🔥 A New Release of PyTorch is Here
the most useful free ML&AI newsletter
📝 Editorial
PyTorch is one of the two most popular deep learning frameworks in the current market and a favorite of the AI research community. For the last few years, every release of PyTorch has made incremental improvements towards making it one of the most complete deep learning stacks for data science teams. That’s the spirit of PyTorch 1.9, the newest release of the popular framework that was just open sourced this week. The new version of PyTorch incorporates native support for many relevant elements of modern deep learning solutions.
Distributed and scalable training is front and center of this new version of PyTorch. The new release incorporates the TorchElastic distributed training framework as part of PyTorch Core. It also includes an upgraded version of PyTorch RPC optimized for large-scale distributed training workloads. Mobile is another area of improvement in PyTorch 1.9 with the beta version of Mobile Interpreter, a streamlined version of the PyTorch runtime optimized for mobile devices. Also on the mobile front, PyTorch 1.9 adds support for the popular TorchVision library for mobile computer vision scenarios. One new area that data science teams will find refreshing in PyTorch 1.9 is the improved support for scientific, computing models built using popular libraries such as torch.linalg, torch.special, and Complex Autograd.
As Pytorch approaches its 2.0 release, it continues to capably support complex and diverse deep learning scenarios in real-world applications. PyTorch 1.9 is not an overwhelming release from the new features standpoint but certainly brings new and exciting capabilities to the PyTorch ecosystem. Go check it out.
🔺🔻TheSequence Scope – our Sunday edition with the industry’s development overview – is free. To receive high-quality content about the most relevant developments in the ML world every Tuesday and Thursday, please subscribe to TheSequence Edge 🔺🔻
🗓 Next week in TheSequence Edge:
Edge#99: we discuss what trust region and proximal policy optimization are; explore RLib – an open-source framework for highly scalable RL; learn how OpenAI used PPO RL to master Dota 2.
Edge#100: about Facebook NetHack challenge, which is likely to become one of the toughest reinforcement learning benchmarks in history.
Now, let’s review the most important developments in the AI industry this week
🔎 ML Research
HuBERT
Facebook AI Research (FAIR) published a paper introducing HuBERT, a self-supervised learning method for speech models ->read more on FAIR team blog
Learning a Physics Simulator
Google Research published a paper outlining reinforcement learning techniques that can learn the physics of an environment with an incredible degree of accuracy ->read more on Google Research blog
Building Task-Oriented Dialog Systems
Microsoft Research published a paper and open source code of SOLOIST, a model based on machine teaching and transfer learning for building more efficient task-specific dialog bots->read more on Microsoft Research blog
🤖 Cool AI Tech Releases
PyTorch 1.9
A new version of PyTorch was released this week with a lot of exciting features ->read a summary of the new release on PyTorch blog
Ludwig 0.4
Ludwig, the low-code ML stack pioneered by Uber, released its 0.4 version (next week we are publishing the interview with Piero Molino, author of Ludwig. Stay tuned!) ->read more on Ludwig team blog
DataRobot 7.1
AutoML powerhouse DataRobot announced the newest release of its platform, which includes integration with Snowflake, automated feature discovery and many other cool capabilities ->read more in this press release from DataRobot
Dataiku Online
Enterprise analytics startup Dataiku unveiled a fully managed cloud service on its platform->read more on Dataiku blog
💬 Useful tweet
Emulating Text in Images
This week FAIR published a paper introducing TextStyleBrush, a model that can emulate the text style in a photo from a single shot ->read more on FAIR team blog
💸 Money in AI
What do you think about this new format of Money in AI? Is it more useful for you this way, or is it better with more details about the companies? Let us know by replying to this email.
Graph database platform Neo4j raised $325 million in a Series F round led by Eurazeo. Hiring for over 120 positions.
Industrial analytics and machine monitoring platform MachineMetrics raised $20 million in a Series B funding round led by Teradyne. Hiring.
Crate.io, which develops the scalable SQL database for machine data, raised $10 million in additional funding. Hiring.
Cord, which works on automating annotation for computer vision, raised $4.5 million in a seed round led by CRV. Hiring.
Human-centered AI platform Vianai Systems raised $140 million in a round led by SoftBank Vision Fund 2.
AI-powered talent marketplace Gloat raised $57 million in a Series C round led by Accel. Hiring for over 60 positions.
Computer vision startup Trigo raised $10 million in funding from REWE Group and Viola Growth. Hiring.
AI-powered talent engagement and communications platform Sense raised $16 million in a Series C round of funding. Almost 60 job openings.
AI-enhanced sales and revenue acceleration platform Introhive raised $100 million in a Series C funding round led by equity firm PSG.
AI-powered cognitive trust platform for cybersecurity Elisity raised $26 million in a funding round led by Two Bear Capital and AllegisCyber Capital. Hiring.