使用transformers过程中出现的bug

1. The following model_kwargs are not used by the model: ['encoder_hidden_states', 'encoder_attention_mask'] (note: typos in the generate arguments will also show up in this list)

使用text_decoder就出现上述错误，这是由于transformers版本不兼容导致的

from transformers import AutoModel, AutoConfig, BertGenerationDecoder
decoder_config = AutoConfig.from_pretrained(args['text_checkpoint'])

text_decoder = BertGenerationDecoder(config=decoder_config)

output = self.text_decoder.generate(input_ids=cls_input_ids, 
                                    encoder_hidden_states=encoder_hidden_states, 
                                    encoder_attention_mask=encoder_attention_mask,  
                                    max_length=self.args['max_seq_length'], 
                                    do_sample=True, 
                                    num_beams=self.args['beam_size'], 
                                    length_penalty=1.0, use_cache=True,
                                    )

解决办法：将transformer的版本换到以下范围, 4.15.0<=transformers<4.22.0，transformers>=4.25.0

比如：pip install transformers==4.25.1 or pip install transformers==4.20.1