1. # coding=utf-8
    2. import os
    3. from multiprocessing import cpu_count
    4. import numpy as np
    5. import paddle
    6. import paddle.fluid as fluid
    7. #import paddlehub as hub
    1. class classify():
    2.     data_root_path = ""
    3.     dict_path = "data/data9658/dict.txt"
    4.     test_data_path = "data/data9658/Test_IDs.txt"
    5.     model_save_dir = "work/classify_nn/"
    6.     save_path = 'work/result.txt'
    8.     def train(self):
    9.         #
    10.         # 学员自行填充训练代码，
    11.         # 完成模型训练。
    12.         #
    13.         print('训练模型保存完成！')
    14.         self.test(self)
    16.         # 获取数据
    17.     def get_data(self,sentence):
    18.         # 读取数据字典
    19.         with open(self.dict_path, 'r', encoding='utf-8') as f_data:
    20.             dict_txt = eval(f_data.readlines()[0])
    21.         dict_txt = dict(dict_txt)
    22.         # 把字符串数据转换成列表数据
    23.         keys = dict_txt.keys()
    24.         data = []
    25.         for s in sentence:
    26.             # 判断是否存在未知字符
    27.             if not s in keys:
    28.                 s = '<unk>'
    29.             data.append(int(dict_txt[s]))
    30.         return data
    32.     def test(self):
    33.         data = []
    34.         # 获取预测数据
    35.         with open(self.test_data_path, 'r', encoding='utf-8') as test_data:
    36.             lines = test_data.readlines()
    37.         for line in lines:
    38.             tmp_sents = []
    39.             for word in line.strip().split(','):
    40.                 tmp_sents.append(int(word))
    41.             data.append(tmp_sents)
    42.         print ('数据加载完毕，数据长度：',len(data))
    43.         #a=self.get_data(self, 'w我是共产主义接班人！')
    44.         #data=[a]
    45.         def load_tensor(data):
    46.             # 获取每句话的单词数量
    47.             base_shape = [[len(c) for c in data]]
    48.             # 创建一个执行器，CPU训练速度比较慢
    49.             #place = fluid.CPUPlace()
    50.             place = fluid.CUDAPlace(0)
    51.             # 生成预测数据
    52.             print('loading tensor')
    53.             tensor_words = fluid.create_lod_tensor(data, base_shape, place)
    55.             #infer_place = fluid.CPUPlace()
    56.             infer_place = fluid.CUDAPlace(0)
    57.             # 执行预测
    58.             infer_exe = fluid.Executor(infer_place)
    59.             # 进行参数初始化
    60.             infer_exe.run(fluid.default_startup_program())
    61.             print('feeder')
    63.             # 从模型中获取预测程序、输入数据名称列表、分类器
    64.             print('loading model')
    65.             [infer_program, feeded_var_names, target_var] = fluid.io.load_inference_model(dirname=self.model_save_dir, executor=infer_exe)
    66.             result=[]
    67.             result = infer_exe.run(program=infer_program,
    68.                                 feed={feeded_var_names[0]: tensor_words},
    69.                                 fetch_list=target_var)
    70.             names = ["财经", "彩票", "房产", "股票", "家居", "教育", "科技",
    71.                     "社会", "时尚", "时政", "体育", "星座", "游戏", "娱乐"]
    72.             # 输出结果
    73.             print('writting')
    74.             for i in range(len(data)):
    75.                 lab = np.argsort(result)[0][i][-1]
    76.                 #print('预测结果标签为：%d， 名称为：%s， 概率为：%f' % (lab, names[lab], result[0][i][lab]))
    77.                 with open(self.save_path, 'a', encoding='utf-8') as ans:
    78.                     #print (names[lab])
    79.                     ans.write( names[lab]+"\n")
    80.             ans.close()
    81.         print('loading 1/4 data')
    82.         load_tensor(data[:int(len(data)/4)])
    83.         print('loading 2/4 data')
    84.         load_tensor(data[int(len(data)/4):2*int(len(data)/4)])
    85.         print('loading 3/4 data')
    86.         load_tensor(data[2*int(len(data)/4):3*int(len(data)/4)])
    87.         print('loading 4/4 data')
    88.         load_tensor(data[3*int(len(data)/4):])
    89.         print('测试输出已生成！')
    91. if __name__ == "__main__":
    92.     classify.train(classify)
    93.     #classify.test(classify)
    1. !rm -rf submit.sh
    2. !wget -O submit.sh http://ai-studio-static.bj.bcebos.com/script/submit.sh
    3. !sh submit.sh work/result.txt 密码

    请点击此处查看本环境基本用法.

    Please click here for more detailed instructions.