Native ReAct Agent can't run on NPU channel-wise model

[JamieVC]

Describe the bug
sample
https://github.com/openvinotoolkit/openvino_notebooks/blob/latest/notebooks/llm-agent-react/llm-agent-react.ipynb

I converted a model for NPU. (Channel-wise quantization)
optimum-cli export openvino -m Qwen/Qwen2.5-3B-Instruct --weight-format int4 --sym --ratio 1.0 --group-size -1 Qwen2.5-3B-Instruct_NPU

Then, tried to run it on NPU but it failed.
Codes that I changed in notebooks/llm-agent-react/llm-agent-react.ipynb:

from pathlib import Path

#llm_model_path = llm_model_id.value.split("/")[-1]
llm_model_path = "Qwen2.5-3B-Instruct_NPU"

#llm_device = device_widget("GPU", exclude=["NPU"])
llm_device = device_widget("NPU")

Expected behavior
NPU should work well like CPU and GPU.

Screenshots

Installation instructions (Please mark the checkbox)
[yes ] I followed the installation guide at https://github.com/openvinotoolkit/openvino_notebooks#-installation-guide to install the notebooks.

[brmarkus]

The Jupyter notebook you are referring to "https://github.com/openvinotoolkit/openvino_notebooks/blob/latest/notebooks/llm-agent-react/llm-agent-react.ipynb" has NPU explicitely excluded, see cell number 4:

from notebook_utils import device_widget
llm_device = device_widget("CPU", exclude=["NPU"])
llm_device

Are you sure the model you want to use is expected to run on the NPU?

[JamieVC]

Thanks so quick response! yes, the code excludes NPU.
In the following sample that RAG starts to run llm on NPU with openvino.genai, so just wonder if the llm ReAct plan to turn in NPU with openvino.genai.
https://github.com/openvinotoolkit/openvino_notebooks/blob/latest/notebooks/llm-rag-langchain/llm-rag-langchain-genai.ipynb