python3.7版本使用accelerate报错

[zhangyu345293721]

210             train_epoch_runner = self.EpochRunner(train_step_runner,should_quiet)
211             train_metrics = {'epoch':epoch}

--> 212 train_metrics.update(train_epoch_runner(train_dataloader))
213
214 for name, metric in train_metrics.items():

/opt/conda3/envs/py37_lgb3.2.1/lib/python3.7/site-packages/torchkeras/kerasmodel.py in call(self, dataloader)
74
75 for step, batch in loop:
---> 76 with self.accelerator.accumulate(self.net):
77 step_losses,step_metrics = self.steprunner(batch)
78 step_log = dict(step_losses,**step_metrics)

AttributeError: 'Accelerator' object has no attribute 'accumulate'