HarryVolek / PyTorch_Speaker_Verification

PyTorch implementation of "Generalized End-to-End Loss for Speaker Verification" by Wan, Li et al.
BSD 3-Clause "New" or "Revised" License
578 stars 164 forks source link

Overfitting #45

Closed biswasingh closed 5 years ago

biswasingh commented 5 years ago

Hi,

I see the model is overfitting on the mix of data I am using. Can you please suggest some regularisation technique which might help here? Shall I add more data?

Sun Jul 7 10:31:55 2019 Epoch:1700[30/79],Iteration:134251 Loss:0.0008 TLoss:0.9285

Sun Jul 7 10:31:58 2019 Epoch:1700[60/79],Iteration:134281 Loss:0.0011 TLoss:0.6694

Sun Jul 7 10:32:05 2019 Epoch:1701[30/79],Iteration:134330 Loss:0.0631 TLoss:0.4391

nidhal1231 commented 5 years ago

@biswasingh could you please explain me how did you know your model is overfitting . is it in relation with Tloss. I didn't understand the signification of Tloss in the model. Thank you.