Sameera Ramasinghe
PulseAugur coverage of Sameera Ramasinghe — every cluster mentioning Sameera Ramasinghe across labs, papers, and developer communities, ranked by signal.
1 day(s) with sentiment data
-
Protocol Models enable efficient decentralized AI training on low-end hardware
Researchers have developed a novel compression algorithm called Protocol Models designed to improve the efficiency of decentralized deep learning training. This method compresses both forward and backward passes of mode…
-
New method enables efficient decentralized training of LLMs with 100K+ context
Researchers have developed a novel method for training large language models with extended context windows in decentralized environments. This technique, called Mixtures of Subspaces, significantly compresses communicat…
-
New 'Architecture Warm-Up' Stabilizes Transformer Training
Researchers have developed a new method to stabilize the training of large Transformer models, which are often prone to instability and divergence. The approach, called "architecture warm-up," involves progressively inc…