A highly optimized LLM inference acceleration engine for Llama and its variants.
zhihu/ZhiLight is cooling — star growth is decelerating by -0.1/day, though it still holds a momentum score of 27.1/100.
Low breakout odds over the next 14 days, no signal is above the cohort median. Confidence is high given the available history.
Transparent heuristic · logged for model training
Reconstructed from 167 snapshots — the time-series GitHub’s API doesn’t expose.
Live score, refreshed every 6 hours. Links back to this page.
[](https://repoclimb.com/repo/zhihu/ZhiLight)Percentile rank within the tracked cohort. The score self-calibrates — it answers “accelerating vs. everything else,” not raw size.