I wanted to train the network on my own .wav files. I downloaded all the .py files, put the .wav files in the data/test and data/train directories, and ran wavenet.py without any arguments (as is shown in the readme). However, it apparently tried to predict using a nonexistent model:
WARNING (theano.configdefaults): g++ not detected ! Theano will be unable to exe
cute optimized C-implementations (for both CPU and GPU) and will default to Pyth
on implementations. Performance will be severely degraded. To remove this warnin
g, set Theano flags cxx to an empty string.
Using Theano backend.
INFO - wavenet - Running command 'main'
WARNING - wavenet - No observers have been added to this run
INFO - wavenet - Started
Configuration (modified, added, typechanged):
batch_size = 16
data_dir = 'test'
data_dir_structure = 'flat'
debug = False
desired_sample_rate = 4410
dilation_depth = 9
early_stopping_patience = 20
final_l2 = 0
fragment_length = 1152
fragment_stride = 128
keras_verbose = 1
learn_all_outputs = True
nb_epoch = 1000
nb_filters = 256
nb_output_bins = 256
nb_stacks = 1
predict_initial_input = ''
predict_seconds = 1
predict_use_softmax_as_input = False
random_train_batches = False
randomize_batch_order = True
res_l2 = 0
run_dir = None
sample_argmax = False
sample_temperature = 1.0
seed = 476062113
test_factor = 0.1
train_only_in_receptive_field = True
train_with_soft_target_stdev = None
use_bias = False
use_skip_connections = True
use_ulaw = True
optimizer:
decay = 0.0
epsilon = None
lr = 0.001
momentum = 0.9
nesterov = True
optimizer = 'sgd'
INFO - main - Running with seed 476062113
ERROR - wavenet - Failed after 0:00:00!
Traceback (most recent calls WITHOUT Sacred internals):
File "C:\Python34\lib\site-packages\wrapt\wrappers.py", line 522, in __call__
args, kwargs)
File "wavenet.py", line 471, in main
os.mkdir(run_dir)
FileNotFoundError: [WinError 3] System nie może odnaleźć określonej ścieżki: 'models\\run_20170123_174859'
(the last line says it can't find the path) After a bit of trial and error, I decided to install basveeling's branch of Keras (I installed the official one previously). However, that only worsened the situation:
WARNING (theano.configdefaults): g++ not detected ! Theano will be unable to exe
cute optimized C-implementations (for both CPU and GPU) and will default to Pyth
on implementations. Performance will be severely degraded. To remove this warnin
g, set Theano flags cxx to an empty string.
Using Theano backend.
Traceback (most recent call last):
File "wavenet.py", line 23, in <module>
from keras.callbacks import ModelCheckpoint, EarlyStopping, ReduceLROnPlatea
u, CSVLogger
ImportError: cannot import name 'ReduceLROnPlateau'
I'm using Windows 7 64-bit without SP1. My Python version is Python 3.4.3 (v3.4.3:9b73f1c3e601, Feb 24 2015, 22:43:06) [MSCv.1600 32 bit (Intel)] on win32.
I wanted to train the network on my own .wav files. I downloaded all the .py files, put the .wav files in the data/test and data/train directories, and ran wavenet.py without any arguments (as is shown in the readme). However, it apparently tried to predict using a nonexistent model:
(the last line says it can't find the path) After a bit of trial and error, I decided to install basveeling's branch of Keras (I installed the official one previously). However, that only worsened the situation:
I'm using Windows 7 64-bit without SP1. My Python version is
Python 3.4.3 (v3.4.3:9b73f1c3e601, Feb 24 2015, 22:43:06) [MSCv.1600 32 bit (Intel)] on win32
.