Closed fracek closed 2 years ago
Looks like the issue now is that the GBT notebook uses gurobi. We should stick to cbc for all notebooks so that everyone can try them out and we can test them. What do you think?
Agreed. @ThebTron, would you please switch to CBC in the black-box optimization notebook? Maybe try making the trees smaller in LightGBM if there is computational trouble with CBC.
Merging #30 (4c8901a) into main (139eb3d) will not change coverage. The diff coverage is
n/a
.
@@ Coverage Diff @@
## main #30 +/- ##
=======================================
Coverage 88.79% 88.79%
=======================================
Files 23 23
Lines 1071 1071
Branches 160 160
=======================================
Hits 951 951
Misses 98 98
Partials 22 22
Continue to review full report at Codecov.
Legend - Click here to learn more
Δ = absolute <relative> (impact)
,ø = not affected
,? = missing data
Powered by Codecov. Last update 139eb3d...4c8901a. Read the comment docs.
The issue was that the second test run (the one testing notebooks) was overwriting the coverage from the main tests. I changed the notebook run to merge coverage with the main tests.
In this pull request, import_network.ipynb still mixes ReLU and sigmoid activation functions. @fracek, possible to just make sure that the activation functions are ones we typically allow together?
I changed it to not use sigmoid anymore.
Like we mentioned, this PR add testing of notebooks.