MetaCollector

MetaCollector is the key component for collect data for evaluation in meta learning circumstance.

Overall, we extend Collector to MetaCollector. The extended modification can be listed briefly as following:

• [Override] self.eval_collect(self, eval_pred: torch.Tensor, data_label: torch.Tensor): Collect data for evaluation.

• [Override] self.data_collect(self, train_data): Collect the evaluation resource from training data.