Article List

PyTorch Mobile provides a runtime environment to execute state-of-the-art machine learning models on mobile devices. Latency is reduced, privacy preserved, and models can run on mobile devices anytime, anywhere.

In this blog post, we provide a quick overview of 10 currently available PyTorch Mobile powered demo apps running various state-of-the-art PyTorch 1.9 machine learning models spanning images, video, audio and text.

It’s never been easier to deploy a state-of-the-art ML model to a phone. You don’t need any domain knowledge in Machine Learning and we hope one of the below examples resonates enough with you to be the starting point for your next project.

Image segmentation model trained from scratch on the Oxford Pets dataset.

This example shows how to do image classification from scratch, starting from JPEG image files on disk, without leveraging pre-trained weights or a pre-made Keras Application model. We demonstrate the workflow on the Kaggle Cats vs Dogs binary classification dataset.

We use the image_dataset_from_directory utility to generate the datasets, and we use Keras image preprocessing layers for image standardization and data augmentation.

This example demonstrates how to implement a deep convolutional autoencoder for image denoising, mapping noisy digits images from the MNIST dataset to clean digits images. This implementation is based on an original blog post titled Building Autoencoders in Keras by François Chollet.

This example demonstrates two techniques for building memory-efficient recommendation models by reducing the size of the embedding tables, without sacrificing model effectiveness:

1. Quotient-remainder trick, by Hao-Jun Michael Shi et al., which reduces the number of embedding vectors to store, yet produces unique embedding vector for each item without explicit definition.
2. Mixed Dimension embeddings, by Antonio Ginart et al., which stores embedding vectors with mixed dimensions, where less popular items have reduced dimension embeddings.

An exploration of threads, processes, and coroutines in Python, with interesting examples that illuminate the differences between each.

Notice: This map is not a precise reflection of the state of the AI field, but just my subjective representation.

This is my first map as of the end of 2020 and will be extended in the future. It contains more than 200 words or phrases, so to describe all of them would be too extensive and overkill. Much more interesting (and useful for me) to tell how this map was gradually building in my head. I will not explain everything, just the main things, so it is normal not to understand something.

Learn how to build a robust and developer-friendly Python microservices infrastructure using gRPC and Kubernetes. You’ll also explore advanced topics such as interceptors and integration testing.
 

Let's pretend we need to build a recommendation engine for an eCommerce web site.

There are basically two approaches you can take: content-based and collaborative-filtering. We'll look at some pros and cons of each approach, and then we'll dig into a simple implementation (ready for deployment on Heroku!) of a content-based engine.

Authentication is definitely a hard and complicated problem. Protocols like OAuth2 try to make it simpler, but in fact they make it harder to understand for beginners, as the reference is quite complex and there aren’t a lot of good practices available online.

The new framework FastAPI is now our go-to web library for all our projects, as it is very efficient to develop with and it supports amazing typing out of the box. The only issue we have is dealing with authentication when using a JS Frontend in front of it. Let’s close this debate once and for all by describing the authentication scheme that I think everyone needs for a simple web application with FastAPI, using an external provider.

Beginner’s guide to ML solutions deployment maintaining concurrency + load testing. Discussion on the Data Scientist’s responsibilities in terms of deployment.

Flash is a collection of tasks for fast prototyping, baselining and fine-tuning scalable Deep Learning models, built on PyTorch Lightning.

Whether you are new to deep learning, or an experienced researcher, Flash offers a seamless experience from baseline experiments to state-of-the-art research. It allows you to build models without being overwhelmed by all the details, and then seamlessly override and experiment with Lightning for full flexibility. Continue reading to learn how to use Flash tasks to get state-of-the-art results in a flash.