def generate(n_docs=300, n_words=10, n_sent_length=5, n_hidden=8):
words = fake_data.fake_data(n_docs, n_words, n_sent_length, n_hidden)
words_flat = words.ravel()
doc_ids = np.repeat(np.arange(words.shape[0]).astype('int32'),
n_sent_length)
doc_ids = doc_ids.ravel()
_, counts = np.unique(words_flat, return_counts=True)
model = LDA2Vec(n_words, n_hidden, counts, n_samples=1)
return model, words_flat, doc_ids
def generate(n_docs=300, n_words=100, n_sent_length=5, n_hidden=8):
words = fake_data.fake_data(n_docs, n_words, n_sent_length, n_hidden)
words_flat = words.ravel()
doc_ids = np.repeat(
np.arange(words.shape[0]).astype('int32'), n_sent_length)
doc_ids = doc_ids.ravel()
_, counts = np.unique(words_flat, return_counts=True)
model = LDA2Vec(n_words, n_hidden, counts, n_samples=1)
return model, words_flat, doc_ids
def test_fake_data():
n_docs = 100
n_words = 10
n_hidden = 2
n_sent_length = 5
data = fake_data.fake_data(n_docs, n_words, n_sent_length, n_hidden)
assert data.dtype == np.dtype('int32')
assert data.shape[0] == n_docs
assert data.shape[1] == n_sent_length
assert np.max(data) <= n_words - 1
