JunHwi/kmhas_multilabel

Pretrained K-mHas with multi-label model with "koelectra-v3"

You can use tokenizer of this model with "monologg/koelectra-v3-base-discriminator"

dataset : https://huggingface.co/datasets/jeanlee/kmhaskoreanhate_speech

pretrained_model : https://huggingface.co/monologg/koelectra-base-v3-discriminator

label maps are like this.

>>

{'origin': 0, 'physical': 1, 'politics': 2, 'profanity': 3, 'age': 4, 'gender': 5, 'race': 6, 'religion': 7, 'nothatespeech': 8}

You can use label map with below code.

from huggingfacehub import hfhub_download

repoid = "JunHwi/kmhasmultilabel"

filename = "kmhasdict.pickle" # 위 repoid에 업로드한 파일 이름

labeldict = hfhubdownload(repoid, filename)

with open(label_dict, "rb") as f: label2num = pickle.load(f)