[Bug]: VLM same chat different image results in serving_chat.py:238 Error in loading image data:

[thealmightygrant]

Your current environment

PyTorch version: 2.3.0+cu121
Is debug build: False
CUDA used to build PyTorch: 12.1
ROCM used to build PyTorch: N/A

OS: Ubuntu 22.04.4 LTS (x86_64)
GCC version: (Ubuntu 11.4.0-1ubuntu1~22.04) 11.4.0
Clang version: Could not collect
CMake version: version 3.29.6
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-6.1.75+-x86_64-with-glibc2.35
Is CUDA available: True
CUDA runtime version: 12.4.131
CUDA_MODULE_LOADING set to: LAZY
GPU models and configuration: GPU 0: NVIDIA A100-SXM4-40GB
Nvidia driver version: 535.161.07
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:                      46 bits physical, 48 bits virtual
Byte Order:                         Little Endian
CPU(s):                             12
On-line CPU(s) list:                0-11
Vendor ID:                          GenuineIntel
Model name:                         Intel(R) Xeon(R) CPU @ 2.20GHz
CPU family:                         6
Model:                              85
Thread(s) per core:                 2
Core(s) per socket:                 6
Socket(s):                          1
Stepping:                           7
BogoMIPS:                           4400.40
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 nonstop_tsc cpuid tsc_known_freq pni pclmulqdq ssse3 fma cx16 pcid sse4_1 sse4_2 x2apic movbe popcnt aes xsave avx f16c rdrand hypervisor lahf_lm abm 3dnowprefetch invpcid_single ssbd ibrs ibpb stibp ibrs_enhanced fsgsbase tsc_adjust bmi1 hle avx2 smep bmi2 erms invpcid rtm mpx avx512f avx512dq rdseed adx smap clflushopt clwb avx512cd avx512bw avx512vl xsaveopt xsavec xgetbv1 xsaves arat avx512_vnni md_clear arch_capabilities
Hypervisor vendor:                  KVM
Virtualization type:                full
L1d cache:                          192 KiB (6 instances)
L1i cache:                          192 KiB (6 instances)
L2 cache:                           6 MiB (6 instances)
L3 cache:                           38.5 MiB (1 instance)
NUMA node(s):                       1
NUMA node0 CPU(s):                  0-11
Vulnerability Gather data sampling: Not affected
Vulnerability Itlb multihit:        Not affected
Vulnerability L1tf:                 Not affected
Vulnerability Mds:                  Mitigation; Clear CPU buffers; SMT Host state unknown
Vulnerability Meltdown:             Not affected
Vulnerability Mmio stale data:      Vulnerable: Clear CPU buffers attempted, no microcode; SMT Host state unknown
Vulnerability Retbleed:             Mitigation; Enhanced IBRS
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 IBRS, IBPB conditional, RSB filling, PBRSB-eIBRS SW sequence
Vulnerability Srbds:                Not affected
Vulnerability Tsx async abort:      Mitigation; Clear CPU buffers; SMT Host state unknown

Versions of relevant libraries:
[pip3] mypy==1.9.0
[pip3] mypy-extensions==1.0.0
[pip3] numpy==1.26.4
[pip3] nvidia-nccl-cu12==2.20.5
[pip3] sentence-transformers==3.0.1
[pip3] torch==2.3.0
[pip3] torchvision==0.18.0
[pip3] transformers==4.41.2
[pip3] triton==2.3.0
[conda] Could not collect
ROCM Version: Could not collect
Neuron SDK Version: N/A
vLLM Version: 0.5.0.post1
vLLM Build Flags:
CUDA Archs: 7.0 7.5 8.0 8.6 8.9 9.0+PTX; ROCm: Disabled; Neuron: Disabled
GPU Topology:
GPU0    CPU Affinity    NUMA Affinity   GPU NUMA ID
GPU0     X  0-11    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

🐛 Describe the bug

Hi y'all, I am accessing Phi 3 Vision via the API's v1/chat/completions endpoint. It seems that if the connection remains open, I am getting the following error after making two requests with two different images:

ERROR 06-25 21:33:32 serving_chat.py:238] Error in loading image data:
/usr/local/lib/python3.10/dist-packages/vllm/entrypoints/openai/api_server.py:103: RuntimeWarning: coroutine 'async_get_and_parse_image' was never awaited
  generator = await openai_serving_chat.create_chat_completion(
RuntimeWarning: Enable tracemalloc to get the object allocation traceback
INFO:     10.24.17.222:0 - "POST /v1/chat/completions HTTP/1.1" 400 Bad Request

If I go ahead and ensure that each request is made on a new connection, then this error does not occur. My only guess is that this line should have an await added to it. I will open a PR referencing this with that change, let me know when you can if you all agree.

[thealmightygrant]

Hi y'all, I looked into this a bit more. I am using open-webui, and I believe what is occurring is that multiple images are being passed into the model as part of providing context from earlier conversations. I am not 100% sure about this though. If that is the case, then this condition is being hit every time and it is nothing more than that there is no support for multi-image inference right now in vLLM.

[ywang96]

Hey @thealmightygrant! If you context does have multiple images (i.e, multiple image_url) in the messages, then this is indeed not supported currently by vLLM. (and in fact, I believe all the vision language models we support on vLLM today do not support multi-image themselves either). Thanks!

[thealmightygrant]

Closing, as I confirmed this is what is occurring. Please reference #4194 for ongoing refactoring of visual language models.

[thealmightygrant]

Oh btw, phi 3 vision is working great for us! Thanks for the hard work @ywang96!

[ywang96]

Oh btw, phi 3 vision is working great for us! Thanks for the hard work @ywang96!

Very glad to hear that! :)