()
| 100 | llm.generate(dummy_prompts, sampling_params=sampling_params, use_tqdm=False, stream=True) |
| 101 | |
| 102 | def run_to_completion(): |
| 103 | start_time = time.perf_counter() |
| 104 | llm_generate() |
| 105 | end_time = time.perf_counter() |
| 106 | latency = end_time - start_time |
| 107 | return latency |
| 108 | |
| 109 | print("Warming up...") |
| 110 | for _ in tqdm(range(args.num_iters_warmup), desc="Warmup iterations"): |