def main():
train_ds = TextDataset.from_folder(IMDB_PATH, name='train', shuffle=True)
valid_ds = TextDataset.from_folder(IMDB_PATH, name='test')
lm_data = [train_ds, valid_ds]
lm_bunch = TextLMDataBunch.create(lm_data, path=LM_PATH)
learner = RNNLearner.language_model(lm_bunch)
n = sum(len(ds) for ds in lm_data)
num_epochs, phases = create_phases(3, n)
callbacks = [
EarlyStopping(learner, patience=2),
SaveModel(learner),
GeneralScheduler(learner, phases)
]
learner.fit(num_epochs, )
def main():
meta = prepare_lyrics(LYRICS_PATH, LYRICS_PATH.parent / 'prepared')
dataset = TextDataset.from_folder(meta.folder)
print(f'Dataset size: {len(dataset)}')