engblogs

Experience the power of row-level secondary indexing in Apache Parquet, enabling 3-20X faster upserts and efficient table ACID operations in Lakehouse architecture.

Reducing computational expense of large language models to make them faster and more robust.

Sharing insights gathered from conversations in 22 countries and discussing how to incorporate them moving forward.

Meta is transferring its IP for Evenstar to the Open Compute Project (OCP) to accelerate the adoption of open RAN technologies, providing the telecom industry with an open, white box radio unit solution.

This technical blogpost discusses the development of on-device diffusion plugins for conditioned text-to-image generation, enabling controllable text-to-image generation to be run on mobile devices.

The blogpost announces the first Machine Unlearning Challenge, which aims to develop efficient, effective, and ethical unlearning algorithms to remove the influence of specific training examples from machine learning models while maintaining model utility and generalization.

Discover 7 Spanish words that cannot be fully translated into English!

Guide on how to use Italian language to order food in Italy

Exploring whether quantization difficulties at scale can be altered and conditioned by optimization choices in Large Language Models (LLMs).

A guide on creating and utilizing custom command models with a focus on finetuning and its advantages.

Learn how to leverage Cohere's Command model in Amazon SageMaker Studio with the help of Amazon SageMaker JumpStart.

This technical blogpost introduces CoarsenConf, an SE(3)-equivariant hierarchical variational autoencoder (VAE) that generates low-energy 3D molecular conformers by pooling information from fine-grain atomic coordinates to a coarse-grain subgraph level representation for efficient autoregressive generation.