Demo VQA app running on Intel dGPU.
Make sure to install GPU driver and OneAPI according to this document.
IPEX-LLM needs python3.11
.
sudo apt install python3.11 python3.11-venv
Create virtual environment.
python -m venv .env
Install IPEX-LLM and dependencies.
source .env/bin/activate
pip install -U pip
pip install --pre --upgrade ipex-llm[xpu] --extra-index-url https://pytorch-extension.intel.com/release-whl/stable/xpu/cn/
pip install -r requirements.txt
Download Qwen-VL pretrained model.
huggingface-cli download Qwen/Qwen-VL-Chat
Run gradio app.
source /opt/intel/oneapi/setvars.sh
GRADIO_SERVER_NAME=0.0.0.0 python qwen_vl_ipex.py --server-name 0.0.0.0