RAHUL YASHWANTKUMAR GUPTA

ryg81

AI & ML interests

None yet

Recent Activity

updated a collection about 4 hours ago
z-image
replied to BestWishYsh's post about 10 hours ago
๐Ÿš€ Introducing Helios: a 14B real-time long-video generation model! Itโ€™s completely wildโ€”faster than 1.3B models and achieves this without using self-forcing. Welcome to the new era of video generation! ๐Ÿ˜Ž๐Ÿ‘‡ ๐Ÿ’ป Code: https://github.com/PKU-YuanGroup/Helios ๐Ÿ  Page: https://pku-yuangroup.github.io/Helios-Page ๐Ÿ“„ Paper: https://huggingface.co/papers/2603.04379 ๐Ÿ”น True Single-GPU Extreme Speed โšก๏ธ No need to rely on traditional workarounds like KV-cache, quantization, sparse/linear attention, or TinyVAE. Helios hits an end-to-end 19.5 FPS on a single H100! Training is also highly accessible: an 80GB VRAM can fit four 14B models. ๐Ÿ”น Solving Long-Video "Drift" from the Core ๐ŸŽฅ Tired of visual drift and repetitive loops? We ditched traditional hacks (like error banks, self-forcing, or keyframe sampling). Instead, our innovative training strategy simulates & eliminates drift directly, keeping minute-long videos incredibly coherent with stunning quality. โœจ ๐Ÿ”น 3 Model Variants for Full Coverage ๐Ÿ› ๏ธ With a unified architecture natively supporting T2V, I2V, and V2V, we are open-sourcing 3 flavors: 1๏ธโƒฃ Base: Single-stage denoising for extreme high-fidelity. 2๏ธโƒฃ Mid: Pyramid denoising + CFG-Zero for the perfect balance of quality & throughput. 3๏ธโƒฃ Distilled: Adversarial Distillation (DMD) for ultra-fast, few-step generation. ๐Ÿ”น Day-0 Ecosystem Ready ๐ŸŒ We wanted deployment to be a breeze from the second we launched. Helios drops with comprehensive Day-0 hardware and framework support: โœ… Huawei Ascend-NPU โœ… HuggingFace Diffusers โœ… vLLM-Omni โœ… SGLang-Diffusion Try it out and let us know what you think!
updated a collection about 22 hours ago
WAN
View all activity

Organizations

None yet