[Bug]: `vllm bench serve` isn't an exact replacement of benchmark_serving.py

[gshtras]

Your current environment

The output of python collect_env.py

Collecting environment information...
==============================
        System Info
==============================
OS                           : Ubuntu 22.04.5 LTS (x86_64)
GCC version                  : (Ubuntu 11.4.0-1ubuntu1~22.04) 11.4.0
Clang version                : Could not collect
CMake version                : version 4.1.0
Libc version                 : glibc-2.35

==============================
       PyTorch Info
==============================
PyTorch version              : 2.7.1+cu128
Is debug build               : False
CUDA used to build PyTorch   : 12.8
ROCM used to build PyTorch   : N/A

==============================
      Python Environment
==============================
Python version               : 3.12.11 (main, Jun  4 2025, 08:56:18) [GCC 11.4.0] (64-bit runtime)
Python platform              : Linux-6.5.0-27-generic-x86_64-with-glibc2.35

==============================
       CUDA / GPU Info
==============================
Is CUDA available            : True
CUDA runtime version         : 12.8.93
CUDA_MODULE_LOADING set to   : LAZY
GPU models and configuration :
GPU 0: NVIDIA H100 80GB HBM3
GPU 1: NVIDIA H100 80GB HBM3
GPU 2: NVIDIA H100 80GB HBM3
GPU 3: NVIDIA H100 80GB HBM3
GPU 4: NVIDIA H100 80GB HBM3
GPU 5: NVIDIA H100 80GB HBM3
GPU 6: NVIDIA H100 80GB HBM3
GPU 7: NVIDIA H100 80GB HBM3

Nvidia driver version        : 570.86.10
cuDNN version                : Could not collect
HIP runtime version          : N/A
MIOpen runtime version       : N/A
Is XNNPACK available         : True

==============================
          CPU Info
==============================
Architecture:                       x86_64
CPU op-mode(s):                     32-bit, 64-bit
Address sizes:                      52 bits physical, 57 bits virtual
Byte Order:                         Little Endian
CPU(s):                             128
On-line CPU(s) list:                0-127
Vendor ID:                          GenuineIntel
Model name:                         Intel(R) Xeon(R) Gold 6430
CPU family:                         6
Model:                              143
Thread(s) per core:                 2
Core(s) per socket:                 32
Socket(s):                          2
Stepping:                           8
CPU max MHz:                        3400.0000
CPU min MHz:                        800.0000
BogoMIPS:                           4200.00
Flags:                              fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush dts acpi mmx fxsr sse sse2 ss ht tm pbe syscall nx pdpe1gb rdtscp lm constant_tsc art arch_perfmon pebs bts rep_good nopl xtopology nonstop_tsc cpuid aperfmperf tsc_known_freq pni pclmulqdq dtes64 monitor ds_cpl vmx smx est tm2 ssse3 sdbg fma cx16 xtpr pdcm pcid dca sse4_1 sse4_2 x2apic movbe popcnt tsc_deadline_timer aes xsave avx f16c rdrand lahf_lm abm 3dnowprefetch cpuid_fault epb cat_l3 cat_l2 cdp_l3 invpcid_single intel_ppin cdp_l2 ssbd mba ibrs ibpb stibp ibrs_enhanced tpr_shadow flexpriority ept vpid ept_ad fsgsbase tsc_adjust bmi1 avx2 smep bmi2 erms invpcid cqm rdt_a avx512f avx512dq rdseed adx smap avx512ifma clflushopt clwb intel_pt avx512cd sha_ni avx512bw avx512vl xsaveopt xsavec xgetbv1 xsaves cqm_llc cqm_occup_llc cqm_mbm_total cqm_mbm_local split_lock_detect avx_vnni avx512_bf16 wbnoinvd dtherm ida arat pln pts hwp hwp_act_window hwp_epp hwp_pkg_req vnmi avx512vbmi umip pku ospke waitpkg avx512_vbmi2 gfni vaes vpclmulqdq avx512_vnni avx512_bitalg tme avx512_vpopcntdq la57 rdpid bus_lock_detect cldemote movdiri movdir64b enqcmd fsrm md_clear serialize tsxldtrk pconfig arch_lbr ibt amx_bf16 avx512_fp16 amx_tile amx_int8 flush_l1d arch_capabilities
Virtualization:                     VT-x
L1d cache:                          3 MiB (64 instances)
L1i cache:                          2 MiB (64 instances)
L2 cache:                           128 MiB (64 instances)
L3 cache:                           120 MiB (2 instances)
NUMA node(s):                       2
NUMA node0 CPU(s):                  0-31,64-95
NUMA node1 CPU(s):                  32-63,96-127
Vulnerability Gather data sampling: Not affected
Vulnerability Itlb multihit:        Not affected
Vulnerability L1tf:                 Not affected
Vulnerability Mds:                  Not affected
Vulnerability Meltdown:             Not affected
Vulnerability Mmio stale data:      Not affected
Vulnerability Retbleed:             Not affected
Vulnerability Spec rstack overflow: Not affected
Vulnerability Spec store bypass:    Mitigation; Speculative Store Bypass disabled via prctl
Vulnerability Spectre v1:           Mitigation; usercopy/swapgs barriers and __user pointer sanitization
Vulnerability Spectre v2:           Mitigation; Enhanced / Automatic IBRS, IBPB conditional, RSB filling, PBRSB-eIBRS SW sequence
Vulnerability Srbds:                Not affected
Vulnerability Tsx async abort:      Not affected

==============================
Versions of relevant libraries
==============================
[pip3] flashinfer-python==0.2.11
[pip3] numpy==2.2.6
[pip3] nvidia-cublas-cu12==12.8.3.14
[pip3] nvidia-cuda-cupti-cu12==12.8.57
[pip3] nvidia-cuda-nvrtc-cu12==12.8.61
[pip3] nvidia-cuda-runtime-cu12==12.8.57
[pip3] nvidia-cudnn-cu12==9.7.1.26
[pip3] nvidia-cufft-cu12==11.3.3.41
[pip3] nvidia-cufile-cu12==1.13.0.11
[pip3] nvidia-curand-cu12==10.3.9.55
[pip3] nvidia-cusolver-cu12==11.7.2.55
[pip3] nvidia-cusparse-cu12==12.5.7.53
[pip3] nvidia-cusparselt-cu12==0.6.3
[pip3] nvidia-nccl-cu12==2.26.2
[pip3] nvidia-nvjitlink-cu12==12.8.61
[pip3] nvidia-nvtx-cu12==12.8.55
[pip3] pyzmq==27.0.1
[pip3] torch==2.7.1+cu128
[pip3] torchaudio==2.7.1+cu128
[pip3] torchvision==0.22.1+cu128
[pip3] transformers==4.55.2
[pip3] triton==3.3.1
[conda] Could not collect

==============================
         vLLM Info
==============================
ROCM Version                 : Could not collect
vLLM Version                 : 0.10.1.1
vLLM Build Flags:
  CUDA Archs: Not Set; ROCm: Disabled
GPU Topology:
        GPU0    GPU1    GPU2    GPU3    GPU4    GPU5    GPU6    GPU7    NIC0    NIC1    CPU Affinity    NUMA Affinity   GPU NUMA ID
GPU0     X      NV18    NV18    NV18    NV18    NV18    NV18    NV18    PXB     NODE    0-31,64-95      0               N/A
GPU1    NV18     X      NV18    NV18    NV18    NV18    NV18    NV18    PIX     NODE    0-31,64-95      0               N/A
GPU2    NV18    NV18     X      NV18    NV18    NV18    NV18    NV18    NODE    PXB     0-31,64-95      0               N/A
GPU3    NV18    NV18    NV18     X      NV18    NV18    NV18    NV18    NODE    PIX     0-31,64-95      0               N/A
GPU4    NV18    NV18    NV18    NV18     X      NV18    NV18    NV18    SYS     SYS     32-63,96-127    1               N/A
GPU5    NV18    NV18    NV18    NV18    NV18     X      NV18    NV18    SYS     SYS     32-63,96-127    1               N/A
GPU6    NV18    NV18    NV18    NV18    NV18    NV18     X      NV18    SYS     SYS     32-63,96-127    1               N/A
GPU7    NV18    NV18    NV18    NV18    NV18    NV18    NV18     X      SYS     SYS     32-63,96-127    1               N/A
NIC0    PXB     PIX     NODE    NODE    SYS     SYS     SYS     SYS      X      NODE
NIC1    NODE    NODE    PXB     PIX     SYS     SYS     SYS     SYS     NODE     X

Legend:

  X    = Self
  SYS  = Connection traversing PCIe as well as the SMP interconnect between NUMA nodes (e.g., QPI/UPI)
  NODE = Connection traversing PCIe as well as the interconnect between PCIe Host Bridges within a NUMA node
  PHB  = Connection traversing PCIe as well as a PCIe Host Bridge (typically the CPU)
  PXB  = Connection traversing multiple PCIe bridges (without traversing the PCIe Host Bridge)
  PIX  = Connection traversing at most a single PCIe bridge
  NV#  = Connection traversing a bonded set of # NVLinks

NIC Legend:

  NIC0: mlx5_0
  NIC1: mlx5_1

==============================
     Environment Variables
==============================
NVIDIA_VISIBLE_DEVICES=all
NVIDIA_REQUIRE_CUDA=cuda>=12.8 brand=unknown,driver>=470,driver<471 brand=grid,driver>=470,driver<471 brand=tesla,driver>=470,driver<471 brand=nvidia,driver>=470,driver<471 brand=quadro,driver>=470,driver<471 brand=quadrortx,driver>=470,driver<471 brand=nvidiartx,driver>=470,driver<471 brand=vapps,driver>=470,driver<471 brand=vpc,driver>=470,driver<471 brand=vcs,driver>=470,driver<471 brand=vws,driver>=470,driver<471 brand=cloudgaming,driver>=470,driver<471 brand=unknown,driver>=535,driver<536 brand=grid,driver>=535,driver<536 brand=tesla,driver>=535,driver<536 brand=nvidia,driver>=535,driver<536 brand=quadro,driver>=535,driver<536 brand=quadrortx,driver>=535,driver<536 brand=nvidiartx,driver>=535,driver<536 brand=vapps,driver>=535,driver<536 brand=vpc,driver>=535,driver<536 brand=vcs,driver>=535,driver<536 brand=vws,driver>=535,driver<536 brand=cloudgaming,driver>=535,driver<536 brand=unknown,driver>=550,driver<551 brand=grid,driver>=550,driver<551 brand=tesla,driver>=550,driver<551 brand=nvidia,driver>=550,driver<551 brand=quadro,driver>=550,driver<551 brand=quadrortx,driver>=550,driver<551 brand=nvidiartx,driver>=550,driver<551 brand=vapps,driver>=550,driver<551 brand=vpc,driver>=550,driver<551 brand=vcs,driver>=550,driver<551 brand=vws,driver>=550,driver<551 brand=cloudgaming,driver>=550,driver<551 brand=unknown,driver>=560,driver<561 brand=grid,driver>=560,driver<561 brand=tesla,driver>=560,driver<561 brand=nvidia,driver>=560,driver<561 brand=quadro,driver>=560,driver<561 brand=quadrortx,driver>=560,driver<561 brand=nvidiartx,driver>=560,driver<561 brand=vapps,driver>=560,driver<561 brand=vpc,driver>=560,driver<561 brand=vcs,driver>=560,driver<561 brand=vws,driver>=560,driver<561 brand=cloudgaming,driver>=560,driver<561 brand=unknown,driver>=565,driver<566 brand=grid,driver>=565,driver<566 brand=tesla,driver>=565,driver<566 brand=nvidia,driver>=565,driver<566 brand=quadro,driver>=565,driver<566 brand=quadrortx,driver>=565,driver<566 brand=nvidiartx,driver>=565,driver<566 brand=vapps,driver>=565,driver<566 brand=vpc,driver>=565,driver<566 brand=vcs,driver>=565,driver<566 brand=vws,driver>=565,driver<566 brand=cloudgaming,driver>=565,driver<566
NCCL_VERSION=2.25.1-1
NVIDIA_DRIVER_CAPABILITIES=compute,utility
NVIDIA_PRODUCT_NAME=CUDA
VLLM_USAGE_SOURCE=production-docker-image
CUDA_VERSION=12.8.1
VLLM_DISABLE_COMPILE_CACHE=1
LD_LIBRARY_PATH=/usr/local/cuda/lib64
NCCL_CUMEM_ENABLE=0
PYTORCH_NVML_BASED_CUDA_CHECK=1
TORCHINDUCTOR_COMPILE_THREADS=1
CUDA_MODULE_LOADING=LAZY

🐛 Describe the bug

Starting with #24411 the benchmark_serving.py is deprecated in favor of vllm bench serve, but the replacement produces different results on the same config

Sample config:
On vllm/vllm-openai:latest, H100 machine
While running the server as
vllm serve meta-llama/Llama-3.1-8B-Instruct --no-enable-prefix-caching

Using
vllm bench serve --model meta-llama/Llama-3.1-8B-Instruct --percentile-metrics ttft,tpot,itl,e2el --dataset-path /models/ShareGPT_Vicuna_unfiltered/ShareGPT_V3_unfiltered_cleaned_split.json --max-concurrency 64 --num-prompts 640 --ignore-eos --sharegpt-output-len 512
produces

============ Serving Benchmark Result ============
Successful requests:                     640
Maximum request concurrency:             64
Benchmark duration (s):                  29.23
Total input tokens:                      654052
Total generated tokens:                  81920
Request throughput (req/s):              21.89
Output token throughput (tok/s):         2802.42
Total Token throughput (tok/s):          25177.00
---------------Time to First Token----------------
Mean TTFT (ms):                          579.39
Median TTFT (ms):                        582.78
P99 TTFT (ms):                           1584.13
-----Time per Output Token (excl. 1st token)------
Mean TPOT (ms):                          18.41
Median TPOT (ms):                        18.33
P99 TPOT (ms):                           22.50
---------------Inter-token Latency----------------
Mean ITL (ms):                           18.41
Median ITL (ms):                         11.13
P99 ITL (ms):                            198.59
----------------End-to-end Latency----------------
Mean E2EL (ms):                          2917.14
Median E2EL (ms):                        2909.60
P99 E2EL (ms):                           3667.08
==================================================

While using benchmark_Serving from a commit prior to #24411
python3 benchmark_serving.py --model meta-llama/Llama-3.1-8B-Instruct --percentile-metrics ttft,tpot,itl,e2el --dataset-path /models/ShareGPT_Vicuna_unfiltered/ShareGPT_V3_unfiltered_cleaned_split.json --max-concurrency 64 --num-prompts 640 --ignore-eos --sharegpt-output-len 512
produces

============ Serving Benchmark Result ============
Successful requests:                     640
Maximum request concurrency:             64
Benchmark duration (s):                  51.11
Total input tokens:                      136352
Total generated tokens:                  327680
Request throughput (req/s):              12.52
Output token throughput (tok/s):         6411.31
Total Token throughput (tok/s):          9079.15
---------------Time to First Token----------------
Mean TTFT (ms):                          246.70
Median TTFT (ms):                        275.99
P99 TTFT (ms):                           415.40
-----Time per Output Token (excl. 1st token)------
Mean TPOT (ms):                          9.50
Median TPOT (ms):                        9.47
P99 TPOT (ms):                           9.96
---------------Inter-token Latency----------------
Mean ITL (ms):                           9.50
Median ITL (ms):                         9.27
P99 ITL (ms):                            12.55
----------------End-to-end Latency----------------
Mean E2EL (ms):                          5102.90
Median E2EL (ms):                        5097.46
P99 E2EL (ms):                           5240.58
==================================================

The main difference seems to be in the amount of tokens grabbed from the dataset for the input: 654052 vs 136352

Before submitting a new issue...

• Make sure you already searched for relevant issues, and asked the chatbot living at the bottom right corner of the documentation page, which can answer lots of frequently asked questions.

[DarkLight1337]

cc @mgoin @yeqcharlotte

[yeqcharlotte]

Will take a look and reproduce.

Looking at the first result, it seems output len is 81920/640 = 128?

[ycma8]

Looks like the issue comes from the default value of --dataset-name.
Old version defaults to sharegpt, but the new version defaults to random.
Not sure if that change was intentional — if not, I can send a quick PR to fix it 🙂

[yeqcharlotte]

Looks like the issue comes from the default value of --dataset-name. Old version defaults to sharegpt, but the new version defaults to random. Not sure if that change was intentional — if not, I can send a quick PR to fix it 🙂

i think random dataset is a better default. @gshtras -- will please use add --dataset-name sharegpt to your command. also fixing the cli error message: #24819.

[gshtras]

Providing a --dataset-path was forcing the old benchmark to use sharegpt instead of its default random. Seems like in the new one it's ignored, breaking existing workflows

[Ramshankar07]

is it a good choice, to add all the choices in validation which infers the dataset path and selects the choice like

if not dataset_path:
        return 'random'
    path_lower = dataset_path.lower()
    if 'sharegpt' in path_lower or 'vicuna' in path_lower:
        return 'sharegpt'
    elif 'sonnet' in path_lower:
        return 'sonnet'
    elif 'burstgpt' in path_lower:
        return 'burstgpt'
    elif 'spec_bench' in path_lower or 'spec-bench' in path_lower:
        return 'spec_bench'

P.S: I'm just a beginner

[Ramshankar07]

I've added this in my repo, If you think this ain't the case. Please help me giving some info to contribute