search

andabi / deep-voice-conversion

Deep neural networks for voice conversion (voice style transfer) in Tensorflow
MIT License
3.92k stars 843 forks source link

Fix broken code #30

Open akshay-ap opened 6 years ago

akshay-ap commented 6 years ago

I have fixed all the issues with the code. Now It runs successfully on my machine. To run 

python3 train1.py case1
python3 train2.py case1 case2
tensorboard --logdir="logdir"
akshay-ap commented 6 years ago

There are some unnecessary logs that I had used for debugging. I shall remove those some time later

Techmind commented 6 years ago

Yo, i found out that your version at least works for me. But why did y change number of epochs and batch_size ?

akshay-ap commented 6 years ago

@Techmind I changed batch_size because I was getting out of memory error in GPU. So, reducing batch_size helped to overcome it. I changing number of epochs doesn't make any difference in program execution. I change number of epochs based on training needed.

guang commented 6 years ago

@akshay-ap thanks for the PR man saved me a lotta time debugging

asssssasssssin commented 6 years ago

I opened tensorboard to check the result. I used arctic dataset. There are A and A_1 which are original sound. And B, B_1 have nothing. Why?

Harsh2040 commented 6 years ago

Provide Pretrained weights of train2.py

thepyninja commented 6 years ago

there is an error while running train1.py:

ImportError: cannot import name preemphasis

Amey1109 commented 4 years ago

Provide Pretrained weights of train2.py

I opened tensorboard to check the result. I used arctic dataset. There are A and A_1 which are original sound. And B, B_1 have nothing. Why?

Can you please provide us the pretrained weights of train2