search

ZhaoZhibin / UDTL

Source codes for the paper "Applications of Unsupervised Deep Transfer Learning to Intelligent Fault Diagnosis: A Survey and Comparative Study" published in TIM
MIT License
400 stars 103 forks source link

last_batch #9

Open wb1997-cell opened 2 years ago

wb1997-cell commented 2 years ago

When using MMD, the 50th epoch will be wrong, because the dimension of Batch is not matched, your parser

parser.add_argument('--last_batch', type=bool, default=False, help='whether using the last batch') drop_last=(True if args.lastbatch and x.split('')[1] == 'train' else False)) The default should be True

ZhaoZhibin commented 2 years ago

Yes, you are right. When using distance-based methods, we should drop the final batch. But when using other methods, we do not need to drop the final batch

wb1997-cell commented 2 years ago

thanks