GOT-OCR

0 Start docker 启动命令可参考: README.md

1 Prepare model

 pip install modelscope
 python downloadModel.py

2 Prepare depandents(Ubuntu_environment)

pip install opencv-python
apt-get update
apt-get install -y libgl1-mesa-glx
pip install matplotlib
conda install -c conda-forge libstdcxx-ng
pip install tiktoken
pip install verovio
pip install 'accelerate>=0.26.0'

3 Convert scripts to musa

musa‑converter ‑r ./stepfun-ai/GOT-OCR2_0 ‑l ./stepfun-ai/GOT-OCR2_0/modeling_GOT.py

4 Download infer_test video

• 访问Video-text-to-text (huggingface.co) 下载测试视频cats_1.mp4
• 把下载的cats_1.mp4视频文件放入./stepfun-ai/GOT-OCR2_0文件下

5 Run run_model.py

cp run_model.py ./stepfun-ai/GOT-OCR2_0
python ./stepfun-ai/GOT-OCR2_0/run_model.py