a-marx
commented
4 months ago Looks great! A few suggestions and pointers. I assume that some of them are on the Todo list anyways.
Notes to README.me:
- use one benchmark suites ->
use one of the benchmark suites - Here I would mention that v1 refers to all tasks used in the NeurIPS paper. We could hint at the new tasks that are mentioned in the documentation.
The config workflow file is easy to understand; thats good :)
What is currently missing is the "How to add your estimator" part. But from what is there, it is clear how to evaluate the implemented estimators.
This PR standardizes running benchmarks. In particular, it changes the workflow structure and improves the documentation.