Make tools to simplify adding traces of agents to an ever growing huggingfaces dataset. * create 2 datasets on hugging face * one that would be an index to be able to easily retrieve traces based on attributes similar to the dataframe when we run `load_result_df` * one that contains actual zipped traces that can be retrieved from a pointer in the index * make code to upload a study trace by trace and easy way to group the traces by study in the index. * legality: * limit adding only from the domains that are whitlisted (e.g. our benchmarks or a subset of them) * based on which LLM and which benchmarks attribute a specific license to it.
Make tools to simplify adding traces of agents to an ever growing huggingfaces dataset.
create 2 datasets on hugging face
load_result_dfmake code to upload a study trace by trace and easy way to group the traces by study in the index.
legality: