Visual language models have evolved significantly recently. However, the existing technology typically only supports one single image. They cannot reason among multiple images, support in context learning or understand videos. Also, they don’t optimize for inference speed.

We developed VILA, a visual language model with a holistic pretraining, instruction tuning, and deployment pipeline that helps our NVIDIA clients succeed in their multi-modal products. VILA achieves SOTA performance both on image QA benchmarks and video QA benchmarks, having strong multi-image reasoning capabilities and in-context learning capabilities. It is also optimized for speed.

It uses 1 ⁄ 4 of the tokens compared to other VLMs and is quantized with 4-bit AWQ without losing accuracy. VILA has multiple sizes ranging from 40B, which can support the highest performance, to 3.5B, which can be deployed on edge devices such as NVIDIA Jetson Orin.

Non-personalized content and ads are influenced by things like the content you’re currently viewing and your location (ad serving is based on general location). Personalized content and ads can also include things like video recommendations, a customized YouTube homepage, and tailored ads based on past activity, like the videos you watch and the things you search for on YouTube. We also use cookies and data to tailor the experience to be age-appropriate, if relevant.

Select “More options” to see additional information, including details about managing your privacy settings. You can also visit g.co/privacytools at any time.

Wake up babe new Optimus video just droppedpic.twitter.com/dsJzQXf3rp.
— Adrian Dittmann (@AdrianDittmann) May 5, 2024.

Charge-recharge cycling of lithium-super-rich iron oxide, a cost-effective and high-capacity cathode for new-generation lithium-ion batteries, can be greatly improved by doping with readily available mineral elements.

The 30th Crystal Award honour the global artists Diébédo Francis Kéré, Michelle Yeoh and Nile Rodgers. The awards celebrate their enduring contributions. #wef24

A new model provides an improved description of the flow of the ocean’s kinetic energy by including friction with the coasts.

Amod Sandhya Lele is interviewed by Céline Leboeuf.

The importance of psychology in resisting China.

Electrify America (EA) increased its total number of chargers to over 4,000 in 2023 – here’s what it’s aiming to achieve in 2024.

In 2022, there were 5.3 million customer charging sessions on Electrify America’s network. In 2023, validated data reflected 10.9 million customer charging sessions – a 106% increase year-over-year.

EA’s EV charging network grew to over 900 stations across 47 US states, the District of Columbia, and six Canadian provinces last year. It also expanded into Hawaii and North Dakota for the first time.