I'm getting loss as NAN values

[Guttappa1238]

Hi,

i'm training single custom class using default parameters when i start training i got following results which loss has nan values.

please can anyone explain why i'm getting this?

Thanks advance

[anhtuan85]

Have you set num_classes = 1?
Then you should check that the data and labels are correct.
And check two loss functions smooth_L1_loss and CrossEntropyLoss

[Guttappa1238]

I set it correctlty, still getting same NAN values.

Please can you add readme doc that explains how to train custom objects? which will be helpful for others too

[anhtuan85]

Hi,
I'm quite busy right now. I will update the code and documentation in the next 1-2 weeks.
Thank you!

[Guttappa1238]

Hi, I'm quite busy right now. I will update the code and documentation in the next 1-2 weeks. Thank you!

Oh okay no problem.

Thanks for quick response