preprocess.py use midi_processor.processor to encode my midi files (PIANO-E-COMPETITION), the vocabulary size is 388, but sequence.py get rid off the notes unused in my midi scores, so the vocabulary size in model decreased to 243,
when I run train.py, it reports:
tensorflow.python.framework.errors_impl.InvalidArgumentError: indices[0,1920] = 361 is not in [0, 243) [Op:ResourceGather] name: encoder/embedding/embedding_lookup/
So how to deal with the vocab size mismatch problem?
preprocess.py use midi_processor.processor to encode my midi files (PIANO-E-COMPETITION), the vocabulary size is 388, but sequence.py get rid off the notes unused in my midi scores, so the vocabulary size in model decreased to 243, when I run train.py, it reports:
tensorflow.python.framework.errors_impl.InvalidArgumentError: indices[0,1920] = 361 is not in [0, 243) [Op:ResourceGather] name: encoder/embedding/embedding_lookup/
So how to deal with the vocab size mismatch problem?