clue/roberta_chinese_base

roberta_chinese_base

Overview

Language model: roberta-base Model size: 392M Language: Chinese Training data: CLUECorpusSmall Eval data: CLUE dataset

Results

For results on downstream tasks like text classification, please refer to this repository.

Usage

NOTE: You have to call BertTokenizer instead of RobertaTokenizer !!!

import torch
from transformers import BertTokenizer, BertModel
tokenizer = BertTokenizer.from_pretrained("clue/roberta_chinese_base")
roberta = BertModel.from_pretrained("clue/roberta_chinese_base")

About CLUE benchmark

Organization of Language Understanding Evaluation benchmark for Chinese: tasks & datasets, baselines, pre-trained Chinese models, corpus and leaderboard.

Github: https://github.com/CLUEbenchmark Website: https://www.cluebenchmarks.com/