This repo provides docker containers to run the folding@home client.
So this is one way of getting Folding@Home done. It has all the usual benefits of Docker containerization. Assuming you are already a docker user the install should take a few minutes, most of which is building the containers. Perhaps I'll push this to dockerhub later, idk.
Be sure docker
19.03 or higher as well as the nvidia-container-toolkit
is
installed. If docker is already installed be sure to restart docker after
installing nvidia-container-toolkit
For arch this would be:
$ yay -S docker nvidia-container-toolkit
$ sudo systemctl restart docker
Ubuntu or whatever will be very similar.
Once this is done run the test-gpu.sh
script to verify your GPU's are available:
❯ ./test-gpu.sh
Thu Apr 16 11:14:52 2020
+-----------------------------------------------------------------------------+
| NVIDIA-SMI 440.82 Driver Version: 440.82 CUDA Version: 10.2 |
|-------------------------------|----------------------|----------------------+
| GPU Name Persistence-M| Bus-Id Disp.A | Volatile Uncorr. ECC |
| Fan Temp Perf Pwr:Usage/Cap| Memory-Usage | GPU-Util Compute M. |
|===============================+======================+======================|
| 0 GeForce RTX 2080 Off | 00000000:01:00.0 On | N/A |
| 0% 43C P8 4W / 260W | 952MiB / 7979MiB | 6% Default |
+-------------------------------|----------------------|----------------------+
+-----------------------------------------------------------------------------+
| Processes: GPU Memory |
| GPU PID Type Process name Usage |
|=============================================================================|
+-----------------------------------------------------------------------------+
So docker-compose does not support the --gpus flag at this time, so you have to
slum it with the run-gpu.sh
script.
$ ./run-gpu.sh -u sthysel -t 248479
If you have FAH passkey make sure its in your environment called FAH_PASSKEY