thuml / Time-Series-Library

A Library for Advanced Deep Time Series Models.
MIT License
6.98k stars 1.1k forks source link

Why test_loader use batchsize=1 #269

Closed NEUCMD closed 1 year ago

NEUCMD commented 1 year ago

if flag == 'test': shuffle_flag = False drop_last = True if args.task_name == 'anomaly_detection' or args.task_name == 'classification': batch_size = args.batch_size else: batch_size = 1 # bsz=1 for evaluation freq = args.freq I am wonder this setting

wuhaixu2016 commented 1 year ago

Hi, this design is to mimc the online serving scenario. If you have an online model, you can use this configuration. But if you only want to propose a normal forecasting model, you can just increase this number.