def test_utt2spk(self, prep_org_data):
corp = bkw.create_corpus(tgt_dir=tgt_dir, speakers=["Mark Djandiomerr"])
assert len(corp.speakers) == 1
assert len(corp.get_train_fns()) < NUM_UTTERS / 2
corp = bkw.create_corpus(tgt_dir=tgt_dir)
assert len(corp.speakers) == NUM_SPEAKERS
assert len(corp.get_train_fns()) == NUM_UTTERS
def test_deterministic_2(self, prep_org_data):
corp_1 = bkw.create_corpus(tgt_dir=tgt_dir)
# Remove the prefix files.
os.remove(str(corp_1.train_prefix_fn))
os.remove(str(corp_1.valid_prefix_fn))
os.remove(str(corp_1.test_prefix_fn))
corp_2 = bkw.create_corpus(tgt_dir=tgt_dir)
assert corp_1.utterances != None
assert corp_1.utterances == corp_2.utterances
assert len(corp_1.utterances) == NUM_UTTERS
assert set(corp_1.get_train_fns()[0]) == set(corp_2.get_train_fns()[0])
assert set(corp_1.get_valid_fns()[0]) == set(corp_2.get_valid_fns()[0])
assert set(corp_1.get_test_fns()[0]) == set(corp_2.get_test_fns()[0])
def train_bkw(num_layers: int) -> None:
exp_dir = prep_exp_dir(directory=config.TEST_EXP_PATH)
corp = bkw.create_corpus(tgt_dir=Path(config.TEST_DATA_PATH) / "bkw")
cr = CorpusReader(corp)
model = rnn_ctc.Model(exp_dir,
cr,
num_layers=num_layers,
hidden_size=250)
model.train(min_epochs=40)
def preprocessed_corpus(prep_org_data):
""" Ensures corpus preprocessing happens before any of the tests
run that rely on it"""
return bkw.create_corpus(tgt_dir=tgt_dir)