FFXZ
FFXZ
@yuanyuan0209 is your loss is decreased? can your share your successful results?
@bailiangze it might be the problem of *.wav file, the timit wav format is different from other. you can try some other wav or use audition software change the format.
@deeplearningzhy do you deal with the problem? I also confused about that.
@zhr1201 the number of validation_samples is random? I can set it without calculating?
how can i deal with it? thanks
I meet same problem, "the other 15 home-made noise types (by USTC): (download at: zip2)". These noises can not be found.
@zfqy do you deal with 0 db problem?
 @zfqy I think you're True. I change noise_only_percent to zero, and then i get correct results as picture shows.
> Hi, thanks for your interest. Both eval_tr_gen and eval_te_gen should use 'test' in the argument because both of them should use the test data generator. The 'train' will use...