The results of reproduction did not meet the expectation

[looking-for-my-magic-bean]

I’m using pytorch to reproduction your work on dcase_t2, the environment built on the windows system. However, using fan_00 as an example, I can’t get a good result as you’ve got (Only got AUC near 0.7). There is my flow chart in the attachment and the parameters in the experiment are from your technical report. Hopefully you can take a look of it and point out the mistakes or omission. Thanks a lot.
Additional context:
Are there any vital tricks that you used have helped improve the efficiency a lot whereas I missed, leading to my works’ deficiency?

[Atriumph]

if name == 'main':

print(list(enumerate_development_datasets()))

for type_, id_ in enumerate_development_datasets():
    _ = MachineDataSet(type_, id_, mode='training')
    _ = MachineDataSet(type_, id_, mode='validation')

[(0, 0), (0, 2), (0, 4), (0, 6), (1, 0), (1, 2), (1, 4), (1, 6), (2, 0), (2, 2), (2, 4), (2, 6), (3, 1), (3, 2), (3, 3), (3, 4), (4, 1), (4, 2), (4, 3), (5, 0), (5, 2), (5, 4), (5, 6)] no EVALUATION_ID_MAP，What's the solution to this situation?