您好，运行train_t5_copy，报prepare_inputs_for_generation()参数错误

[zyDotwei]

File "/home/root/v2/train_t5_copy.py", line 84, in validation_step pred = self.predict_batch(batch) File "/home/root/v2/train_t5_copy.py", line 60, in predict_batch pred = self.model.generate(eos_token_id=tokenizer.sep_token_id, File "/opt/conda/lib/python3.8/site-packages/torch/autograd/grad_mode.py", line 26, in decorate_context return func(*args, **kwargs) File "/opt/conda/lib/python3.8/site-packages/transformers/generation_utils.py", line 965, in generate return self.beam_search( File "/opt/conda/lib/python3.8/site-packages/transformers/generation_utils.py", line 1604, in beam_search model_inputs = self.prepare_inputs_for_generation(input_ids, **model_kwargs) File "/home/root/v2/t5_copy.py", line 64, in prepare_inputs_for_generation res = super().prepare_inputs_for_generation(input_ids, TypeError: prepare_inputs_for_generation() takes from 2 to 6 positional arguments but 9 were given

[vanpersie32]

I met the same problem

[renmada]

如果transformers的版本高于4.12.0的话，降级试试

[zyDotwei]

如果transformers的版本高于4.12.0的话，降级试试

可能是我的太低了；改到4.10.0就好了，感谢～