def build_wavenet(batch_size=1, sample_length=64000): config = Config() x = tf.placeholder(tf.float32, shape=[batch_size, sample_length]) graph = config.build({"wav": x}, is_training=False) graph.update({"X": x}) return graph
def load_style_nsynth(initial): """Load the NSynth autoencoder network for stylizing.""" config = Config() with tf.device("/gpu:0"): initial = initial.reshape([1, -1, 1]) # [Batch_size, length, channel] x = tf.Variable(initial) graph = config.build({"wav": x}, is_training=False) graph.update({"X": x}) return graph
def load_nsynth(batch_size=1, sample_length=40000): # Load the NSynth autoencoder network. config = Config() print("Inside load_nsynth function") with tf.device("/device:GPU:0"): print("Loading nsynth") x = tf.placeholder(tf.float32, shape=[batch_size, sample_length]) graph = config.build({"wav": x}, is_training=False) graph.update({"X": x}) return graph
def load_nsynth(batch_size=1, sample_length=64000): """Load the NSynth autoencoder network. Args: batch_size: Batch size number of observations to process. [1] sample_length: Number of samples in the input audio. [64000] Returns: graph: The network as a dict with input placeholder in {"X"} """ config = Config() with tf.device("/gpu:0"): x = tf.placeholder(tf.float32, shape=[batch_size, sample_length]) graph = config.build({"wav": x}, is_training=False) graph.update({"X": x}) return graph