def setup_mnist_corpus(datafiles,
labels,
ngram_size,
shuffle=True,
limit_instance_count=None,
subsets=("train", )):
file_loader = partial(load_mnist_files, subsets=subsets)
viewpoints = (MnistVP((28, 28)), )
corpus = Corpus(file_loader, viewpoints)
LOGGER.info('loading data files')
corpus.load_files(datafiles)
LOGGER.info('constructing ngrams')
corpus.set_to_ngram(ngram_size)
if shuffle:
corpus.shuffle_instances()
if limit_instance_count is not None:
corpus.limit_instance_count(limit_instance_count)
return corpus
def setup_midi_corpus(datafiles, ngram_size, rebuild=False):
viewpoints = (MidiPitchVP(min_pitch=36, max_pitch=100), )
corpus = Corpus(load_midi_files, viewpoints)
LOGGER.info('loading data files')
corpus.load_files(datafiles, clear=rebuild)
corpus.set_to_ngram(ngram_size=ngram_size)
return corpus
def setup_interval_corpus(datafiles, ngram_size, rebuild=False):
viewpoints = (MidiPitchIntervalVP(), )
corpus = Corpus(load_midi_files, viewpoints)
LOGGER.info('loading data files')
corpus.load_files(datafiles, clear=rebuild)
corpus.set_to_ngram(ngram_size=1)
return corpus