R0.13#73
Conversation
jameswind
left a comment
jameswind
left a comment
There was a problem hiding this comment.
Does this work for a lower version of cuda, say, cuda-9.2?
It mainly depends on your TensorFlow version. If you use a lower version of TensorFlow, like TensorFlow-1.12, I think CUDA-9* is fine. Check this for a closer look. |
| ### Cmake conditional compilation | ||
| We introduced the USE-CUDA-TOOLKIT parameter as a control variable for whether to compile using the CUDA environment. If you want to build DeePMD-kit with CUDA-toolkit support, then execute cmake | ||
| ```bash | ||
| cmake -DTF_GOOGLE_BIN=true -DUSE_CUDA_TOOLKIT=true -DTENSORFLOW_ROOT=$tensorflow_root \ |
There was a problem hiding this comment.
So, we don’t need to specify the CUDA’s PATH? Does it just use TensorFlow’s settings?
There was a problem hiding this comment.
Cmake has a function, named "find_package(CUDA REQUIRED)", which can detect the CUDA environment by default. If you load the CUDA module or set the CUDA environment in your system, when you set "-DUSE_CUDA_TOOLKIT=true" , CMake will set the CUDA environment by default.
There was a problem hiding this comment.
user can specify the cuda tool-kit path by CUDA_TOOLKIT_ROOT_DIR, please check
https://cmake.org/cmake/help/latest/module/FindCUDA.html
Imported from jinzhezenggroup/computational-chemistry-agent-skills. Upstream-Commit: jinzhezenggroup/computational-chemistry-agent-skills@37ed5a2 Upstream-Paths: - machine-learning-potentials/deepmd-finetune-dpa3 - machine-learning-potentials/deepmd-python-inference - machine-learning-potentials/deepmd-train-dpa3 - machine-learning-potentials/deepmd-train-se-e2-a - molecular-dynamics/lammps-deepmd
4 participants
Add multi-GPU support unit