LoRA Serving Trace

This repo contains the statistics of 126 LoRA inference services within one day. All data is at a 1 - minute granularity. The metrics include:

1. Request rate
2. Average input token length
3. Average output token length
4. Total input token number
5. Total output token number
6. Input token number percentiles
7. Output token number percentiles

All data has been normalized. Only relative values are included.