class_weight option in model.fit_generator

[alirezaniki]

Dear @smousavi05

Recently, I've faced a problem running EQT training.py code. The issue arises from the class_weight option in the model.fit_generator which is by default set to {0: 0.11, 1: 0.89}. The error (which originates from the TF version) asks me to replace the keys (0 and 1) with "detector", "picker_P", and "picker_S". As there are 3 classes in EQT, my question is that what classes do the keys 0 and 1 refer to?

Thank you in advance, Alireza.

[alirezaniki]

Dear @smousavi05 My apologies for mixing the questions, but I am facing some weird usage of memory while training my dataset (It's only a ~6 Gb hdf5 file and a 16 Mb metadata file). Have you ever faced such a thing? It's occupying more and more memory after each epoch.

[smousavi05]

No, that is not Normal. I never had that problem before

On Sat, Apr 23, 2022, 2:08 PM Alireza Niksejel @.***> wrote:

Dear @smousavi05 https://github.com/smousavi05 My apologies for mixing the questions, but I am facing some weird usage of memory while training my dataset (It's only a ~6 Gb hdf5 file and a 16 Mb metadata file). Have you ever faced such a thing? It's occupying more and more memory after each epoch.

[image: Screenshot from 2022-04-23 17-26-55] https://user-images.githubusercontent.com/24492517/164945963-e0d55e25-2535-49df-ae62-feba1237e864.png

— Reply to this email directly, view it on GitHub https://github.com/smousavi05/EQTransformer/issues/127#issuecomment-1107649768, or unsubscribe https://github.com/notifications/unsubscribe-auth/ADHWAPC3WFBLOSFMBTQX3R3VGRRF3ANCNFSM5TTHGDMA . You are receiving this because you were mentioned.Message ID: @.***>

[alirezaniki]

Thank you for your response

[Chouyuhin]

Thank you for your response

Hello， I have the same problem with you , did you figure it out? Lokking forward to your reply. If necessary，my email is chouyuhin777@163.com

Thanks!