google-deepmind / google-deepmind/alphafold
error running with a nvidia rtx pro 4500/5000
Open
- Dominant language
- Python
- Stars
- 14.9k
- Forks
- 2.9k
- PR merge metrics
- No merged PRs in 30d
Description
Hi, everyone
When I was using the Nvidia RTX PRO 4500/5000, I modified the Dockerfile, and the relevant information is as follows:
1. using the nvidia/cuda:12.8.0-cudnn-devel-ubuntu22.04 image
2. using jax[cuda12]==0.6.0 and 0.6.1 and 0.8
3. upgrading to dm-haiku==0.0.14
4. installing tensorflow==2.12.0 and latest
I can successfully compile it, and AlphaFold runs, but only using the CPU; it cannot utilize the GPU (nvidia-smi shows no processes). However, if I enter the AlphaFold container image, JAX can detect the GPU.
any suggestions?
Contributor guide
Assessment
This issue has not been assessed yet.