vocab_size=feature_encoder.vocabulary_size, word_embedding_dim=48, # input_length=None, input_length=sentence_padding_length, num_labels=4, conv_filter_type=[ [100, 2, 'full'], [100, 4, 'full'], # [100,6,5,'valid'], ], ktop=1, embedding_dropout_rate=0.5, output_dropout_rate=0.5, nb_epoch=10, earlyStoping_patience=5, ) # dcnn.print_model_descibe() # 训练模型 # dcnn.model_from_pickle('model/modelA.pkl') dcnn.fit((feature_encoder.train_padding_index, trian_y), (map(feature_encoder.transform_sentence, test_X), test_y)) quit() print(dcnn.predict(feature_encoder.transform_sentence(test_X[0]))) dcnn.accuracy((map(feature_encoder.transform_sentence, test_X), test_y)) print(dcnn.batch_predict(map(feature_encoder.transform_sentence, test_X))) # 保存模型 # dcnn.save_model('model/modelA.pkl') # 从保存的pickle中加载模型 # print onehot_cnn.predict(feature_encoder.transform_sentence('你好吗'))
# -------------- region end : 3. 初始化CNN模型并训练 --------------- # -------------- region start : 4. 预测 ------------- if verbose > 1: logging.debug('-' * 20) print '-' * 20 logging.debug('4. 预测') print '4. 预测' # -------------- code start : 开始 ------------- y_predict = map(rand_embedding_cnn.predict,test_X_feature) test_data[u'Y_PRED'] = [index_to_label[item] for item in y_predict] data_util.save_data(test_data,path=result_file_path) quit() rand_embedding_cnn.predict(feature_encoder.transform_sentence('你好吗')) print index_to_label[rand_embedding_cnn.predict(feature_encoder.transform_sentence('你好吗'))] y_pred, is_correct, accu,f1 = rand_embedding_cnn.accuracy((test_X_feature, test_y)) logging.debug('F1(macro)为:%f'%(np.average(f1[:-1]))) print 'F1(macro)为:%f'%(np.average(f1[:-1])) test_data[u'IS_CORRECT'] = is_correct test_data[u'PREDICT'] = [index_to_label[item] for item in y_pred] # data_util.save_data(test_data,'tmp.tmp') # quit() data_util.save_data(test_data, path=result_file_path) # -------------- region start : 生成深度特征编码 -------------
verbose=1, batch_size=2, vocab_size=feature_encoder.vocabulary_size, word_embedding_dim=48, # input_length=None, input_length=sentence_padding_length, num_labels=4, conv_filter_type=[[100, 2, 'full'], [100, 4, 'full'], # [100,6,5,'valid'], ], ktop=1, embedding_dropout_rate=0.5, output_dropout_rate=0.5, nb_epoch=10, earlyStoping_patience=5, ) dcnn.print_model_descibe() # 训练模型 # dcnn.model_from_pickle('model/modelA.pkl') dcnn.fit((train_X_features, trian_y), (test_X_features, test_y)) print(dcnn.predict(feature_encoder.transform_sentence(test_X[0]))) dcnn.accuracy((test_X_features, test_y)) print(dcnn.batch_predict(test_X_features)) # 保存模型 # dcnn.save_model('model/modelA.pkl') # 从保存的pickle中加载模型 # print onehot_cnn.predict(feature_encoder.transform_sentence('你好吗'))
if verbose > 2: logging.debug('-' * 20) print('-' * 20) # -------------- region end : 3. 初始化CNN模型并训练 --------------- # -------------- region start : 4. 预测 ------------- if verbose > 1: logging.debug('-' * 20) print('-' * 20) logging.debug('4. 预测') print('4. 预测') # -------------- code start : 开始 ------------- print(index_to_label[dcnn_model.predict(feature_encoder.transform_sentence('你好吗'))]) y_pred, is_correct, accu,f1 = dcnn_model.accuracy((map(feature_encoder.transform_sentence, test_X), test_y)) logging.debug('F1(macro)为:%f'%(np.average(f1[:-1]))) print('F1(macro)为:%f' % (np.average(f1[:-1]))) test_data[u'IS_CORRECT'] = is_correct test_data[u'PREDICT'] = [index_to_label[item] for item in y_pred] # data_util.save_data(test_data,'tmp.tmp') # quit() data_util.save_data(test_data, path=result_file_path) # -------------- code start : 结束 ------------- if verbose > 1: logging.debug('-' * 20)
# -------------- code start : 结束 ------------- if verbose > 2: logging.debug('-' * 20) print('-' * 20) # -------------- region end : 3. 初始化CNN模型并训练 --------------- # -------------- region start : 4. 预测 ------------- if verbose > 1: logging.debug('-' * 20) print('-' * 20) logging.debug('4. 预测') print('4. 预测') # -------------- code start : 开始 ------------- print(index_to_label[dcnn_model.predict( feature_encoder.transform_sentence('你好吗'))]) y_pred, is_correct, accu, f1 = dcnn_model.accuracy( (map(feature_encoder.transform_sentence, test_X), test_y)) logging.debug('F1(macro)为:%f' % (np.average(f1[:-1]))) print('F1(macro)为:%f' % (np.average(f1[:-1]))) test_data[u'IS_CORRECT'] = is_correct test_data[u'PREDICT'] = [index_to_label[item] for item in y_pred] # data_util.save_data(test_data,'tmp.tmp') # quit() data_util.save_data(test_data, path=result_file_path) # -------------- code start : 结束 ------------- if verbose > 1: logging.debug('-' * 20)
if verbose > 2: logging.debug('-' * 20) print '-' * 20 # -------------- region end : 3. 初始化CNN模型并训练 --------------- # -------------- region start : 4. 预测 ------------- if verbose > 1: logging.debug('-' * 20) print '-' * 20 logging.debug('4. 预测') print '4. 预测' # -------------- code start : 开始 ------------- print index_to_label[rand_embedding_cnn.predict(feature_encoder.transform_sentence('你好吗'))] y_pred, is_correct, accu,f1 = rand_embedding_cnn.accuracy((map(feature_encoder.transform_sentence, test_X), test_y)) logging.debug('F1(macro)为:%f'%(np.average(f1[:-1]))) print 'F1(macro)为:%f'%(np.average(f1[:-1])) test_data[u'IS_CORRECT'] = is_correct test_data[u'PREDICT'] = [index_to_label[item] for item in y_pred] # data_util.save_data(test_data,'tmp.tmp') # quit() result_file_path = ''.join(config['result_file_path']) data_util.save_data(test_data, path=result_file_path) # -------------- code start : 结束 ------------- if verbose > 1:
# -------------- region end : 3. 初始化CNN模型并训练 --------------- # -------------- region start : 4. 预测 ------------- if verbose > 1: logging.debug('-' * 20) print '-' * 20 logging.debug('4. 预测') print '4. 预测' # -------------- code start : 开始 ------------- y_predict = map(rand_embedding_cnn.predict, test_X_feature) test_data[u'Y_PRED'] = [index_to_label[item] for item in y_predict] data_util.save_data(test_data, path=result_file_path) quit() rand_embedding_cnn.predict(feature_encoder.transform_sentence('你好吗')) print index_to_label[rand_embedding_cnn.predict( feature_encoder.transform_sentence('你好吗'))] y_pred, is_correct, accu, f1 = rand_embedding_cnn.accuracy( (test_X_feature, test_y)) logging.debug('F1(macro)为:%f' % (np.average(f1[:-1]))) print 'F1(macro)为:%f' % (np.average(f1[:-1])) test_data[u'IS_CORRECT'] = is_correct test_data[u'PREDICT'] = [index_to_label[item] for item in y_pred] # data_util.save_data(test_data,'tmp.tmp') # quit() data_util.save_data(test_data, path=result_file_path) # -------------- region start : 生成深度特征编码 -------------