How does the PEMS-BAY datasets preprocess and train?

[0shelter0]

Hello, thanks for sharing your work about STTN! Recently, I run your original code using default configs but I get the results much worse than yours. I doubt that it is caused by my runtime envs different with yours. My runtime consists of tensorflow1.15, cuda11.5, cudnn8.3.3, python3.8 and other requirements as README tells, and my experiments are conducted on a single RTX3090 and ubuntu20.04 system. In a word, it's hard for me to reproduce your reported results.

Another ask for you is how the PEMS-BAY datasets realize preprocessing and training.

Looking for your reply, thanks sincerely.

Regard!