orion
orion copied to clipboard
Epistimio
Asynchronous Distributed Hyperparameter Optimization.
The description of the parallelization is not faithful anymore in documentation: https://orion.readthedocs.io/en/stable/user/parallel.html. The synchronisation is now done across Runners, not workers, and the algorithm is shared across the Runners using...
Ex: https://orion.readthedocs.io/en/stable/user/algorithms.html The table of content should have the form of a bullet list, with sub sections being more indented. It currently renders as a sequence of lines, indented, with...
The documentation states that the API should be included in v0.2.0, but it is not. We should update this and plan properly in which release this API would be included....
Following #602, broken trials we never be set a lie objective (although it does not make much difference as HB, ASHA & EES where meant to be used with `StubParallelStrategy(stub_value=None)`)....
Could be something like https://scikit-learn.org/dev/about.html#funding Or the section 'Organizations using and contributing to MLflow' here: https://mlflow.org/#community
This is very unclear, and terrifying for a new user! ```python Traceback (most recent call last): File “/media/martin/the_milano/anaconda3/bin/orion”, line 11, in sys.exit(main()) File “/media/martin/the_milano/anaconda3/lib/python3.7/site-packages/orion/core/cli/__init__.py”, line 39, in main orion_parser.execute(argv) File...
When a user runs Orion, it should print out something like: ``` you are using the Algorithm {name}, please include the following citation in your publication {algo.citation_string} ``` Researchers building...
When a trial is broken, it is assigned an objective using the lying strategy the same way it is done for pending trials. This is very bad. For sequential algorithms,...
# Description Detects if a by-design crashing black-box script Traceback is visible in the stderr output of orion. The test succeeds, which indicates that the stacktrace is indeed visible by...
We are running Orion 0.2.1 on a HTCondor cluster using Dask with a single MongoDB instance hosted on a virtual machine. We run multiple experiments sequentially using the same client...
