Unable to reproduce the results of anomaly detection in the paper

[51348761z]

Hello, thank you very much for the work you have done, I downloaded the code and tested the model metrics using the SMD dataset with the following results:

>>>>>>>testing : anomaly_detection_SMD_GPT4TS_SMD_ftM_sl100_ll48_pl0_dm768_nh8_el2_dl1_df768_fc1_ebtimeF_dtTrue_test_0<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<
test 7084
train 7084
Threshold : 3.7866930925845512
pred:    (708400,)
gt:      (708400,)
pred:  (708400,)
gt:    (708400,)
Accuracy : 0.9872, Precision : 0.8746, Recall : 0.8072, F-score : 0.8395 

this is not as the paper stated(Precision:88.89, Recall:84.98, F-score:86.89) using the parameters in the latest scripts file,
Can you please provide your settings?