rm lighteval sha from id

added precision

fix rounding

get model type info from request file

only display the scores for the latest result file

Added icons for types + fixed pending queue

fix new config name

FT: precision and adapter models

Small fix - we do not want to display models where the MMLU is old with models where the MMLU is new - however, since version is displayed in the results, we keep the files

Using the new backend

merge refactor