Hello, could someone help me please? Im using tacotron training notebook on google collab and I have a error :
hparams to Tune
#These two are the most important
hparams.batch_size = 10 # Controls how fast the model trains. Don’t set this too high, or else it will GPU will OOM (out of memory). 30-ish is usually a good number if you have a bigger dataset. If the number of audio files is more than/about the same as this number, it won’t train properly, and you won’t be able to use it.
hparams.epochs = 500 # Maxmimum epochs (number of times the AI looks through the dataset) to train
#The rest aren’t that important
hparams.p_attention_dropout=0.1
hparams.p_decoder_dropout=0.1
hparams.decay_start = 15000 # wait till decay_start to start decaying learning rate
hparams.A_ = 5e-4 # Start/Max Learning Rate
hparams.B_ = 8000 # Decay Rate
hparams.C_ = 0 # Shift learning rate equation by this value
hparams.min_learning_rate = 1e-5 # Min Learning Rate
generate_mels = True # Don’t change
hparams.show_alignments = True
alignment_graph_height = 600
alignment_graph_width = 1000
hparams.load_mel_from_disk = True
hparams.ignore_layers = [] # Layers to reset (None by default, other than foreign languages this param can be ignored)
torch.backends.cudnn.enabled = hparams.cudnn_enabled
torch.backends.cudnn.benchmark = hparams.cudnn_benchmark
output_directory = ‘/content/drive/My Drive/colab/outdir’ # Location to save Checkpoints
log_directory = ‘/content/tacotron2/logs’ # Location to save Log files locally
log_directory2 = ‘/content/drive/My Drive/colab/logs’ # Location to copy log files (done at the end of each epoch to cut down on I/O)
checkpoint_path = output_directory+(r’/’)+model_filename
if generate_mels:
create_mels()
RuntimeError: shape ‘[1, 1, 83256]’ is invalid for input of size 166512