11/22/2020 17:14:20 - INFO - __main__ -   *** Evaluate ***
11/22/2020 17:14:20 - INFO - seq2seq.utils.utils -   using task specific params for boolq: {'max_length': 4}
Traceback (most recent call last):
  File "finetune_t5_trainer.py", line 233, in <module>
    main()
  File "finetune_t5_trainer.py", line 188, in main
    result = trainer.evaluate(eval_datasets, compute_metrics_fn)
  File "/home/rabeeh/internship/seq2seq/t5_trainer.py", line 175, in evaluate
    prediction_loss_only=True if self.compute_metrics is None else None, # self.compute_metrics[eval_task]
  File "/opt/conda/envs/internship/lib/python3.7/site-packages/transformers/trainer.py", line 1417, in prediction_loop
    loss, logits, labels = self.prediction_step(model, inputs, prediction_loss_only)
  File "/home/rabeeh/internship/seq2seq/t5_trainer.py", line 249, in prediction_step
    generated_tokens = model.generate(
  File "/opt/conda/envs/internship/lib/python3.7/site-packages/torch/nn/modules/module.py", line 779, in __getattr__
    type(self).__name__, name))
torch.nn.modules.module.ModuleAttributeError: 'DataParallel' object has no attribute 'generate'