bug: lmcache/vllm-openai for RTX 5090

Describe the bug

What version of lmcache/vllm-openai should I take to run on RTX 5090?

I'm trying to run production-stack in k3s and get the error:

NVIDIA GeForce RTX 5090 with CUDA capability sm_120 is not compatible with the current PyTorch installation.
The current PyTorch install supports CUDA capabilities sm_50 sm_60 sm_70 sm_75 sm_80 sm_86 sm_90.
If you want to use the NVIDIA GeForce RTX 5090 GPU with PyTorch, please check the instructions at https://pytorch.org/get-started/locally/

I have already tried many versions of the image, but none of them worked on the RTX 5090.

To Reproduce

Driver Version: 570.133.07
CUDA Version: 12.8

$ nvcc --version
nvcc: NVIDIA (R) Cuda compiler driver
Copyright (c) 2005-2025 NVIDIA Corporation
Built on Fri_Feb_21_20:23:50_PST_2025
Cuda compilation tools, release 12.8, V12.8.93
Build cuda_12.8.r12.8/compiler.35583870_0

$ python3 -c "
import torch
print(f'PyTorch version: {torch.version}')
print(f'CUDA available: {torch.cuda.is_available()}')
print(f'CUDA version in PyTorch: {torch.version.cuda}')
print(f'cuDNN version: {torch.backends.cudnn.version()}')
if torch.cuda.is_available():
print(f'Current GPU: {torch.cuda.get_device_name(torch.cuda.current_device())}')
"
PyTorch version: 2.7.1+cu128
CUDA available: False
CUDA version in PyTorch: 12.8
cuDNN version: 90701

Expected behavior

No response

Additional context

No response

Start by checking the lmcache/vllm-openai image versions used by production-stack and compare their PyTorch CUDA support with the reported RTX 5090, driver, and CUDA 12.8 details. Reproduce the deployment in k3s and verify whether a compatible image exists; done means identifying and documenting a working version or the required compatibility change.