| { | |
| "concurrency": 1, | |
| "requests": 8, | |
| "output_tokens_per_request": 512, | |
| "wall_seconds": 38.2679453890305, | |
| "aggregate_output_tps": 107.03475084330286, | |
| "median_decode_tps": 107.2369849327996, | |
| "median_ttft_seconds": 0.07549251493765041, | |
| "sampling": "greedy; ignore_eos for this throughput test only", | |
| "speculation_counter_deltas": { | |
| "vllm:spec_decode_num_drafts_total": 1316.0, | |
| "vllm:spec_decode_num_draft_tokens_total": 3948.0, | |
| "vllm:spec_decode_num_accepted_tokens_total": 2780.0 | |
| }, | |
| "decode_tps_note": "Client stream timing estimate; speculative decoding may deliver several tokens in one event." | |
| } | |