Table of Contents
- General
- auto exists with error code 137
- Dockerfile parse error
- Unknown Flag --profile
- Output is a always green image
- Found no NVIDIA driver on your system even though the drivers are installed and nvidia-smi shows it
- Linux
- Windows / WSL
- Build fails at The Shell command, /bin/bash not found in WSL.
- Build fails with credentials errors when logged in via SSH on WSL2/Windows
- unable to access 'https://github.com/...': Could not resolve host: github.com or any domain
- Other build errors on windows
- AWS
General
Unfortunately, AMD GPUs #63 and MacOs #35 are not supported, contributions to add support are very welcome.
auto exists with error code 137
This is an indicator that the container does not have enough RAM, you need at least 12GB, recommended 16GB.
You might need to adjust the size of the docker virtual machine RAM depending on your OS.
Dockerfile parse error
Error response from daemon: dockerfile parse error line 33: unknown instruction: GIT
ERROR: Service 'model' failed to build : Build failed
Update docker to the latest version, and make sure you are using docker compose instead of docker-compose. #16, also, try setting the environment variable DOCKER_BUILDKIT=1
Unknown Flag --profile
Update docker to the latest version, and see this comment, try setting the environment variable mentioned in the previous point.
Output is a always green image
use --precision full --no-half. #9
Found no NVIDIA driver on your system even though the drivers are installed and nvidia-smi shows it
add NVIDIA_DRIVER_CAPABILITIES=compute,utility and NVIDIA_VISIBLE_DEVICES=all to container can resolve this problem #348
Linux
Error response from daemon: could not select device driver "nvidia" with capabilities: [[gpu]]
Install NVIDIA Container Toolkit and restart the docker service #81
docker compose --profile auto up --build fails with OSError
This might be related to the overlay2 storage driver used by docker overlayed on zfs, change to the zfs storage driver #433
Windows / WSL
Build fails at The Shell command, /bin/bash not found in WSL.
Edit the corresponding docker file, and change the SHELL from /bin/bash to //bin/bash #21, note: this is a hack and something in your wsl is messed up.
Build fails with credentials errors when logged in via SSH on WSL2/Windows
You can try forcing plain text auth creds storage by removing line with "credStore" from ~/.docker/config.json (in WSL). #56
unable to access 'https://github.com/...': Could not resolve host: github.com or any domain
Set the build/network of the service you are starting to host #114
Other build errors on windows
- Make sure:
- Windows 10 release >= 2021H2 (required for WSL to see the GPU)
- WSL2 (check with
wsl -l -v) - Latest Docker Desktop
- You might need to create a
.wslconfigand increase memory, if you have 16GB RAM, set the limit to something around 12GB, #34 #64 - You might also need to force wsl to allow file permissions
AWS
You have to use one of AWS's GPU-enabled VMs and their Deep Learning OS images. These have the right divers, the toolkit and all the rest already installed and optimized. #70