Closed ck37 closed 2 years ago
Hey @ck37
You can always use a custom metric as explained in this guide from HF.
If this issue needs to be contributed to (for enhancing the metric API) I think this link would be helpful for the MAE
metric.
You can use a local metric script just by providing its path instead of the usual shortcut name
Is your feature request related to a problem? Please describe.
I am uploading our dataset and models for the "Constructing interval measures" method we've developed, which uses item response theory to convert multiple discrete labels into a continuous spectrum for hate speech. Once we have this outcome our NLP models conduct regression rather than classification, so binary metrics are not relevant. The only continuous metrics available at https://huggingface.co/metrics are pearson & spearman correlation, which don't ensure that the prediction is on the same scale as the outcome.
Describe the solution you'd like I would like to be able to tag our models on the Hub with the following metrics:
Describe alternatives you've considered
I don't know if there are any alternatives.
Additional context Our preprint is available here: https://arxiv.org/abs/2009.10277 . We are making it available for use in Jigsaw's Toxic Severity Rating Kaggle competition: https://www.kaggle.com/c/jigsaw-toxic-severity-rating/overview . I have our first model uploaded to the Hub at https://huggingface.co/ucberkeley-dlab/hate-measure-roberta-large
Thanks, Chris