Thomas Moreau
Thomas Moreau
There are several random failures in appveyor. Let's try to report all of them here: - [timeout in teardown in `test_in_callback_submit_with_crash`](https://ci.appveyor.com/project/tomMoral/loky/build/1.0.1350/job/vnjibh15dh3l9udm). In this one, it hangs while joining the `QueueManagerThread`,...
Extend the existing doc: - [ ] Add examples of the fixed deadlocks - [ ] after worker crash - [x] after result unpickle errors - [ ] Add example...
For now, we only embed the javascript in result pages. It would be nice to do the same for all pages (`index` and the benchmark pages) so we avoid conflict....
We have a very aggressive way to increase the `stop_val`, which can lead to very long run, blowing up the timeout for no good reason. The way we get the...
The use of `as_completed` and `rich.progress.track` might lead to timeout on the frontal of SLURM cluster. This should be properly investigated, to make sure we are not overloading by using...
For users unused to SLURM, we could ease the use of `benchopt` by better introducing what will happen and pointing out that in most cases, `benchopt` should be run in...
We should properly test the `--slurm` option. We could use the [cluster arg in AutoExecutor](https://github.com/facebookincubator/submitit/blob/main/submitit/auto/auto.py#L48) in debug or local mode for this.
The tests of benchopt with no conda are very long for not so good reasons. (~11min for linux and >17min for OSX) The goal of the tests is simply to...
We should have a standard way to share a benchmark outside github, for instance in paper sup mat. We could create `benchopt share .` to make a zip file with...
The coverage of the project is not superb yet. We should aim at improving it from this page: https://app.codecov.io/gh/benchopt/benchopt/ Possible way to do it: - [ ] stop `xfailing` julia,...