huggingfaceR icon indicating copy to clipboard operation
huggingfaceR copied to clipboard

hf_load_dataset issues

Open jpcompartir opened this issue 2 years ago • 0 comments

Function + arguments break too easily for datasets other than emotions e.g.

x <- hf_load_dataset("go_emotions", as_tibble = TRUE, label_name = "int2str")

Doesn't deal with go_emotions' labels being lists. Do we need to take care of this? User can in this example with:

x <- x %>%
  tidyr::unnest(labels)

But this same feature also prevents label_name working as intended which seems a bigger problem. Currently user would have to go and fetch the int2str lookup table.

Playing around with IMDB and there are similar related issues.

jpcompartir avatar Nov 12 '22 14:11 jpcompartir