benchmark_args=Namespace(backend='sglang', base_url=None, host='0.0.0.0', port=30000, ready_check_timeout_sec=60, dataset_name='random', dataset_path='', dataset_offset=0, agentic_max_turns=None, speed_bench_category=None, speed_bench_output_len=512, model=None, served_model_name=None, tokenizer=None, num_prompts=16, sharegpt_output_len=None, sharegpt_context_len=None, random_input_len=512, random_output_len=128, random_range_ratio=1.0, image_count=1, image_resolution='1080p', random_image_count=False, image_format='jpeg', image_content='random', request_rate=inf, use_trace_timestamps=False, max_concurrency=1, output_file='/private/tmp/sglang-research/capture/out/bench-random-c1.jsonl', output_details=False, print_requests=False, disable_tqdm=False, disable_stream=False, return_logprob=False, top_logprobs_num=0, token_ids_logprob=None, logprob_start_len=-1, return_routed_experts=False, cache_report=False, seed=7, disable_ignore_eos=False, temperature=0.0, top_p=1.0, extra_request_body=None, apply_chat_template=False, profile=False, plot_throughput=False, profile_activities=['CPU', 'GPU'], profile_start_step=None, profile_steps=None, profile_num_steps=None, profile_by_stage=False, profile_stages=None, profile_output_dir=None, profile_prefix=None, lora_name=None, lora_request_distribution='uniform', lora_zipf_alpha=1.5, prompt_suffix='', pd_separated=False, profile_prefill_url=None, profile_decode_url=None, flush_cache=False, flush_cache_timeout=60.0, warmup_requests=1, tokenize_prompt=False, gsp_num_groups=64, gsp_prompts_per_group=16, gsp_system_prompt_len=2048, gsp_question_len=128, gsp_output_len=256, gsp_range_ratio=1.0, gsp_fast_prepare=False, gsp_send_routing_key=False, gsp_num_turns=1, gsp_ordered=False, gsp_group_distribution='uniform', gsp_zipf_alpha=None, mooncake_slowdown_factor=1.0, mooncake_num_rounds=1, mooncake_workload='conversation', fake_prefill=False, tag=None, header=None, rust_client=False, rust_worker_threads=None)
Waiting up to 60s for http://0.0.0.0:30000/v1/models to become ready...
Server ready in 0.0s.
Namespace(backend='sglang', base_url=None, host='0.0.0.0', port=30000, ready_check_timeout_sec=60, dataset_name='random', dataset_path='', dataset_offset=0, agentic_max_turns=None, speed_bench_category=None, speed_bench_output_len=512, model='mlx-community/Qwen3-0.6B-4bit', served_model_name=None, tokenizer=None, num_prompts=16, sharegpt_output_len=None, sharegpt_context_len=None, random_input_len=512, random_output_len=128, random_range_ratio=1.0, image_count=1, image_resolution='1080p', random_image_count=False, image_format='jpeg', image_content='random', request_rate=inf, use_trace_timestamps=False, max_concurrency=1, output_file='/private/tmp/sglang-research/capture/out/bench-random-c1.jsonl', output_details=False, print_requests=False, disable_tqdm=False, disable_stream=False, return_logprob=False, top_logprobs_num=0, token_ids_logprob=None, logprob_start_len=-1, return_routed_experts=False, cache_report=False, seed=7, disable_ignore_eos=False, temperature=0.0, top_p=1.0, extra_request_body=None, apply_chat_template=False, profile=False, plot_throughput=False, profile_activities=['CPU', 'GPU'], profile_start_step=None, profile_steps=None, profile_num_steps=None, profile_by_stage=False, profile_stages=None, profile_output_dir=None, profile_prefix=None, lora_name=None, lora_request_distribution='uniform', lora_zipf_alpha=1.5, prompt_suffix='', pd_separated=False, profile_prefill_url=None, profile_decode_url=None, flush_cache=False, flush_cache_timeout=60.0, warmup_requests=1, tokenize_prompt=False, gsp_num_groups=64, gsp_prompts_per_group=16, gsp_system_prompt_len=2048, gsp_question_len=128, gsp_output_len=256, gsp_range_ratio=1.0, gsp_fast_prepare=False, gsp_send_routing_key=False, gsp_num_turns=1, gsp_ordered=False, gsp_group_distribution='uniform', gsp_zipf_alpha=None, mooncake_slowdown_factor=1.0, mooncake_num_rounds=1, mooncake_workload='conversation', fake_prefill=False, tag=None, header=None, rust_client=False, rust_worker_threads=None)

#Input tokens: 8192
#Output tokens: 2048
Starting warmup with 1 sequences...
Warmup completed with 1 sequences. Starting main benchmark run...

============ Serving Benchmark Result ============
Backend:                                 sglang    
Traffic request rate:                    inf       
Max request concurrency:                 1         
Successful requests:                     16        
Benchmark duration (s):                  11.69     
Total input tokens:                      8192      
Total input text tokens:                 8192      
Total generated tokens:                  2048      
Total generated tokens (retokenized):    2048      
Request throughput (req/s):              1.37      
Input token throughput (tok/s):          700.59    
Output token throughput (tok/s):         175.15    
Peak output token throughput (tok/s):    231.00    
Peak concurrent requests:                3         
Total token throughput (tok/s):          875.74    
Concurrency:                             0.99      
----------------End-to-End Latency----------------
Mean E2E Latency (ms):                   726.66    
Median E2E Latency (ms):                 708.96    
P90 E2E Latency (ms):                    875.20    
P95 E2E Latency (ms):                    902.33    
P99 E2E Latency (ms):                    910.82    
---------------Time to First Token----------------
Mean TTFT (ms):                          170.08    
Median TTFT (ms):                        162.98    
P90 TTFT (ms):                           253.98    
P95 TTFT (ms):                           266.29    
P99 TTFT (ms):                           270.18    
-----Time per Output Token (excl. 1st token)------
Mean TPOT (ms):                          4.38      
Median TPOT (ms):                        4.38      
P90 TPOT (ms):                           5.03      
P95 TPOT (ms):                           5.12      
P99 TPOT (ms):                           5.24      
---------------Inter-Token Latency----------------
Mean ITL (ms):                           4.38      
Median ITL (ms):                         4.14      
P90 ITL (ms):                            5.08      
P95 ITL (ms):                            5.32      
P99 ITL (ms):                            8.40      
Max ITL (ms):                            42.03     
==================================================
/private/tmp/sglang-research/sglang-v0.5.22/python/sglang/srt/layers/quantization/awq/awq.py:53: UserWarning: Only CUDA, HIP and XPU support AWQ currently.
  warnings.warn(f"Only CUDA, HIP and XPU support AWQ currently.")
/private/tmp/sglang-research/sglang-v0.5.22/python/sglang/srt/layers/quantization/gguf.py:69: UserWarning: Only CUDA, MUSA and NPU support GGUF quantization currently.
  warnings.warn(f"Only CUDA, MUSA and NPU support GGUF quantization currently.")
Warning: You are sending unauthenticated requests to the HF Hub. Please set a HF_TOKEN to enable higher rate limits and faster downloads.

  0%|          | 0/16 [00:00<?, ?it/s]
  6%|▋         | 1/16 [00:00<00:08,  1.79it/s]
 12%|█▎        | 2/16 [00:01<00:07,  1.86it/s]
 19%|█▉        | 3/16 [00:01<00:07,  1.73it/s]
 25%|██▌       | 4/16 [00:02<00:07,  1.64it/s]
 31%|███▏      | 5/16 [00:03<00:07,  1.56it/s]
 38%|███▊      | 6/16 [00:03<00:07,  1.36it/s]
 44%|████▍     | 7/16 [00:04<00:06,  1.37it/s]
 50%|█████     | 8/16 [00:05<00:05,  1.33it/s]
 56%|█████▋    | 9/16 [00:06<00:05,  1.26it/s]
 62%|██████▎   | 10/16 [00:07<00:04,  1.23it/s]
 69%|██████▉   | 11/16 [00:08<00:04,  1.23it/s]
 75%|███████▌  | 12/16 [00:08<00:03,  1.26it/s]
 81%|████████▏ | 13/16 [00:09<00:02,  1.32it/s]
 88%|████████▊ | 14/16 [00:10<00:01,  1.37it/s]
 94%|█████████▍| 15/16 [00:10<00:00,  1.42it/s]
100%|██████████| 16/16 [00:11<00:00,  1.34it/s]
100%|██████████| 16/16 [00:11<00:00,  1.38it/s]
