[Usage]: How can i user scheduler with OpenAI Compatible Server

[anhnh2002]

Your current environment

PyTorch version: 2.4.0+cu121
Is debug build: False
CUDA used to build PyTorch: 12.1
ROCM used to build PyTorch: N/A

OS: Ubuntu 20.04.6 LTS (x86_64)
GCC version: (Ubuntu 9.4.0-1ubuntu1~20.04.2) 9.4.0
Clang version: Could not collect
CMake version: version 3.29.3
Libc version: glibc-2.31

Python version: 3.10.14 (main, May  6 2024, 19:42:50) [GCC 11.2.0] (64-bit runtime)
Python platform: Linux-5.4.0-193-generic-x86_64-with-glibc2.31
Is CUDA available: True
CUDA runtime version: 10.1.243
CUDA_MODULE_LOADING set to: LAZY
GPU models and configuration: 
GPU 0: NVIDIA A100-PCIE-40GB
GPU 1: NVIDIA A100-PCIE-40GB

Nvidia driver version: 535.183.01
cuDNN version: Probably one of the following:
/usr/lib/x86_64-linux-gnu/libcudnn.so.8.9.7
/usr/lib/x86_64-linux-gnu/libcudnn_adv_infer.so.8.9.7
/usr/lib/x86_64-linux-gnu/libcudnn_adv_train.so.8.9.7
/usr/lib/x86_64-linux-gnu/libcudnn_cnn_infer.so.8.9.7
/usr/lib/x86_64-linux-gnu/libcudnn_cnn_train.so.8.9.7
/usr/lib/x86_64-linux-gnu/libcudnn_ops_infer.so.8.9.7
/usr/lib/x86_64-linux-gnu/libcudnn_ops_train.so.8.9.7
HIP runtime version: N/A
MIOpen runtime version: N/A
Is XNNPACK available: True

CPU:
Architecture:                       x86_64
CPU op-mode(s):                     32-bit, 64-bit
Byte Order:                         Little Endian
Address sizes:                      40 bits physical, 57 bits virtual
CPU(s):                             52
On-line CPU(s) list:                0-51
Thread(s) per core:                 1
Core(s) per socket:                 52
Socket(s):                          1
NUMA node(s):                       1
Vendor ID:                          GenuineIntel
CPU family:                         6
Model:                              134
Model name:                         Intel Xeon Processor (Icelake)
Stepping:                           0
CPU MHz:                            2194.848
BogoMIPS:                           4389.69
Virtualisation:                     VT-x
Hypervisor vendor:                  KVM
Virtualisation type:                full
L1d cache:                          1.6 MiB
L1i cache:                          1.6 MiB
L2 cache:                           208 MiB
L3 cache:                           16 MiB
NUMA node0 CPU(s):                  0-51
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:      Vulnerable: Clear CPU buffers attempted, no microcode; SMT Host state unknown
Vulnerability Retbleed:             Not affected
Vulnerability Spec store bypass:    Mitigation; Speculative Store Bypass disabled via prctl and seccomp
Vulnerability Spectre v1:           Mitigation; usercopy/swapgs barriers and __user pointer sanitization
Vulnerability Spectre v2:           Mitigation; Enhanced / Automatic IBRS; IBPB conditional; RSB filling; PBRSB-eIBRS Not affected; BHI SW loop, KVM SW loop
Vulnerability Srbds:                Not affected
Vulnerability Tsx async abort:      Mitigation; TSX disabled
Flags:                              fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush mmx fxsr sse sse2 ss ht syscall nx pdpe1gb rdtscp lm constant_tsc rep_good nopl xtopology cpuid tsc_known_freq pni pclmulqdq vmx ssse3 fma cx16 pcid sse4_1 sse4_2 x2apic movbe popcnt tsc_deadline_timer aes xsave avx f16c rdrand hypervisor lahf_lm abm 3dnowprefetch cpuid_fault invpcid_single ssbd ibrs ibpb stibp ibrs_enhanced tpr_shadow vnmi flexpriority ept vpid ept_ad fsgsbase tsc_adjust bmi1 avx2 smep bmi2 erms invpcid avx512f avx512dq rdseed adx smap avx512ifma clflushopt clwb avx512cd sha_ni avx512bw avx512vl xsaveopt xsavec xgetbv1 xsaves wbnoinvd arat avx512vbmi umip pku ospke avx512_vbmi2 gfni vaes vpclmulqdq avx512_vnni avx512_bitalg avx512_vpopcntdq rdpid md_clear arch_capabilities

Versions of relevant libraries:
[pip3] flake8==7.0.0
[pip3] flake8-bugbear==24.4.26
[pip3] flashinfer==0.1.2+cu124torch2.4
[pip3] mypy-extensions==1.0.0
[pip3] numpy==1.26.2
[pip3] nvidia-nccl-cu12==2.20.5
[pip3] onnxruntime==1.18.1
[pip3] sentence-transformers==3.0.1
[pip3] torch==2.4.0
[pip3] torch-stoi==0.2.1
[pip3] torchaudio==2.4.0
[pip3] torchtext==0.18.0
[pip3] torchvision==0.19.0
[pip3] transformers==4.44.2
[pip3] triton==3.0.0
[conda] flashinfer                0.1.2+cu124torch2.4          pypi_0    pypi
[conda] numpy                     1.26.4                   pypi_0    pypi
[conda] nvidia-nccl-cu12          2.20.5                   pypi_0    pypi
[conda] sentence-transformers     3.0.1                    pypi_0    pypi
[conda] torch                     2.4.0                    pypi_0    pypi
[conda] torch-stoi                0.2.1                    pypi_0    pypi
[conda] torchaudio                2.4.0                    pypi_0    pypi
[conda] torchtext                 0.18.0                   pypi_0    pypi
[conda] torchvision               0.19.0                   pypi_0    pypi
[conda] transformers              4.44.2                   pypi_0    pypi
[conda] triton                    3.0.0                    pypi_0    pypi
ROCM Version: Could not collect
Neuron SDK Version: N/A
vLLM Version: N/A
vLLM Build Flags:
CUDA Archs: Not Set; ROCm: Disabled; Neuron: Disabled
GPU Topology:
GPU0    GPU1    CPU Affinity    NUMA Affinity   GPU NUMA ID
GPU0     X      NV12    0-51    0               N/A
GPU1    NV12     X      0-51    0               N/A

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

How would you like to use vllm

I am currently serving an LLM via the vLLM as an OpenAI Compatible Server. I make API calls as shown below:

completion = llm_client.chat.completions.create(
    model=LLM_MODEL_NAME,
    messages=messages,
    top_p=0.5
)
response = completion.choices[0].message.content.strip()

I would like to introduce a scheduler or prioritize requests in my API calls. Is there a way to specify a priority or use a scheduler directly in the API call, similar to the following?

completion = llm_client.chat.completions.create(
    model=LLM_MODEL_NAME,
    messages=messages,
    top_p=0.5,
    priority=1
)
response = completion.choices[0].message.content.strip()

Could you provide guidance on implementing this feature or suggest an alternative approach?

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]

I think this isn't a primary concern for vLLM. It may be better to implement a separate load-balancing / scheduling layer on top of the vLLM server.

[anhnh2002]

I think this isn't a primary concern for vLLM. It may be better to implement a separate load-balancing / scheduling layer on top of the vLLM server.

I am using FastAPI wrapped on a vLLM server. I have tried searching for documents to create a priority queue for FastAPI but haven't found any. Do you have any suggestions for me?

[DarkLight1337]

I haven't done this myself, so I wouldn't consider myself qualified to provide concrete suggestions. You can check out #4873 for more details.

[DarkLight1337]

Update: I found #5958 which may serve your purpose but currently the PR only considers offline inference.

[anhnh2002]

Update: I found #5958 which may serve your purpose but currently the PR only considers offline inference.

thanks a lot

[DarkLight1337]

#5958 has been merged. It should be quite straightforward to expose this argument to the OpenAI-compatible API.