我用2.0版本进行命名实体识别模型训练,想请教下不同的transformer有什么区别?代码如下,官方提供的transformer可选项包括:[‘uncased_L-12_H-768_A-12’, ‘uncased_L-24_H-1024_A-16’, ‘cased_L-12_H-768_A-12’, ‘cased_L-24_H-1024_A-16’, ‘multi_cased_L-12_H-768_A-12’, ‘multilingual_L-12_H-768_A-12’, ‘chinese_L-12_H-768_A-12’, ‘wwm_uncased_L-24_H-1024_A-16’, ‘wwm_cased_L-24_H-1024_A-16’, ‘albert_base_zh’, ‘albert_large_zh’, ‘albert_xlarge_zh’, ‘albert_xxlarge_zh’, ‘albert_base’, ‘albert_large’, ‘albert_xlarge’, ‘albert_xxlarge’]
recognizer = TransformerNamedEntityRecognizer()
recognizer.fit(‘C:/Users/jane/Desktop/train_cusdic_2018.tsv’, ‘C:/Users/jane/Desktop/valid_cusdic_2018.tsv’, save_dir, transformer=’********’, metrics=‘f1’)