The results on ETTh1 cannot be reproduced

[yzhHoward]

Hello, I try to reproduce this work based on Autoformer.
However, I cannot reproduce the result on ETTh1 especially on long series.
The result in the paper claims that Triformer achieves 0.488 on MSE on the ETTh1, whereas I get the MSE about 0.570.
Can you indicate how to reproduce the results correctly?