seb.

https://h2oai.atlassian.net/browse/PUBDEV-8315 Restoring the full model rendering for default output or `print(model)` in iPython/Jupyter or Py REPL

[nocheck] consistent handling of scoring support logic in CV

1

@wendycwong @michalkurka : does this make sense?

https://h2oai.atlassian.net/browse/PUBDEV-7799

PUBDEV-8416, PUBDEV-8417: AutoML propagate early stopping criteria to SEs + GLM

1

https://h2oai.atlassian.net/browse/PUBDEV-8416 https://h2oai.atlassian.net/browse/PUBDEV-8417

https://h2oai.atlassian.net/browse/PUBDEV-8472 building 2 GLMs: - a very basic one, trained in first priority group. - one with lambda & alpha search (former default) that often takes much longer, and we...

[nocheck] Adding Py support until 3.10

5

Refactored groovy scripts + docker scripts + makefile in such a way that we can more clearly see which versions are really used and for which purpose. Py 3.10 is...

https://h2oai.atlassian.net/browse/PUBDEV-8302 Even when the results represent indices (e.g. result of `h2o.column_by_types`, those are represented as floats by the Py client (it's working as expected in R). This fix allow ints...

In some cases, when upgrading some frameworks locally using the `python runbenchmark my_framework --setup force` command, there can be dependency version conflicts during the setup as the previous virtual env...

Adding file lock on global results file

1

addressing https://github.com/openml/automlbenchmark/issues/135 applied on top of improvements regarding saving logic in https://github.com/openml/automlbenchmark/pull/452

Save results on each job completion

1

Results are currently saved when all jobs are completed, which is a bit annoying for so many reasons (what if process hangs or get killed before completion? can't easily access...