== sglang-dev+DFlash2 (gittensor-model-hub/Qwen3.8-27B-NVFP4-RTX5090) warmup...
  N=1  agg= 294.6 tok/s  per-req=294.7 tok/s  p50= 2.22s p95= 2.22s  (653 tok/2.22s)
  N=2  agg= 616.6 tok/s  per-req=308.4 tok/s  p50= 2.38s p95= 2.38s  (1470 tok/2.38s)
  N=4  agg= 656.9 tok/s  per-req=215.1 tok/s  p50= 4.22s p95= 4.22s  (2772 tok/4.22s)
  N=8  agg= 653.9 tok/s  per-req=130.8 tok/s  p50= 5.99s p95= 7.96s  (5208 tok/7.96s)
JSON {"label": "sglang-dev+DFlash2", "model": "gittensor-model-hub/Qwen3.8-27B-NVFP4-RTX5090", "rows": [{"n": 1, "wall_s": 2.22, "total_tok": 653, "agg_tok_s": 294.6, "per_req_tok_s": 294.7, "p50_lat_s": 2.22, "p95_lat_s": 2.22}, {"n": 2, "wall_s": 2.38, "total_tok": 1470, "agg_tok_s": 616.6, "per_req_tok_s": 308.4, "p50_lat_s": 2.38, "p95_lat_s": 2.38}, {"n": 4, "wall_s": 4.22, "total_tok": 2772, "agg_tok_s": 656.9, "per_req_tok_s": 215.1, "p50_lat_s": 4.22, "p95_lat_s": 4.22}, {"n": 8, "wall_s": 7.96, "total_tok": 5208, "agg_tok_s": 653.9, "per_req_tok_s": 130.8, "p50_lat_s": 5.99, "p95_lat_s": 7.96}]}