performance of test result

[JongMokKim]

Hello! Thank you for great work.

I'm interested in training the network with my custom data.
To validate your training script, I ran train.py with not changing train parameters (only data path)
What I'm concerning is that the performance is much more poor than you reported:
acc : 92.02, F1 : 56.95, Prec:43.99, Recall : 87.72
(your report : acc :95.5, F1:65, Prec:55, Recall:81)

which parameter should I tune? or is there any other factor can affect the performance?

thanks in advance!

[NickLucche]

Sorry for the delayed answer, I might have submitted the wrong model, let me look into that and I'll get back to you.
In the meantime, you should try running the test.py script instead, making sure to test on the provided test set.