Posted by Nari Yoon, Hee Jung, DevRel Group Supervisor / Soonson Kwon, DevRel Program Supervisor
Let’s discover highlights and accomplishments of huge Google Machine Studying communities over the past quarter of 2022. We’re enthusiastic and grateful about all of the actions by the worldwide community of ML communities. Listed below are the highlights!
ML at DevFest 2022
 |
ML Group Summit 2022
 |
TensorFlow
 |
gMLP: What it’s and methods to use it in apply with Tensorflow and Keras? by ML GDE Radostin Cholakov (Bulgaria) demonstrates the state-of-the-art outcomes on NLP and pc imaginative and prescient duties utilizing lots much less trainable parameters than corresponding Transformer fashions. He additionally wrote Differentiable discrete sampling in TensorFlow.
Constructing Laptop Imaginative and prescient Mannequin utilizing TensorFlow: Half 2 by TFUG Pune for the builders who wish to deep dive into coaching an object detection mannequin on Google Colab, inspecting the TF Lite mannequin, and deploying the mannequin on an Android utility. ML GDE Nitin Tiwari (India) coated detailed facets for end-to-end coaching and deployment of object mannequin detection.
Creation of Code 2022 in pure TensorFlow (days 1-5) by ML GDE Paolo Galeone (Italy) fixing the Creation of Code (AoC) puzzles utilizing solely TensorFlow. The articles include an outline of the options of the Creation of Code puzzles 1-5, in pure TensorFlow.
 |
Construct tensorflow-lite-select-tf-ops.aar and tensorflow-lite.aar recordsdata with Colab by ML GDE George Soloupis (Greece) guides how one can shrink the ultimate measurement of your Android utility’s .apk by constructing tensorflow-lite-select-tf-ops.aar and tensorflow-lite.aar recordsdata with out the necessity of Docker or private PC setting.
TensorFlow Lite and MediaPipe Utility by ML GDE XuHua Hu (China) explains methods to use TFLite to deploy an ML mannequin into an utility on gadgets. He shared experiences with creating a movement sensing recreation with MediaPipe, and methods to remedy issues that we could meet normally.
Keras
 |
 |
TFX
 |
Usha Rengaraju (India) shared TensorFlow Prolonged (TFX) Tutorials (Half 1, Half 2, Half 3) and the next TF initiatives: TensorFlow Choice Forests Tutorial and FT Transformer TensorFlow Implementation.
JAX/Flax
JAX Excessive-performance ML Analysis by TFUG Taipei and ML GDE Jerry Wu (Taiwan) launched JAX and methods to begin utilizing JAX to resolve machine studying issues.
 |
Kaggle
Low-light Picture Enhancement utilizing MirNetv2 by ML GDE Soumik Rakshit (India) demonstrated the duty of Low-light Picture Enhancement.
 |
Cloud AI
Higher {Hardware} Provisioning for ML Experiments on GCP by ML GDE Sayak Paul (India) mentioned the ache factors of provisioning {hardware} (particularly for ML experiments) and the way we will get higher provision {hardware} with code utilizing Vertex AI Workbench cases and Terraform.
 |
 |
Extra sensible time-series mannequin with BQML by ML GDE JeongMin Kwon (Korea) launched BQML and time-series modeling and confirmed some sensible purposes with BQML ARIMA+ and Python implementations.
Analysis & Ecosystem
AI in Healthcare by ML GDE Sara EL-ATEIF (Morocco) launched AI purposes in healthcare and the challenges dealing with AI in its adoption into the well being system.
Girls in AI APAC completed their journey at ML Paper Studying Membership. Throughout 10 weeks, members gained data on excellent machine studying analysis, realized the newest methods, and understood the notion of “ML analysis” amongst ML engineers. See their session right here.
A Pure Language Understanding Mannequin LaMDA for Dialogue Functions by ML GDE Jerry Wu (Taiwan) launched the pure language understanding (NLU) idea and shared the operation mode of LaMDA, mannequin fine-tuning, and measurement indicators.
Python library for Arabic NLP preprocessing (Ruqia) by ML GDE Ruqiya Bin (Saudi Arabia) is her first python library to serve Arabic NLP.
 |
Anatomy of Capstone ML Initiatives 🫀by ML GDE Sayak Paul (India) mentioned engaged on capstone ML initiatives that can stick with you all through your profession. He coated varied subjects starting from drawback choice to tightening up the technical gotchas to presentation. And in Enhancing as an ML Practitioner he shared his studying from expertise within the subject engaged on a number of facets.
 |
Transcending Scaling Legal guidelines with 0.1% Additional Compute by ML GDE Grigory Sapunov (UK) reviewed a latest Google article on UL2R. And his posting Discovering sooner matrix multiplication algorithms with reinforcement studying defined how AlphaTensor works and why it will be important.
Again in Particular person – Prompting, Directions and the Way forward for Massive Language Fashions by TFUG Singapore and ML GDE Sam Witteveen (Singapore) and Martin Andrews (Singapore). This occasion coated latest advances within the subject of huge language fashions (LLMs).
ML for Manufacturing: The artwork of MLOps in TensorFlow Ecosystem with GDG Casablanca by TFUG Agadir mentioned the motivation behind utilizing MLOps and the way it might help organizations automate a variety of ache factors within the ML manufacturing course of. It additionally coated the instruments used within the TensorFlow ecosystem.
