Issue with moments.pkl about wavegan HOT 11 CLOSED

Hi there.

Can you give me a bit more context? Did you run the script to extract the moments (spectrogram bin means and variances) from your dataset? Was there any output from that script (it may take a while)? Can you locate the moments.pkl file?

python train_specgan.py moments ./train \
	--data_dir ./data/dir_with_mp3s \
	--data_moments_fp ./train/moments.pkl

from wavegan.

Hi Chris,

Thanks for getting back to me.

I am currently running SpecGAN using the srun functionality. I have imported all the required packages to execute the file. Yes, I did run the moments script to extract the mean and variance but I get an error saying that the moments.pkl file cannot be found.

Do you recommend that I update any existing libraries in my system to achieve success?

PS: Big fan of your work :)

from wavegan.

from wavegan.

I think this means that the loader is not successfully loading any data (the spectrogram array is empty). How many audio files does the script say it found? Are they all the exact same format? Do you have librosa installed?

To get an idea of where exactly the script is breaking, change lines 202/203 to:

    while True:
      try:
        _X_lmag = sess.run(X_lmag)
      except Exception as e:
        print(e)
        break

from wavegan.

Hi Chris,

I implemented the code that you suggested and it worked. I had made a big blunder while running the program. My directory extension wasn't correct, so instead of importing audio files, it was importing the folders, resulting in the error.

Thanks for your help!

from wavegan.

Hi Chris,

I have the same issue computing moments from SC09 dataset.

Traceback (most recent call last):
  File "train_specgan.py", line 758, in <module>
    moments(fps, args)
  File "train_specgan.py", line 612, in moments
    _X_lmags = np.concatenate(_X_lmags, axis=0)
ValueError: need at least one array to concatenate

Below is from your checklist above.
len(fps) = 2494
fps[0] = '/home/data/sc09/valid/Eight_060cd039_nohash_0.wav'
librosa==0.6.2 is installed.
and except Exception as e: print(e) gave me

End of sequence
	 [[{{node IteratorGetNext}} = IteratorGetNext[output_shapes=[[1,16384,1,1]], output_types=[DT_FLOAT]](IteratorFromStringHandleV2)]]
	 [[node loader/ExperimentalFunctionBufferingResourceGetNext (defined at /home/wavegan/loader.py:198)  = ExperimentalFunctionBufferingResourceGetNext[output_types=[DT_FLOAT], _device="/job:localhost/replica:0/task:0/device:GPU:0"](loader/ExperimentalFunctionBufferingResource)]]
	 [[{{node Log/_9}} = _Recv[client_terminated=false, recv_device="/job:localhost/replica:0/task:0/device:CPU:0", send_device="/job:localhost/replica:0/task:0/device:GPU:0", send_device_incarnation=1, tensor_name="edge_50_Log", tensor_type=DT_FLOAT, _device="/job:localhost/replica:0/task:0/device:CPU:0"]()]]

Could you help me solving this issue?

from wavegan.

Hi @youngjung. I'm not 100% sure what's going on. Are you using the correct version of tensorflow (1.12.0)? Can you try installing scipy and using --data_fast_wav as a command line input to use scipy instead of librosa?

from wavegan.

Yes I am using tf-1.12.0

$ pip show tensorflow-gpu
Name: tensorflow-gpu
Version: 1.12.0

I tried

$ python train_specgan.py moments ./experiments     --data_dir ~/data/sc09/valid/     --data_moments_fp ./experiments/moments.pkl --data_fast_wav
Found 2494 audio files in specified directory

and it still gives me

End of sequence
	 [[{{node IteratorGetNext}} = IteratorGetNext[output_shapes=[[1,16384,1,1]], output_types=[DT_FLOAT]](IteratorFromStringHandleV2)]]
	 [[node loader/ExperimentalFunctionBufferingResourceGetNext (defined at /home/wavegan/loader.py:198)  = ExperimentalFunctionBufferingResourceGetNext[output_types=[DT_FLOAT], _device="/job:localhost/replica:0/task:0/device:GPU:0"](loader/ExperimentalFunctionBufferingResource)]]
	 [[{{node Log/_9}} = _Recv[client_terminated=false, recv_device="/job:localhost/replica:0/task:0/device:CPU:0", send_device="/job:localhost/replica:0/task:0/device:GPU:0", send_device_incarnation=1, tensor_name="edge_50_Log", tensor_type=DT_FLOAT, _device="/job:localhost/replica:0/task:0/device:CPU:0"]()]]
stat...
len = 0
Traceback (most recent call last):
  File "train_specgan.py", line 760, in <module>
    moments(fps, args)
  File "train_specgan.py", line 614, in moments
    _X_lmags = np.concatenate(_X_lmags, axis=0)
ValueError: need at least one array to concatenate

from wavegan.

@youngjung Have you solved the issue? I am facing the same error.

from wavegan.

I'm having the same issue here with working with train_specgan.py and sc09/train. Can we get a reopen? Happy to help...

python packages: (I'm using virtualenv--very helpful!); Ubuntu 16.04:
absl-py==0.9.0
astor==0.8.1
audioread==2.1.8
cffi==1.14.0
cycler==0.10.0
decorator==4.4.2
gast==0.3.3
grpcio==1.29.0
h5py==2.10.0
importlib-metadata==1.6.1
joblib==0.14.1
Keras-Applications==1.0.8
Keras-Preprocessing==1.1.2
kiwisolver==1.1.0
librosa==0.6.2
llvmlite==0.32.1
Markdown==3.2.2
matplotlib==3.0.2
numba==0.47.0
numpy==1.16.4
protobuf==3.12.2
pycparser==2.20
pyparsing==2.4.7
python-dateutil==2.8.1
resampy==0.2.2
scikit-learn==0.22.2.post1
scipy==1.0.0
six==1.15.0
SoundFile==0.10.3.post1
tensorboard==1.12.2
tensorflow-gpu==1.12.0
termcolor==1.1.0
Werkzeug==1.0.1
zipp==1.2.0

from wavegan.

I have same error.

from wavegan.