Dernière minute

Senate Intelligence Committee Postpones Jay Clayton Hearing After Trump Intervention Windows and Linux users urged to update Secure Boot keys before June 24 deadline Google Home Speaker Finally Gets a Release Date and Price UniQure to Seek FDA Approval for Huntington's Disease Gene Therapy Mastodon Bets on Email Newsletters for Audience Growth Tropical Storm Arthur Forms Near Gulf Coast, Threatening Texas and Louisiana with Heavy Rain and Flooding The Rise of 'Slow Tech': Why People Are Ditching Smartphones for Simpler Devices Thread Group Launches Beta of New Thread Networks Diagnostics App Best Apple Deals Ahead of Prime Day: AirPods, Apple Watch, iPads, and More Geothermal Startup Critical Energy Raises $19 Million Seed Funding Senate Intelligence Committee Postpones Jay Clayton Hearing After Trump Intervention Windows and Linux users urged to update Secure Boot keys before June 24 deadline Google Home Speaker Finally Gets a Release Date and Price UniQure to Seek FDA Approval for Huntington's Disease Gene Therapy Mastodon Bets on Email Newsletters for Audience Growth Tropical Storm Arthur Forms Near Gulf Coast, Threatening Texas and Louisiana with Heavy Rain and Flooding The Rise of 'Slow Tech': Why People Are Ditching Smartphones for Simpler Devices Thread Group Launches Beta of New Thread Networks Diagnostics App Best Apple Deals Ahead of Prime Day: AirPods, Apple Watch, iPads, and More Geothermal Startup Critical Energy Raises $19 Million Seed Funding

DiffusionGemma

Stable2 articles2 sourcesDernière mise à jour: 6 g önce

Sujets associéstext generation Google AI open model Hugging Face NVIDIA speculative decoding bidirectional attention

Derniers articles

Google Releases DiffusionGemma, an Open AI Model for Faster Text Generation

En développement

Tech·6 g önceRésumé IA

Google Releases DiffusionGemma, an Open AI Model for Faster Text Generation

Google has launched DiffusionGemma, an open AI model that generates text using a diffusion process, achieving up to 1,000 tokens per second on NVIDIA H100. While faster than traditional autoregressive models, it currently trails in output quality and requires specific hardware and configurations for efficient local execution.

Google DeepMind Unveils DiffusionGemma, a New AI Model for Faster Text Generation

En développement

Tech·6 g önceRésumé IA

Google DeepMind Unveils DiffusionGemma, a New AI Model for Faster Text Generation

Google DeepMind has released DiffusionGemma, an experimental AI model that generates text in parallel, unlike traditional autoregressive models. This approach, similar to image diffusion, offers faster and more efficient text generation on local hardware, potentially improving tasks like in-line editing and Sudoku solving.