search

xiph / LPCNet

Efficient neural speech synthesis
BSD 3-Clause "New" or "Revised" License
1.12k stars 295 forks source link

Concatenate Error #184

Open Kristopher-Chen opened 2 years ago

Kristopher-Chen commented 2 years ago

Hi, have you ever met this problem?

in "training_tf2/lpcnet.py", line 246, in new_lpcnet_model, ValueError: A Concatenate layer requires inputs with matching shapes except for the concat axis. Got inputs shapes: [(128, None, 20), (None, None, 64)]

Thank you!

norihirosunada commented 2 years ago

Any answer to this issue?

I got same problem while trying to train my own model. I think I've followed right steps. I have used src/concat.sh to get input.s16 and used dump_data to get features.f32 and data.s16. And runned training_tf2/train_lpcnet.py. Anything else to concern about?

Thanks!

exercise-book-yq commented 2 years ago

No one can answer to this issue?

norihirosunada commented 1 year ago

Version of Python might be the answer to this issue. I've rebuilt my environment and this error got away. I was using Python 3.6.8 and got this error, but rebuilt the environment with 3.8 and concatenate error didn't appear. Hope this will help you.