Tencent выложила Hy-Nomas: опенсорс-модель интерактивного видеомира, которая запускается на одной RTX 4090 🌍 Модель строит мир по сегментам видео, а вы управляете ею текстом и траекторией камеры: на каждом шаге можно сменить сцену, стиль, действие или ракурс, и память мира сохраняется между сегментами. Выход 720x1280 при 24 fps. Для узких областей вроде робототехники есть лёгкая адаптация через LoRA. На одной RTX 4090 модель идёт с INT8-весами, INT8 KV-кэшем и выгрузкой на CPU, стандартная конфигурация рассчитана на восемь GPU. В репозитории код обучения, инференса и демо, лицензия Apache-2.0. Бенчмарков для самой модели пока нет, технический отчёт обещают позже. Демо: https://hy-nomas.hy.tencent.com/demo Веса: https://huggingface.co/tencent/Hy-Nomas @ai_machinelearning_big_data
From the original source:
Telegram — AI Machine Learning Big Data
Read full article at Telegram — AI Machine Learning Big DataTencent releases Hy-Nomas: open-source interactive video world model that runs on a single RTX 4090 🌍
Tencent releases Hy-Nomas: open-source interactive video world model that runs on a single RTX 4090 🌍 The model builds the world from video segments, and you control it with text and camera trajectory: at each step you can change the scene, style, action, or angle, and the world's memory…
AI Summary & Key Takeaways
- Tencent releases Hy-Nomas: open-source interactive video world model that runs on a single RTX 4090 🌍 The model builds the world from video segments, and you control it with text and camera trajectory: at each step you can change the scene, style, action, or angle, and the world's memory…
Tencent releases Hy-Nomas: open-source interactive video world model that runs on a single RTX 4090 🌍
Related stories
Webinar to discuss how the FCC robotics ruling affects automation buyers
Vecna Robotics makes its systems in the U.S. and is not directly affected by the FCC ruling. Credit: Eugene Demaitre In July, the Federal Communications Commission, or FCC, said it had determined that certain foreign-made advanced robots pose a cybersecurity and competitive risk…
Multi-step Reinforcement Learning on Amazon SageMaker AI
AWS Machine Learning describes multi-step reinforcement learning on SageMaker AI.
Chinese AI model Kimi K3.1 Pro is closing in on Anthropic's Claude Opus 5.5
Moonshot AI's Kimi K3.1 Pro scores 55 points on the Artificial Analysis Intelligence Index, an 11-point jump over its predecessor that puts it in fourth place. Only three Anthropic and Meta models rank higher. If Moonshot releases the weights as planned, the gap between open and…
Generative Bionics to launch Physical AI Data Factory in Italy
The first Data Factory hub is set to become one of Europe’s largest physical AI training facilities for humanoid robots. | Source: Generative Bionics Generative Bionics this week announced plans to build a proprietary, multi-site “Physical AI Data Factory” in Italy. The company…