Skip to content

Latest commit

 

History

History
32 lines (22 loc) · 945 Bytes

README.md

File metadata and controls

32 lines (22 loc) · 945 Bytes

iaml

Reproduce results from YAHPO Gym iaml.

Installation

You can install the released version from GitHub using:

remotes::install_github("sumny/iaml")

Note that for exact reproduction, you need the R packages with the exact version as specified in the DESCRIPTION (i.e., == instead of >=) and also R version 4.0.5. Also note that some measures (ram and time related ones) can never be exactly reproduced because they depend on system hardware and system load.

By default, in all eval_* functions, the original fixed seed is used and therefore, runs are fully deterministic, (except for the ram and time measures mentioned above).

To change this, simply provide a custom seed.

Note that data is obtained via mlr3oml and is cached via qs.

Example

library(iaml)
scenario = "iaml_glmnet"
configuration = list(alpha = 0.1, s = 0.1, trainsize = 0.05, task_id = "40981")
eval_yahpo(scenario, configuration)