Search Articles — Sudonull

Search Results

In this project

ruGPT3XL with 8k context: sparse attention

https://sudonull.com/rugpt3xl-with-8k-context-sparse-attention

Expansion of ruGPT3XL to 8k tokens via sparse attention and fine-tuning. PPL 13.00 on gazeta, Triton/SDPA optimization ×1.85. Instructions for developers.

Kimi K2.6: code optimization by AI agent +185%

https://sudonull.com/kimi-k2-6-code-optimization-by-ai-agent-185

How open-source model Kimi K2.6 rewrote the exchange engine in 13 hours and increased throughput by 185%. Analysis of architecture, benchmarks, and practical cases.

From other projects

Trending Now