[Bug]: Gemma3 multimodal always returns the same response for different image input request #1113
Description
Your current environment
The output of commands above
Collecting environment information...
uv is set
==============================
System Info
==============================
OS : Ubuntu 22.04.5 LTS (x86_64)
GCC version : (Ubuntu 11.4.0-1ubuntu1~22.04.2) 11.4.0
Clang version : Could not collect
CMake version : version 4.1.2
Libc version : glibc-2.35
==============================
PyTorch Info
==============================
PyTorch version : 2.8.0+cu128
Is debug build : False
CUDA used to build PyTorch : 12.8
ROCM used to build PyTorch : N/A
==============================
Python Environment
==============================
Python version : 3.11.13 (main, Jun 4 2025, 08:57:29) [GCC 11.4.0] (64-bit runtime)
Python platform : Linux-6.8.0-1015-gcp-x86_64-with-glibc2.35
==============================
CUDA / GPU Info
==============================
Is CUDA available : False
CUDA runtime version : No CUDA
CUDA_MODULE_LOADING set to : N/A
GPU models and configuration : No CUDA
Nvidia driver version : No CUDA
cuDNN version : No CUDA
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): 180
On-line CPU(s) list: 0-179
Vendor ID: AuthenticAMD
Model name: AMD EPYC 9B14
CPU family: 25
Model: 17
Thread(s) per core: 2
Core(s) per socket: 90
Socket(s): 1
Stepping: 1
BogoMIPS: 5199.99
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 tsc_known_freq pni pclmulqdq ssse3 fma cx16 pcid sse4_1 sse4_2 x2apic movbe popcnt aes xsave avx f16c rdrand hypervisor lahf_lm cmp_legacy cr8_legacy abm sse4a misalignsse 3dnowprefetch osvw topoext ssbd ibrs ibpb stibp vmmcall 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 avx512_bf16 clzero xsaveerptr wbnoinvd arat avx512vbmi umip avx512_vbmi2 gfni vaes vpclmulqdq avx512_vnni avx512_bitalg avx512_vpopcntdq rdpid fsrm
Hypervisor vendor: KVM
Virtualization type: full
L1d cache: 2.8 MiB (90 instances)
L1i cache: 2.8 MiB (90 instances)
L2 cache: 90 MiB (90 instances)
L3 cache: 384 MiB (12 instances)
NUMA node(s): 1
NUMA node0 CPU(s): 0-179
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 Reg file data sampling: 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==2.3.4
[pip3] nvidia-cublas-cu12==12.8.4.1
[pip3] nvidia-cuda-cupti-cu12==12.8.90
[pip3] nvidia-cuda-nvrtc-cu12==12.8.93
[pip3] nvidia-cuda-runtime-cu12==12.8.90
[pip3] nvidia-cudnn-cu12==9.10.2.21
[pip3] nvidia-cufft-cu12==11.3.3.83
[pip3] nvidia-cufile-cu12==1.13.1.3
[pip3] nvidia-curand-cu12==10.3.9.90
[pip3] nvidia-cusolver-cu12==11.7.3.90
[pip3] nvidia-cusparse-cu12==12.5.8.93
[pip3] nvidia-cusparselt-cu12==0.7.1
[pip3] nvidia-nccl-cu12==2.27.3
[pip3] nvidia-nvjitlink-cu12==12.8.93
[pip3] nvidia-nvtx-cu12==12.8.90
[pip3] pyzmq==27.1.0
[pip3] torch==2.8.0
[pip3] torchax==0.0.7
[pip3] torchvision==0.23.0
[pip3] transformers==4.57.1
[pip3] triton==3.4.0
[conda] Could not collect
==============================
vLLM Info
==============================
ROCM Version : Could not collect
vLLM Version : 0.11.1
vLLM Build Flags:
CUDA Archs: Not Set; ROCm: Disabled
GPU Topology:
Could not collect
==============================
Environment Variables
==============================
PYTORCH_NVML_BASED_CUDA_CHECK=1
TORCHINDUCTOR_COMPILE_THREADS=1
🐛 Describe the bug
I am testing the multi modal functionality of gemma3-27b model using below. However, it appears to return the same response even if different image url is passed in the client request.
I can confirm that another multimodal (qwen2.5 VL 7b) is working fine. Using the same client requests with different image urls, I get expected and different responses.
Just gemma3 multimodal seem to be behaving strangely.
gemma3 vLLM server command
vllm serve google/gemma-3-27b-it --download-dir /data --seed 42 --max-model-len 16384 --max-num-batched-tokens 16384 --gpu-memory-utilization 0.95 --tensor-parallel-size 4 &
Should return a response about a yellow rubber ducky
(env) ikwak_google_com@t1v-n-bfb3e0ee-w-0:~$ curl -X POST "http://localhost:8000/v1/chat/completions" -H "Content-Type: application/json" -d '{
"model": "google/gemma-3-27b-it",
"messages": [
{
"role": "user",
"content": [
{
"type": "text",
"text": "What is in this image?"
},
{
"type": "image_url",
"image_url": {
"url": "https://encrypted-tbn0.gstatic.com/images?q=tbn:ANd9GcSbS7Kx2nyYRKLeNA7AqJyI6NMox_VC9mylKg&s"
}
}
]
}
],
"max_tokens": 100
}'
(APIServer pid=389703) INFO: 127.0.0.1:54440 - "POST /v1/chat/completions HTTP/1.1" 200 OK
{"id":"chatcmpl-789503ed3220405e9c3b8bcd0821ab00","object":"chat.completion","created":1763240620,"model":"google/gemma-3-27b-it","choices":[{"index":0,"message":{"role":"assistant","content":"The image consists entirely of a repeating pattern of a symbol that appears to be a Devanagari letter. Specifically, it's the letter \"दा\" (dā). \n\nIt's a densely packed, abstract image made up solely of this script. Without further context, it's hard to say *why* this image was created. It could be a design element, a representation of a word or phrase, or a visual exploration of the letter itself.","refusal":null,"annotations":null,"audio":null,"function_call":null,"tool_calls":[],"reasoning_content":null},"logprobs":null,"finish_reason":"stop","stop_reason":106,"token_ids":null}],"service_tier":null,"system_fingerprint":null,"usage":{"prompt_tokens":275,"total_tokens":372,"completion_tokens":97,"prompt_tokens_details":null},"prompt_logprobs":null,"prompt_token_ids":null,"kv_transfer_params":null}
Should return a response about a puppy
(env) ikwak_google_com@t1v-n-bfb3e0ee-w-0:~$ curl -X POST "http://localhost:8000/v1/chat/completions" -H "Content-Type: application/json" -d '{
"model": "google/gemma-3-27b-it",
"messages": [
{
"role": "user",
"content": [
{
"type": "text",
"text": "What is in this image?"
},
{
"type": "image_url",
"image_url": {
"url": "https://www.nylabone.com/-/media/project/oneweb/nylabone/images/dog101/9-fluffy-dog-breeds/pomeranian-sitting-in-grass.jpg"
}
}
]
}
],
"max_tokens": 100
}'
(APIServer pid=389703) INFO: 127.0.0.1:37836 - "POST /v1/chat/completions HTTP/1.1" 200 OK
{"id":"chatcmpl-53d7ea4aa7214807bea4a99cde622f54","object":"chat.completion","created":1763240495,"model":"google/gemma-3-27b-it","choices":[{"index":0,"message":{"role":"assistant","content":"The image consists entirely of a repeating pattern of a symbol that appears to be a Devanagari letter. Specifically, it's the letter \"दा\" (dā). \n\nIt's a densely packed, abstract image made up solely of this script. Without further context, it's hard to say *why* this image was created. It could be a design element, a representation of a word or phrase, or a visual exploration of the letter itself.","refusal":null,"annotations":null,"audio":null,"function_call":null,"tool_calls":[],"reasoning_content":null},"logprobs":null,"finish_reason":"stop","stop_reason":106,"token_ids":null}],"service_tier":null,"system_fingerprint":null,"usage":{"prompt_tokens":275,"total_tokens":372,"completion_tokens":97,"prompt_tokens_details":null},"prompt_logprobs":null,"prompt_token_ids":null,"kv_transfer_params":null}(env) ikwak_google_com@t1v-n-bfb3e0ee-w-0:~$ (APIServer pid=389703) INFO 11-15 21:01:44 [loggers.py:181] Engine 000: Avg prompt throughput: 27.5 tokens/s
Before submitting a new issue...
- Make sure you already searched for relevant issues and checked the documentation page, which can answer lots of frequently asked questions.