adcc
adcc copied to clipboard
Conda-forge deployment
I'd like to test deployment of libtensorlight/adcc via conda-forge. Comments etc. can be added to this issue 😄
- [x] deploy libtensorlight
- [x] deploy adcc
- [x] adapt documentation and badges
- [x] nuke conda deployment from repo
libtensorlight already up on c-f ✅
Very nice. Let me know if you need to discuss something. In case there is need: I know some people from the conda-forge team I could summon.
I had some problems yesterday but everything is running smoothly now I guess: https://github.com/conda-forge/staged-recipes/pull/17477
BTW I can add you as package maintainer if you want 😬
I don't have strong feelings about it. As you wish ...
Why did you disable the sha256sum check by the way. I'm not sure this is a good idea.
The sha256sum (older version) on c-f VM does not support the --ignore-missing option. Maybe you have a suggestion how one could patch the script instead? 😬
I guess what you could do is "preprocess" the SHA256SUM file to only contain the files which are truely there, e.g. something like
while read line; do
[ -f "$line" ] && echo "$line" >> SHA256SUM.filtered
done < SHA256SUM
mv SHA256SUM.filtered SHA256SUM
I guess it even makes sense to update this in adcc directly.
I'm keeping track of the next steps here (in order):
- [x] merge #141
- [x] remove own conda pipeline & add conda-based build to CI (such that we don't break anything conda-related in
setup.py) #143 - [x] release
v0.15.10 - [x] adapt recipe on c-f for
v0.15.10(remove patch & dependencies) (@maxscheurer) - [x] Smoke tests #144
- [x] release
v0.15.11 - [x] adapt recipe on c-f for
v0.15.11(@maxscheurer) - [x] manually deploy
v0.15.11toadccchannel (@maxscheurer) - [x] adapt documentation and badges once
adccis on c-f - [x] upload new docs (currently online is
v0.15.12(@mfherbst) - [x] add Psi4 to conda CI to increase test coverage (#137)
- [ ] retire
adccchannel for0.16.0.
upload new docs (currently online is v0.15.7)
Seriously ... that's my bad.
We should probably somehow discontinue the adcc channel on anaconda.org. I don't know if we should leave it up there for some "grace period"? Opinions, @mfherbst and @loriab (if you need the current adcc channel versions)?
Agree, we should leave it up for some time. Perhaps we could push a last version by a manual build (?), which includes a warning upon import?
That's a good suggestion! Can do the manual build for 0.15.11 and upload it to the adcc channel. I would say we remove our own channel once 0.16.0 comes out.
Sounds fair. That'll probably be the case when another major feature (LR-PCM ?) goes in.
Yes, PCM (#132) and gradients (#124) should be in 0.16.0.
@mfherbst if you find the time, could you push the v0.15.13 docs to the website?
Done.