#python#gym#gym_environment#reinforcement_learning#reinforcement_learning_agent#reinforcement_learning_environments#rl_environment#rl_training
NeMo Gym helps you build and run reinforcement‑learning training environments for large language models, letting you develop, test, and collect verified rollouts separately from the training loop and integrate with your preferred RL framework and model endpoints (OpenAI, vLLM, etc.). It includes ready resource servers, datasets, and patterns for multi‑step, multi‑turn, and tool‑using scenarios, runs on a typical dev machine (no GPU required), and is early-stage with evolving APIs and docs. Benefit: you can generate high‑quality, verifiable training data faster and plug it into existing training pipelines to improve model behavior.
https://github.com/NVIDIA-NeMo/Gym
📰 FFmpeg Introduces Vulkan-Accelerated 360 Degree Video Conversion
Beyond the capabilities of just the Vulkan Video API, the FFmpeg multimedia library has made interesting Vulkan-accelerated adaptations using compute shaders. With Vulkan compute they've implemented Apple ProRes video acceleration, FFV1 decode, and other features. The newest Vulkan feature now in place for FFmpeg is 360 degree video conversion...
🔗 Source: https://www.phoronix.com/news/FFmpeg-360-Degree-Vulkan
#ffmpeg
📰 FFmpeg 8.1 Released With Experimental xHE-AAC MPS212, More Vulkan Acceleration
FFmpeg 8.1 is out today as the newest stable release of this widely-used, open-source multimedia library...
🔗 Source: https://www.phoronix.com/news/FFmpeg-8.1-Released
#opensource#ffmpeg
📰 Gentoo-Based Redcore Linux Hardened 2601 Released with Kernel 6.19
Redcore Linux Hardened 2601 Vulpecula has been released, featuring Linux kernel 6.19, FFmpeg 8, and updates to the Sisyphus package manager.
🔗 Source: https://linuxiac.com/gentoo-based-redcore-linux-hardened-2601-released-with-kernel-6-19/
#ffmpeg#kernel#linux
Кстати! Я всем советую попробовать HandBrake — программу для ужатия видео.
Это графический интерфейс для мощной библиотеки ffmpeg, которая заточена на перегонку видео в разные форматы.
Я в основном использую для уменьшения размера файлов, чтобы телеграм проигрывал их без звука автоматически
Тут подробнее писал как это работает
Анимацию постом выше я просто закинул и без изменения настроек перегнал из mpg в mp4. Размер изменился с 4.5мб до 300кб
А для тизера «Коротких вопросов» нажав две кнопки, смог уменьшить файл со 167 мб до (!) 3 мб. За счет уменьшения размера сторон с 4к до 720p.
На мобиле разницы скорее всего не будет видно, а вовлечение в просмотр повысится — сплошная польза!
Работает на всех платформах, в том числе на маке и линуксе!
🎤Ссылки на утро — второй канал
⏲Ускорить YouTube за звезду (VPN за 2₽)
#ToolReview@cogload#ffmpeg@cogload#telegram@cogload
🔖 The creator of ffmpeg Fabrice Bellard(1) is truly a genius and has created so many amazing software that have been an amazing benefit to the world. | Hacker News #pinboard#pantheon#ffmpeg#person
Yeah definitely. If programming genius had a unit it should be called the bellard.
One 10x programmer = a 1/10th Bellard?
100x 工程师，可以放在万神殿了
https://news.ycombinator.com/item?id=25487711
This is Probably the Best Video Downloader App (And it is Free and Open Source) | itsFOSS
VidBee allows you to download videos from YouTube, Facebook, X, Instagram, etc. In fact, it supports over 1,800 websites.
It is built on top of popular command line tools like yt-dlp and #ffmpeg. For the interface, it uses the Electron framework. I understand that some people dislike Electron framework as it runs a web browser underneath, but the 'advantage' of this framework is that you get the same interface in all the operating systems. At least, it's an advantage for the developers as they don't have to build the interface separately for #Linux, #Windows and #macOS.
The source code for VidBee is available on its GitHub repository.
#VidBee - Free Open Source Video Downloader
https://vidbee.org/