Offline Throughput Benchmark Summary

• model: gpt2
• dtype: fp16
• num_prompts: 128
• input_len: 256
• output_len: 64
• sampling: temperature=0.7, top_p=0.9, top_k=50
• wall_s: 18.06826117899618
• run_start_time: 2025-12-30T18:51:52+08:00
• run_end_time: 2026-01-01T08:12:29+08:00
• versions: git_rev=a547c7c, rosellm=0.1.0, vllm=0.7.2, sglang=0.4.6, tensorrt_llm=1.1.0, torch=2.6.0, transformers=4.51.3, python=3.11.11

backend	req/s	output tok/s	total tok/s	total latency (s)
roseinfer	201.13	12872.49	64362.44	0.636
roseinfer (+chunk bucket)	202.57	12964.54	64822.70	0.632
roseinfer (in-proc)	204.11	13062.83	65314.13	0.627
SGLang	243.20	15564.48	77822.40	0.526
TensorRT-LLM	248.69	15916.24	79581.21	0.515
vLLM	140.44	8988.14	44940.70	0.911