search

TL-UESTC / Domain-Adaptive-Remaining-Useful-Life-Prediction-with-Transformer

Pytorch implementation for Domain Adaptive Remaining Useful Life Prediction with Transformer
47 stars 7 forks source link

CMAPSS result not producing what is claimed in paper #7

Open ZeeshanAbbas opened 7 months ago

ZeeshanAbbas commented 7 months ago

Hi, I was trying to reproduce the result, but it's not producing the result as claimed in the repo/article, not even for the pre-trained models, "best-performing models saved in folder save/final," as I run "python validation_cmapss.py --source FD001 --target FD003" it results min RMSE = 26.0132, min score = 55565 and in article you claimed RMSE = 15.5 and score=2974, similarly for others. I can understand there may be hyper-tuning and model convergence issues that can arise during training, but I'm a little bit worried about why the "best-performing models saved in folder save/final" don't produce the result. Can you please tell me what I'm doing wrong? I have attached the screen shoot of the above model evaluation result. Weixin Screenshot_20240201161231

956077450 commented 7 months ago

Please make sure you are using the exact version of PyTorch and torchvision required. Check #1.

ZeeshanAbbas commented 6 months ago

Thanks for your reply.

ZeeshanAbbas commented 6 months ago

I have other questions/ confusion; please clear my confusion. 1) Why didn't you explain the preprocessing step in the paper and code? 2) You haven't tested your model on the given test set, and it's RUL; instead, you have test on the part of the preprocessed dataset; why is it so?

image