[Bug]: Error running Inference on integrated GPU

[nilutpolkashyap]

OpenVINO Version

2024.0

Operating System

Windows System

Device used for inference

GPU

Framework

None

Model used

GroundedSAM

Issue description

I am trying to run the grounded-segment-anything.ipynb notebook.

I can successfully convert the GroundingDINO to OpenVINO IR format. I can successfully run inference on my Intel CPU and get inference results correctly.

But when I try to run inference with my Intel integrated GPU, I get error as

RuntimeError: Exception from src\inference\src\cpp\core.cpp:109:
Exception from src\inference\src\dev\plugin.cpp:54:
Check 'input_pshape.is_static()' failed at src\plugins\intel_gpu\src\plugin\ops\roll.cpp:24:
Dynamic shapes are not supported for Roll operation yet

System Specifications -
CPU - Intel(R) Core(TM) i7-9750H CPU
GPU - integrated Intel(R) UHD Graphics 630
RAM - 16 GB
OS - Windows 10

Step-by-step reproduction

I am following the steps from the openvino notebook for Object detection and masking from prompts with GroundedSAM (GroundingDINO + SAM) and OpenVINO from grounded-segment-anything.ipynb

CPU inference runs perfectly.But GPU inference fails

Relevant log output

RuntimeError                              Traceback (most recent call last)
Cell In[14], line 1
----> 1 ov_compiled_grounded_dino = core.compile_model(ov_dino_model, device.value)

File F:\github\openvino_projects\grand_theft_autonomous\.openvino-venv\lib\site-packages\openvino\runtime\ie_api.py:515, in Core.compile_model(self, model, device_name, config)
    509 if device_name is None:
    510     return CompiledModel(
    511         super().compile_model(model, {} if config is None else config),
    512     )
    514 return CompiledModel(
--> 515     super().compile_model(model, device_name, {} if config is None else config),
    516 )

RuntimeError: Exception from src\inference\src\cpp\core.cpp:109:
Exception from src\inference\src\dev\plugin.cpp:54:
Check 'input_pshape.is_static()' failed at src\plugins\intel_gpu\src\plugin\ops\roll.cpp:24:
Dynamic shapes are not supported for Roll operation yet

Issue submission checklist

• I'm reporting an issue. It's not a question.
• I checked the problem with the documentation, FAQ, open issues, Stack Overflow, etc., and have not found a solution.
• There is reproducer code and related data files such as images, videos, models, etc.

[avitial]

@nilutpolkashyap thanks for reaching out. I see that you are using OpenVINO 2024.0, would it be possible to try on the latest 2024.1 release or latest nightly build? PR #23617 may have addressed this issue already.

[nilutpolkashyap]

Hi @avitial ,
I tried the latest 2024.1 release, and it just kills the kernel when I try to run inference on GPU.

With the nightly release, the inference on GPU goes on forever, it's been almost 1.5 hours since I started inference and its utilizing 100% of my integrated GPU. Is this due to hardware limitations?