You are attempting to use Flash Attention 2.0 with a model not initialized on GPU.

[lesca]

安装：使用README中指定的安装方式

conda create -n deepseek-ocr python=3.12.9 -y
conda activate deepseek-ocr
pip install -r requirements.txt
# Install Torch First
pip install flash-attn==2.7.3 --no-build-isolation

执行报错：

python start_ocr_webui.py                                                                                                        (deepseek-ocr) 
Using default model path: deepseek-ai/DeepSeek-OCR
To use a custom model path, set the environment variable: DEEPSEEK_OCR_MODEL_PATH
* Running on local URL:  http://0.0.0.0:7860
* To create a public link, set `share=True` in `launch()`.
=== Starting image processing ===
Number of images: 1
Prompt: <|grounding|>Convert the document to markdown.
Loading model from: deepseek-ai/DeepSeek-OCR
You are using a model of type deepseek_vl_v2 to instantiate a model of type DeepseekOCR. This is not supported for all configurations of models and can yield errors.
You are attempting to use Flash Attention 2.0 without specifying a torch dtype. This might lead to unexpected behaviour
You are attempting to use Flash Attention 2.0 with a model not initialized on GPU. Make sure to move the model to GPU after initializing it on CPU with `model.to('cuda')`.
fish: Job 1, 'python start_ocr_webui.py' terminated by signal SIGKILL (Forced quit)

请问这个怎么处理？