Ulysses
PulseAugur coverage of Ulysses — every cluster mentioning Ulysses across labs, papers, and developer communities, ranked by signal.
3 day(s) with sentiment data
-
Wave Forcing enables efficient streaming video generation
Researchers have introduced Wave Forcing (WF), a novel framework designed for efficient and high-quality continuous long-video generation. This method co-designs the generation, training, and execution processes to enab…
-
New Layerwise Repacking method optimizes hybrid-attention model training
Researchers have developed a new technique called Layerwise Repacking to optimize the training of hybrid-attention models. This method addresses the conflicting parallelism requirements between full-attention and sparse…
-
Luma Labs details multimodal sequence processing innovations
Luma Labs has released details on their advancements in efficient multimodal sequence processing, particularly for packed sequences. They have developed techniques that improve performance and reduce computational overh…
-
Nolan's Odysseus film reflects Western elite's existential angst
Christopher Nolan's film "Odysseus" is interpreted as a modern reflection on the Western elite's existential anxieties and their historical pursuit of power. The film's portrayal of the Homeric hero Odysseus, a figure o…
-
LiveAnimate achieves real-time, stable long-form human animation
Researchers have developed LiveAnimate, a novel system capable of generating stable, long-form human animations in real-time. This system utilizes a 14-billion parameter video diffusion transformer, enhanced with specia…
-
Mesh-Attention improves LLM scaling with better data locality · arXiv paper
Researchers have developed Mesh-Attention, a novel communication-efficient distributed attention mechanism designed to improve data locality for scaling large language models (LLMs) to long contexts. Unlike previous met…
-
New X-Stage pipeline optimization boosts DiT inference speed
Researchers have identified a new pipeline stage, termed X-Stage, that can optimize communication-computation overlap during the inference of Diffusion Transformers (DiTs). This stage focuses on the period after communi…
-
Wan-Streamer v0.2 boosts audio-visual interaction resolution while preserving low latency
Researchers have released Wan-Streamer v0.2, an updated model for end-to-end audio-visual interaction that significantly increases output resolution while maintaining low latency. The new version raises the interactive …
-
New 'process sidecar' method allows precise memory revocation in language models
Researchers have introduced "process sidecars" as a novel method for revoking learned information from language models after safety training. This technique aims to precisely remove specific memories without negatively …
-
Hugging Face releases AI updates for LeRobot, Ulysses, and RL training
Hugging Face has released updates across several AI projects. LeRobot v0.5.0 introduces scaling across all dimensions, while Ulysses implements sequence parallelism for training with a 1 million token context window. Ad…