mezark

Recent Submissions

1. 1. ▲ Moe inference optimizations: 15% lower expert load by request reordering (blog.doubleword.ai) 3 points · 14 days ago · 0 comments
2. 2. ▲ Tensor Network Attention (mainlymatmul.com) 2 points · 28 days ago · 0 comments
3. 3. ▲ Redundant Information in LLM Weights (fergusfinn.com) 5 points · 1 month ago · 0 comments
4. 4. ▲ Tans: Precomputing RANS (fergusfinn.com) 3 points · 1 month ago · 0 comments
5. 5. ▲ Also-RANS: Asymmetric Numeral Systems for Entropy Coding (fergusfinn.com) 25 points · 1 month ago · 0 comments
6. 6. ▲ 70x faster cold(ish) starts for SGLang (fergusfinn.com) 4 points · 1 month ago · 0 comments
7. 7. ▲ QueueSpec – drafting speculation tokens while a request queues (blog.doubleword.ai) 1 point · 4 months ago · 0 comments
8. 8. ▲ ZeroDP: Just-in-Time Weight Offloading over NVLink for Data Parallelism (mainlymatmul.com) 1 point · 4 months ago · 0 comments
9. 9. ▲ Parallel Primitives for Multi-Agent Workflows (fergusfinn.com) 1 point · 4 months ago · 0 comments
10. 10. ▲ New fastest AI Model Gateway – 450x less overhead than LiteLLM (github.com) 2 points · 7 months ago · 0 comments