jakeatx/llamAmpere-LSA-quality-Qwen3.8-27B download history

jakeatx/llamAmpere-LSA-quality-Qwen3.8-27B is a dataset on the Hugging Face Hub. In the last 30 days it was downloaded 66 times (66 in the last 7 days), and 66 times in total. It ranks #209,436 among datasets by monthly downloads.

LSA quality results: how much does sparse attention change the output? llamAmpere's LSA (Latent Sparse Attention, Prefrontal AI, arXiv 2610.42147; our engine is a rewrite of their method, using their released distilled index map) lets a 27B model run 200K+ context on 8, 10 and 12 GB c

Open jakeatx/llamAmpere-LSA-quality-Qwen3.8-27B on Hugging Face

Sister project: Paper Pulse, the upvote history of every Hugging Face Daily Paper.