scores = []
cui_scores = []
for train_idx, test_idx in skf.split(x, y):
tv = TokenVectorizer(list(x[train_idx]),
use_pos=[],
stopword=True,
stemming=True,
as_numpy=True)
train_vec = tv.vectorizer()
in_seq_size = len(train_vec[0])
model = MLP(in_seq=in_seq_size,
out_vec_size=len(collections.Counter(y)),
dropout_rate=0.2,
loss=MODEL_PARAMS['loss'],
optimizer=MODEL_PARAMS['optimizer'])
mlp = model.build_model()
mlp.fit(train_vec,
y_onehot[train_idx],
batch_size=MODEL_PARAMS['batch_size'],
epochs=MODEL_PARAMS['epochs'],
verbose=1)
test_vec = tv.vectorizer(x[test_idx])
score = mlp.evaluate(test_vec, y_onehot[test_idx])
scores.append(score[1])
print(f'SCORE: {score[1]}')
mlp.save(f'../models/mlp_split{split_cnt}.h5', include_optimizer=False)
# CUI訓練用データ作成部 ---
rm_nv_tokens, _ = tv.rm_nv_tokenizer(y_id[train_idx])
rm_nv_vectors, rm_nv_labels = tv.rm_nv_vectorizer(y_id[train_idx])
multi_label_dict = tv.create_multi_label(rm_nv_tokens, rm_nv_labels)
