TDNN
TDNN copied to clipboard
How to train the model?
Hi Thank you in advance, I am trying to implement the X-vector using your TDNN model, but when I try to train it and the loss can not down and the rate of right is so low. Thank you, could you give me some suggestions? my input: 24-dim MFCC
Yours sincerely
Ma
Hi Thank you in advance, I am trying to implement the X-vector using your TDNN model, but when I try to train it and the loss can not down and the rate of right is so low. Thank you, could you give me some suggestions? my input: 24-dim MFCC
Yours sincerely Ma
i have the same question, when i train the model, the loss always nan, than i exchange the relu and batchnorm, but the eer is very high
Thanks for your reply, I am a freshman and have not to compute the eer because the final exam.
Hi Thank you in advance, I am trying to implement the X-vector using your TDNN model, but when I try to train it and the loss can not down and the rate of right is so low. Thank you, could you give me some suggestions? my input: 24-dim MFCC
Yours sincerely Ma
i have the same question, when i train the model, the loss always nan, than i exchange the relu and batchnorm, but the eer is very high
Hi,thank you first, I make the score with cosine, and about 0.998 between the same speaker utt and 0.993 between the different, so I predict that eer will be high, can you give me some suggestions?
Ma
Hi Thank you in advance, I am trying to implement the X-vector using your TDNN model, but when I try to train it and the loss can not down and the rate of right is so low. Thank you, could you give me some suggestions? my input: 24-dim MFCC
Yours sincerely Ma
i have the same question, when i train the model, the loss always nan, than i exchange the relu and batchnorm, but the eer is very high
Hi ,if you have solved the question ,loss always nan .I have meet the problem too , but i have no method to solve . if you know how please teach me . Thanks!
@qinxiaoyi