def train()

in torchrecipes/core/base_train_app.py [0:0]

• 25 lines of code
• 4 McCabe index (conditional complexity)

    def train(self) -> TrainOutput:
        trainer, log_params = self._get_trainer()

        start_time = time.monotonic()
        got_exception = None
        try:
            trainer.fit(self.module, datamodule=self.datamodule)
        except Exception as ex:
            got_exception = ex

        # log trainer status to Scuba and Hive
        total_run_time = int(time.monotonic() - start_time)
        log_params["global_rank"] = trainer.global_rank
        log_params["world_size"] = trainer.world_size
        log_params["total_run_time"] = total_run_time
        if got_exception is None:
            log_params["run_status"] = JobStatus.COMPLETED.value
            log_run(**log_params)
        else:
            log_params["error_message"] = str(got_exception)
            log_params["stacktrace"] = traceback.format_stack()
            log_params["run_status"] = JobStatus.FAILED.value
            log_run(**log_params)
            raise got_exception
        best_model_path = getattr(self._checkpoint_callback, "best_model_path", None)
        return TrainOutput(
            tensorboard_log_dir=self.log_dir, best_model_path=best_model_path
        )