[Bug] Incompatible between bmf 0.1.0 and official runtime docker image

[PYNing]

Environment：

GPU 4090
Driver 535.54.03

Description & Reproduce：

（1）Run docker
docker run --gpus all -e NVIDIA_DRIVER_CAPABILITIES=all -it babitmf/bmf_runtime:latest bash
（2）It‘s fine to compile and run 0.0.11(inside the latest official runtime images)

cd ~/bmf
export CMAKE_ARGS="-DBMF_ENABLE_CUDA=ON"
./build.sh

(3) However, when I replaced the existing code with version 0.1.0, it compiled successfully, but an error occurred during runtime.

python3
>>> import bmf

Here is the error:

Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
  File "/data/ningpeiyang/2025/0427_build_bmf/bmf/output/bmf/__init__.py", line 4, in <module>
    from bmf.python_sdk.module_functor import make_sync_func
  File "/data/ningpeiyang/2025/0427_build_bmf/bmf/output/bmf/python_sdk/__init__.py", line 1, in <module>
    from .module_functor import make_sync_func, ProcessDone
  File "/data/ningpeiyang/2025/0427_build_bmf/bmf/output/bmf/python_sdk/module_functor.py", line 1, in <module>
    import bmf.lib._hmp
ImportError: /data/ningpeiyang/2025/0427_build_bmf/bmf/output/bmf/lib/_hmp.cpython-38-x86_64-linux-gnu.so: undefined symbol: _ZN3hmp4cuda5Event5blockERKSt8optionalINS_6StreamEE

Related

#188

[taoboyang]

It seems that the hmp linked here does not contain GPU-related implementations. Is the library link wrong? The default bmf in the image is the CPU version. Have you reset LD_LIBRARY_PATH and PYTHONPATH after recompiling? The link behavior will be set to the library in /root/bmf by default in image.