Closed yingtaoren closed 1 year ago
This is my implicit and python version implicit Version: 0.7.2 python Version:3.10 CUDA Version: 12.2
We currently are shipping cuda wheels on pypi - but only for cuda 11.* on linux x86_64. The only cuda12 binaries we're shipping right now is on anaconda.
If you don't want to use conda, you can either try building from source (cloning the repo and then running something like python -m pip wheel .
to build the cuda 12 wheel yourself) or you could run implicit in a 11.8 docker container (by going something like :
docker run -it --gpus=all nvidia/cuda:11.8.0-devel-ubuntu22.04 bash
apt update
apt install python3 python3-pip
pip3 install implicit
)
Thank you so much for your help! Your solution indeed resolved my issue.
The command "pip install implicit" does not install the necessary GPU extensions, resulting in an error: (implicit) production-cuda-server:/data$ python implicit_test.py /home/xingyuan/anaconda3/envs/implicit/lib/python3.10/site-packages/implicit/cpu/als.py:95: RuntimeWarning: OpenBLAS is configured to use 6 threads. It is highly recommended to disable its internal threadpool by setting the environment variable 'OPENBLAS_NUM_THREADS=1' or by calling 'threadpoolctl.threadpool_limits(1, "blas")'. Having OpenBLAS use a threadpool can lead to severe performance issues here. check_blas_config() <class 'implicit.cpu.als.AlternatingLeastSquares'> Traceback (most recent call last): File "/data/implicit_test.py", line 21, in
model = model.to_gpu()
File "/home/anaconda3/envs/implicit/lib/python3.10/site-packages/implicit/cpu/als.py", line 442, in to_gpu
ret = implicit.gpu.als.AlternatingLeastSquares(
File "/home/anaconda3/envs/implicit/lib/python3.10/site-packages/implicit/gpu/als.py", line 59, in init
raise ValueError("No CUDA extension has been built, can't train on GPU.")
ValueError: No CUDA extension has been built, can't train on GPU.
To enable GPU inference and training, it is required to use "conda install -c conda-forge implicit implicit-proc=*=gpu". How can I resolve this issue? I need to use pip installing the package