inteligence Attacks Audiobooks Augmented Reality Australia Auto-GPT AutoML Avatar 2 Bachelors Banned Bard AI Based Because Before Bernie Sanders Big data BigQuery Bill Gates Bitwise Blind Blockchain Developer Blockchain Technology Books Brave Brave Browser Brazil C charger CPU CPU temperature CTEVT CV Cases Casting Changed ChatGBT Chery Chinese Citroën C5 Cloud Factory Cloud Factory Nepal Club House Colab Command Comparison Compute Concatenate Contactless Contactless payment system Copa America Copilot Couple Challenge Crash test Create your first Project on Python Crossover Cup DNS DRS Gaming Dark mode Datalab Deep Fake Deep Learinig Deep Learning with Python Deep Neural Networks Deepfake Demat Dept Development in predictive analytics Didn't Digital avatars Discontinuing Do not Dodge Dogecoin DuckDuckGo E-task EA ETF EU EV Earbuds Earth 2 Earthquake Edge Computing El Salvador Elected Electric Vehicles Electrical Elon Embedded Application Embedded Application (EA) Emoji Estimators Ethical Hacking Euro NCAP European Even Everyone Evolve Explained Explosion Express WiFi FPS Facebook Messenger Facebook's Facets Fears Federal Reserve System Finance Firefox FiveG Fixed wireless Follow Forge Fraud Call Freefire Freelancing GIF Git Gold Google Chat Google Cloud Google Meet Google Play Music Google Plus Google Plus code Google Workspace Google search Green room Greenroom. Spotify Guest Mode HDMI Happy Birthday Health sector Holi Honest Honeygain Huawei Hyundai ID IMD IP ISP Identify Implementing Includes Increasing Indonesia Inflation InfoSec Input Inspiration Installation Integrated circuit Intel Intelligent Internet of Things (IoT) Introduction Iranian Island Isn't JBL JPG JPMorgan Chase & Co Jack Ma January JavaScript Jio Joker Virus Jungle Jupyter Jupyter Notebooks Keys Korean LAN LLM LP Large Language Models Launch of better autonomous systems Lee Kun-hee Library Line Linux Logical Lucky MDMS Nepal ML Engine MSN MaAfee Mark Zuckerberg Max Meet Membership Mero Share Metaverse Microsoft Office Microsoft Teams Military Military weapons Mobile Operating System Module Mouse Mukesh Ambani Music Must NASA's NEA NFT NFTs Natural language processing (NLP) Nepal. radio mapping Nepali businesses Nepali game Nepali youth Nepalis NetTV Neural Network Neural Networks New Technology No Nokia North Korea Note Object Detection Open-source Opera Operating PDF PNG PPT PUBG Pandas Paytm Pendrive Photoshoot Pi Network Pip Plan Play Store Pokémon Pokémon Go Police Premium Preparations Prerequisite Prime Pro's Process Process discovery Pycharm Pyenv Python Programming Python Tutorial Python Tutorials Python for Beginners Python on Windows Quick Draw RCS Race Radically Ransomware Rashtra Bank Reboot Recommender Recommender Systems Redmi Reinforcement Reinforcement learning Reliance Reliance Jio Remove. bg Replacing Revolution Rice that grows for years once planted Rises Robot Sophia Roles Ronaldo Routine of Nepal Banda S&P 500 S&P Global Ratings SD Scale Scaling Scikit Screen Pinning Selection Seven Shorts Singapore Sitting SixG Snapchat Sophia South Korea Space X Spam Stable Coin Starlink Steve Jobs Stock market String Success Sundar Pichai Supermarket Supervised Supervised Learning Supervised Machine Learning Supply Chain Attack Supports Swift TIFF Telecom TensorBoard TensorFLow Hub Thes Tiktok stop Time Travel Tool Training Data Transforming Trojan Truecaller Trump Trusting Type-C US Congress USA USB United States Unnecessary Unsupervised Unsupervised Learning Unsupervised LearningUnsupervised Machine Learning Unsupervised Machine Learning Upcoming Upcoming Technology Urges Using a drone VPNs VR Vehicles Virtual reality Virtualenv Visualize WWW Wait Walkthrough Walmart WeChat Wha What are Assignment Operators in Python What are Comparison Operators in Python What are Logical Operators in Python What are Operators in Python What are the basic laws of quantum physics What is What is Chat GPT What is Google Adsense What is Pycharm What is Python What is String in Python What is Variable in Python Whose Wi-Fi 6 Wikipedia WordPress Wrangling data Write X X8 series XAI XOR XSS YouTuber Ziglar Zipty Zuckerberg admin advertisers again agency agricultural ai beauty air aircraft aired alert algorithm almost along alpha alternative analytics ancient angles announcement announces another answer answering antivirus anyone anything appear appearance appliances approach approaching approaching science meaning apps. google article artificial blood vessels arts associated attention audience automatic automatically autonomous avatars back backed ban bans bar basic batteries becoming beginner benefit benefits bitcoin mine bitcoins black block boarding bogged book bought box brand break brings broadband brought browsing bug bounty build but buttons buying bypass cable internet cables calculus calls campaign can't cancer cannot car cards careeer carry cave center challenge channel charger charging chat.com cheap cheaper checkmarks chess child children choose. a class clicking climbers clock closest club coding colleges color combat common communicate compensates compete competing computer mouse computer science concept connect cons control controls controversies could countries credit crisis criteria crore crores crowdsourcing culture cyberattack d about damaged danger dark data center data science dating apps day debit dedicated delete data depression destination devices diary die different digit digital cameras digital land digital privacy disappeared disappearing discovered discovery displaced display document dog dollars doodle door downloads dream drone drug trafficking e features e-Rupee e-books e-passport e-sewa eBooks ePassport each earn money from Nepal easier eating economy edit effective electronic else email server emails emerged emergency emojis employee employees end enough espionage etflix ethics except excessive excuse existence expected expire extracts eye face app facial verification facts family far farm fax fdown.net fee feet fiber fight file film final five flying foldable food footprint forced foreigners forget forgotten form formats foundation free upgrade frequency freshman from search fruit fuel game tips gamer gas gasoline geometry gets gives glasses goes good content goodbye goods google docs gossip granted great groups growing had hall hand handy happen happy harmful he head headphones headset heater hobby human brain human intelligence human trafficking hundreds hurting hydrogen hype iCloud iPhone 12 Pro illegal data illicit trade image processing processor images impair inbox incidents income increased incur insecure instant instrument interest internal storage internet speed into intranet introduced invented invention invest investment invites jack join journalists journey kit laboratory lakh languages last later latest launched launching lawmakers laws leak leaks legalize let letter letters light likes link lives loaded location locked longest lose loss love machine vision made main main features makes man manage management system mango marketplace martial mask matches measuring meetings melting meme messaging meta microphone middle million. downloads mine mistake mistakes mobile number moble moment monitors mountain move movie moving mute name-x naming near necessary neural neural networking new code new look new windows news anchor night mode non notes notifications now.gg nuclear energy obscene official offline open source opened operate operated operating system opposed optic optical fiber optimization option options other others our outbreak oversold owner page paid pandemic paper participant participate passports password. passwords patent pattern paying payment pen drive permanent permission person personal perspective phone confidential picture pictures pirated placed planting platform platforms political pop-up popular popularity port possible powered practice predictive pregnant prepared principles private prize processor product key programmatically programming languages project prompt property pros protected proxies proxy quantum computer quantum internet quires quota r daily radio rain rainy season rate reach reading ready real reason rebranding record recovery reform refresh refreshes refrigerator regarding registered registration regulators relationship released remain remove removes removing repairing replace report requiring reset residence resolution responsibilities restaurants returned revenue review rings risks risky road robotic dog rocket room rooms round ruin rules running safely safety sale satellite saying says scary schedule scheme schools screen screens search engines secret secretly secure selectric cars sell semi-final semiconductor sending series server services shared ships shocked shortage should shoulders shuffled shuts shutting sidebar simple since sites sky sleeping smartblock smartly social engineering hacking software. tech solve somewhere soon source sources space center space debris spacecraft spaceships special spectrum spend spending sponsors sports spying star starship start starting starvation steps stocks stolen stop stories strategy streaming student studying subject subscribers successful suggested suggestions suitable suitcase surface surprised survive t are tag tagging talent talk teach team technlogy technoloy technonlogy telecommunication terminology test text they think thousand thread threat to through throwaway timer tinder toilet too took topic tossing touch pad tracking trackpad trading transact transport travel trending trends trip turn turns tweets unbuyable unemployed unemployment unpleasant unregistered unsafe unseen unveils upgrades useful uses various versatility very view viral virtual virtual currency virtual world vishing visit visiting vulnerabilities warning washing waterproof we weapons web design websites week well went were wet willing woman works workspace world war worrie worth writer written wrong young
Show more