np.save(result_src, total_y_pre)
return y, total_y_pre
if __name__ == "__main__":
# 模型路径
model_save_src = "data/model/2_layer_lstm_model"
num_category = 9
# 向量化后的数据集
x_src = "data/vectorized_data/test/x.npy"
y_src = "data/vectorized_data/test/y.npy"
result_src = "data/results/rnn_pre.npy"
vocab_src = "data/middle_result/vocab.npy"
data = Data()
vocab, _ = data.load_vocab(vocab_src)
# 模型
config = TRNNConfig()
config.vocab_size = len(vocab)
model = TextRNN(config)
# 测试
print("Begin Testing")
start_time = time.time()
y, y_pre = test(x_src, y_src, result_src)
print("the time is {}".format(get_time_dif(start_time)))
# 评估
precision_score, recall_score, f1_val, accuracy = evaluate(y, y_pre)
for i in range(num_category):
if total_batch - last_improved > require_improvement:
# 验证集正确率长期不提升，提前结束训练
print("No optimization for a long time, auto-stopping...")
flag = True
break # 跳出循环
if flag: # 同上
break
print("the best acc on validation is {}".format(best_acc_val))
if __name__ == '__main__':
train_dir = "data/vectorized_data/train"
val_dir = "data/vectorized_data/validation"
vocab_dir = "data/file_dict/train/vocab.npy"
save_dir = 'data/model2'
data_process = Data()
config = CharCNNConfig()
if not os.path.exists(vocab_dir):
data_process.build_vocab(train_dir, vocab_dir)
words, word_to_id = data_process.load_vocab(vocab_dir)
config.vocab_size = len(words)
model = CharCNN(config)
train()
