Add build matrix for docker images, update images#1936
Add build matrix for docker images, update images#1936MMathisLab merged 16 commits intoDeepLabCut:mainfrom
Conversation
docker/build.sh
Outdated
| 2.2.1.1 | ||
| do | ||
| for stage in base core test gui jupyter; do | ||
| tag=${deeplabcut_version}-${stage}-cuda${cuda_version} |
|
@jeylau can we fix this? // get it merged DLC broader docker support is quite important for a lot of applications, including in our lab ;) |
|
"core" now works; I will test the GUI tomorrow on Ubuntu. |
|
@MMathisLab, can you please, merge this branch?
Once it merged, I will PR to:
|
|
@MMathisLab Here is my PR to @stes PR! Hope this time it will make the affair! 🙌 |
Co-authored-by: Steffen Schneider <steffen@bethgelab.org>
Co-authored-by: Steffen Schneider <steffen@bethgelab.org>
|
All right, some progress here! Thanks a lot to @maryapp for taking over and adding support for the latest deeplabcut version / upgrading the images. The changelog can be seen in my forked repo here: stes#3 If all tests pass (for docker --- I assume @maryapp checked these), this is good to go for merging into main from my end. |
|
An open Q is maybe the faith of the deeplabcut-docker python package. It seems GUI support will be dropped with the latest version of this PR --- so I think we should decide whether to remove that, or use the legacy images in the pypi package. Opinions? |
Yes, all passed.
I would vote for the 1st option. I will check on the python package code (also how it picks up the right cuda version support, etc.) |
We need to keep the GUI support in legacy versions. Please then decide if this can be merged ... |
|
@stes can you, please, elaborate what is "legacy" images for you? You are referencing the very first old images that are distributed via docker pip package now because of the tag? |
|
I am merging as this is becoming unmanageable -- we need docker support... |
Open:
Build and test logs
https://gist.github.com/stes/80412bac41a37cabb6248db5e5d2884e
Testing
If you come across this PR and want to test it, please see this discussion for instructions (until this package is updated):
deeplabcut-dockerThis should open the deeplabcut UI. If you are on a linux machine, you might run
xhost +before executing the last command if you get a permission error.Fixes #2164