Preparing v0.8 #218
Preparing v0.8 #218
Conversation
fixes #194
pin sparse to 0.9.1
Change to allow grouping order to be preserved
…le parameters in numpy.
|
Can I rebase this branch to dev? Just so that we can merge through there. Once both this and the corresponding PR on batchglm are in the respective dev branches, I would then advance them to release, incremementing the dependency version of batchglm in diffxpy as well. |
|
Ah yes of course! I was wondering what was up with that - it seems like master is ahead of |
Those were small fixes it seems, I rebased them into dev. I think the rebase caused two minor conflicts, as shown in the summary of the PR. |
| jinja2 | ||
| docutils | ||
|
|
||
| sparse==0.9.1 |
There was a problem hiding this comment.
Do we need to pin sparse exactly or can we use a >= dependency?
There was a problem hiding this comment.
I could remove the pin but this issue cropped up: #212
ilan-gold
force-pushed
the
ig/batchglm_api_update
branch
from
f3c69a6 to
5d9595e
Compare
4 participants
Pairs with theislab/batchglm#146
This PR aims to upgrade
batchglm. The main outstanding question mentioned on slack is whether the 5 failing tests intest_constrained.pyandtest_pairwise_null.pyshould be deleted. It seems to me that they are trying to do the impossible by fitting more than one scale parameter but I feel somewhat uncertain here because I have had trouble with constraints before and I am not familiar with Z-tests. Other than that, feel free to ask why I did anything!