[Bug]: new beam search implementation ignores stop conditions

[nFunctor]

Your current environment

The output of `python collect_env.py`

Collecting environment information...
/usr/local/lib/python3.10/dist-packages/vllm/connections.py:8: RuntimeWarning: Failed to read commit hash:
No module named 'vllm._version'
  from vllm.version import __version__ as VLLM_VERSION
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 22.04.3 LTS (x86_64)
GCC version: (Ubuntu 11.4.0-1ubuntu1~22.04) 11.4.0
Clang version: Could not collect
CMake version: Could not collect
Libc version: glibc-2.35

Python version: 3.10.12 (main, Nov 20 2023, 15:14:05) [GCC 11.4.0] (64-bit runtime)
Python platform: Linux-6.5.0-35-generic-x86_64-with-glibc2.35
Is CUDA available: True
CUDA runtime version: 12.1.105
CUDA_MODULE_LOADING set to: LAZY
GPU models and configuration: GPU 0: NVIDIA L40S
Nvidia driver version: 550.54.14
cuDNN version: Could not collect
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
Address sizes:                      48 bits physical, 48 bits virtual
Byte Order:                         Little Endian
CPU(s):                             128
On-line CPU(s) list:                0-127
Vendor ID:                          AuthenticAMD
Model name:                         AMD EPYC 75F3 32-Core Processor
CPU family:                         25
Model:                              1
Thread(s) per core:                 2
Core(s) per socket:                 32
Socket(s):                          2
Stepping:                           1
Frequency boost:                    enabled
CPU max MHz:                        4041.8450
CPU min MHz:                        1500.0000
BogoMIPS:                           5900.35
Flags:                              fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush mmx fxsr sse sse2 ht syscall nx mmxext fxsr_opt pdpe1gb rdtscp lm constant_tsc rep_good nopl nonstop_tsc cpuid extd_apicid aperfmperf rapl pni pclmulqdq monitor ssse3 fma cx16 pcid sse4_1 sse4_2 movbe popcnt aes xsave avx f16c rdrand lahf_lm cmp_legacy svm extapic cr8_legacy abm sse4a misalignsse 3dnowprefetch osvw ibs skinit wdt tce topoext perfctr_core perfctr_nb bpext perfctr_llc mwaitx cpb cat_l3 cdp_l3 invpcid_single hw_pstate ssbd mba ibrs ibpb stibp vmmcall fsgsbase bmi1 avx2 smep bmi2 invpcid cqm rdt_a rdseed adx smap clflushopt clwb sha_ni xsaveopt xsavec xgetbv1 xsaves cqm_llc cqm_occup_llc cqm_mbm_total cqm_mbm_local clzero irperf xsaveerptr rdpru wbnoinvd amd_ppin brs arat npt lbrv svm_lock nrip_save tsc_scale vmcb_clean flushbyasid decodeassists pausefilter pfthreshold v_vmsave_vmload vgif v_spec_ctrl umip pku ospke vaes vpclmulqdq rdpid overflow_recov succor smca
Virtualization:                     AMD-V
L1d cache:                          2 MiB (64 instances)
L1i cache:                          2 MiB (64 instances)
L2 cache:                           32 MiB (64 instances)
L3 cache:                           512 MiB (16 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: Mitigation; Safe RET
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; Retpolines; IBPB conditional; IBRS_FW; STIBP always-on; RSB filling; PBRSB-eIBRS Not affected; BHI Not affected
Vulnerability Srbds:                Not affected
Vulnerability Tsx async abort:      Not affected

Versions of relevant libraries:
[pip3] numpy==1.26.3
[pip3] nvidia-cublas-cu12==12.1.3.1
[pip3] nvidia-cuda-cupti-cu12==12.1.105
[pip3] nvidia-cuda-nvrtc-cu12==12.1.105
[pip3] nvidia-cuda-runtime-cu12==12.1.105
[pip3] nvidia-cudnn-cu12==9.1.0.70
[pip3] nvidia-cufft-cu12==11.0.2.54
[pip3] nvidia-curand-cu12==10.3.2.106
[pip3] nvidia-cusolver-cu12==11.4.5.107
[pip3] nvidia-cusparse-cu12==12.1.0.106
[pip3] nvidia-ml-py==12.560.30
[pip3] nvidia-nccl-cu12==2.20.5
[pip3] nvidia-nvjitlink-cu12==12.3.101
[pip3] nvidia-nvtx-cu12==12.1.105
[pip3] pyzmq==24.0.1
[pip3] torch==2.4.0
[pip3] torchaudio==2.2.0
[pip3] torchvision==0.19.0
[pip3] transformers==4.45.2
[pip3] triton==3.0.0
[conda] Could not collect
ROCM Version: Could not collect
Neuron SDK Version: N/A
vLLM Version: dev
vLLM Build Flags:
CUDA Archs: Not Set; ROCm: Disabled; Neuron: Disabled
GPU Topology:
GPU0    NIC0    NIC1    CPU Affinity    NUMA Affinity   GPU NUMA ID
GPU0     X      SYS     SYS     32-63,96-127    1               N/A
NIC0    SYS      X      PIX
NIC1    SYS     PIX      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

Model Input Dumps

No response

🐛 Describe the bug

Command used to serve vllm serve $MODEL --port $PORT --gpu-memory-utilization 0.97 --max-model-len 4096 --dtype "auto" --trust-remote-code --enable-prefix-caching --disable-log-requests --num_scheduler-steps 1 --max-num-seqs 128

Tried with MODEL="hugging-quants/Meta-Llama-3.1-70B-Instruct-AWQ-INT4" and "hugging-quants/Meta-Llama-3.1-8B-Instruct-AWQ-INT4", vllm served via OpenAI, nightly build + python_only_dev.py

Both models ignore stop conditions specified in client.completions.create when run with beam search. It appears to me that they actually generate up to the requested number of max_tokens regardless whether tokens in stop or <eot_id> were produced. For an example, run the code below.

NB: I am using n=4 for beam search since the API commits seem to suggest that this is how you pass the number of beams now.

# ... assume you have a client and a model_id

tokenizer = AutoTokenizer.from_pretrained(model_id)

# Create a simple conversation prompt
system_prompt = "You are a helpful assistant that continues number sequences."
user_prompt = "Please continue this sequence: one, two, three, four, five..."

prompt = tokenizer.apply_chat_template([
    {"role": "system", "content": system_prompt},
    {"role": "user", "content": user_prompt}
], tokenize=False, add_generation_prompt=True)

# First run
result_1 = client.completions.create(
    model=model_id,
    prompt=prompt,
    max_tokens=500,
    temperature=0.0,
    top_p=1.0,
    stop=["six"],
)

print("First run result:")
print(result_1.choices[0].text)
print("'six' in result_1:", "six" in result_1.choices[0].text)

# Second run
result_2 = client.completions.create(
    model=model_id,
    prompt=prompt,
    max_tokens=500,
    temperature=0.0,
    top_p=1.0,
    n=4,
    stop=["six"],
    extra_body={"use_beam_search": True}
)

print("\nSecond run result:")
print(result_2.choices[0].text)
print("'six' in result_2:", "six" in result_2.choices[0].text)

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.

[nFunctor]

Addendum, after going through the API code and understanding things better. Not so much of a bug but rather a lack of implementation (shame on me, I've seen that remark...), neither the class BeamSearchParams nor its initialisation from CompletionRequest allow for stop, and it is not implemented in the beam search method. I am currently trying to figure out what should be the logic of stop check (thought of simply passing everything to a generation step but ignore EOS seems to be treated outside of sampling params for beam generation), but if someone goes ahead with it I will be happy to see.

[github-actions]

This issue has been automatically marked as stale because it has not had any activity within 90 days. It will be automatically closed if no further activity occurs within 30 days. Leave a comment if you feel this issue should remain open. Thank you!

[github-actions]

This issue has been automatically closed due to inactivity. Please feel free to reopen if you feel it is still relevant. Thank you!