基于 NLP 情感识别的电子宠物社交应用

• 萌宠菠菠乐园
• 2024-08-18 12:26

0 介绍视频

https://www.bilibili.com/video/BV1k44y117aQ/

1 项目背景

社会背景

当前，中国经济高速增长，社会节奏快，人口逐渐老龄化，城镇化进展显著。

年轻人的生活压力大，精神情感需求得不到满足，中老年人空巢特征明显，同样缺少情感寄托和社交途径，代际间社交隔阂明显。

市场呼吁新模态游戏、社交应用的出现，满足情感需求，「旅行青蛙」、「摩尔庄园」或长或短的火爆都是由此，但是，传统的宠物养成囿于固定的养成模式，仅有的定制化方式局限于可选装饰等，同时，旧有匿名社交难以保证对文字图片等内容的有效审查。

用户需求分析

经过团队成员的前期调研，可以发现应用的目标用户主要存在以下痛点：

情感需求无法得到满足，缺少精神寄托

缺少触及灵魂的社交，过于直接的交流让社交恐惧症患者难为情，同时害怕隐私泄露

饱受信息轰炸，在推荐算法中迷失自我，很少反思自我

空间距离给不同关系的人之间情感交流造成阻碍，同时快节奏的生活使得缺少时间和亲友交流感情

通过访谈和问卷的形式，我们总结出了以下需求：

希望能有合适精神寄托的应用，能让情感充分投入，同时得到更加具象的表现

设计基于情感表达的交互模式，通过对文本、语音等多种材料的情感分析转化成宠物成长的参数

希望能够有方便安全有趣的匿名社交应用

定义「模糊社交」，以宠物代表自我的开放式社交

希望应用能真正反映自我，从应用中窥见自我

多种数据生成参数，参数表达生成宠物，宠物即是自我的化身

希望能够方便地和亲友情感交流，避免直接交流的尴尬

提供能够影响现实的宠物养成交互方式，对信息进行有效的过滤和转化

用户画像

林未

女，17岁，高中生 形象：天真烂漫的高中生 场景：希望能和姐妹一起使用新潮好玩的社交应用，一起玩养成游戏，同时父母也很关心自己，担心所使用应用的隐私问题 需求：有趣好玩安全的养成应用、社交应用

张洋

女，21岁，大学生 形象：和男友异地恋的大学生 场景：平时和男友比较难见面，希望能够和男友借助一些移动应用培养感情，记录点滴 需求：带有记录机制反映感情可以点对点交互的应用

黄征

男，27岁，互联网大厂员工 形象：从家乡出来独自在大城市打拼，忙于工作很少有时间和父母沟通感情 场景：随着工作逐步进入正轨和生产逐渐恢复越来越忙，和父母的通讯越来越少，有时会担心远在家乡的父母的健康状况 需求：能够减少和长辈交流情感的尴尬，同时一定程度上进行健康监控的应用

杨建

男，55岁，接近退休 形象：在家乡小城事业单位工作，平时并不忙碌，挂念去大城市打拼的女儿 场景：经常在微信上和女儿嘘寒问暖，表示关心，但女儿确实忙碌，有时会忘记回复消息，有时不知道女儿的状况 需求：能够激励孩子和父母交流，同时一定程度上反映子女状态的应用

竞品分析

产品名 中国式家长 旅行青蛙 摩尔庄园 ZEPETO Soul 产品定位 基于消除类游戏和日程事件的养成类游戏 简单方便的放置类养成轻手游 充满情怀结合社交的庄园手游 创建自己的3D虚拟人物捏脸社交应用 基于兴趣匹配算法的年轻人社交应用 目标用户 休闲类网络游戏玩家 以女性用户为主的一二线城市20-40岁的用户 以大学生为主的旧摩尔庄园网游玩家 女性用户居多的学生和刚上班不久的年轻人 生活乏味燥，环境单一，交友欲望强的年轻人 功能分析 随机故事线
养成游戏
消除类小游戏 放置游戏
物品收集
图片分享
庄园游戏
家园社交 3D人物捏脸
好友社交
虚拟合影、游戏 兴趣算法匹配
聊天
动态广场 盈利模式 游戏购买 充值：购买道具
电商增值业务 游戏充值：皮肤售卖等 充值：装扮售卖 匹配增值服务
广告业务
电商业务 优势 具有中国特色的游戏故事线设计
设计风格 设计风格
有特色的物品收集 经典IP
社交宣传 虚拟人物设计
社交宣传 匹配算法 缺点不足 游戏玩法单一 游戏玩法单一 游戏玩法单一
换皮手游 社交方式有局限 内容审查运营压力大

可以看到，市面上的同类电子宠物养成应用或社交应用存在相当的不足。纯粹的电子宠物养成应用往往只能采用传统的交互模式，玩法单一；传统的社交应用往往存在着内容审查和用户粘性方面的问题。因此我们提出将电子宠物养成与社交结合的方式，这一创意属于全球首创。

2 项目介绍

团队项目成果是一款应用新技术用情感养成宠物，让宠物诠释自己、代表自己的模糊社交应用。

成果运行于 iOS 平台上，通过交互和情感投入转化参数实现宠物养成，通过宠物开展社交，打造模糊社交新模态。

通过这一应用，你可以定制并长期培养自己的专属宠物，通过喂养日记，语音交互，提供定位和健康运动信息，甚至使用其他应用等信息让你的宠物逐渐成长，你的「一举一动」都会对宠物产生影响，使其作出反应，毛玻璃下富于变化的宠物形象满足你的情感需求。

同时，用户可以进入社交模式，在社区中随机浏览和临时领养他人宠物，通过观察和喂养宠物感受宠物主人的生活方式和性格情感，并在社交模式中完成任务解锁特殊成就，崭新的社交形式赋予模糊社交更多可能。

社交模式还支持点对点的宠物长期绑定机制，在绑定机制下，宠物有着更丰富的交互模式，以及允许进行跌倒检测、屏幕时间监控等以满足亲子家庭等关系的特殊需求。

同时，应用可以与智能机器人联动，扩展感知反馈和作用范围。

APP UI界面；

产品突破性成果

情感养成宠物：全屏幕观察隐藏在毛玻璃下的小宠物，通过触摸屏幕和晃动手机和ta玩耍，试试和ta玩玩游戏，还可以向ta投喂日记，你的行为举止和喜怒哀乐会让ta慢慢成长成「你的样子」 参数表达诠释：投喂日记，提供定位，运动感应等多样化交互均可提供参数，多自由度的参数定量描述生活情感，参数化表达生成真正意义的「专属」宠物 定义模糊社交：只见宠物，不见其人，你看到的小宠物背后反映的是对方的一举一动和喜怒哀乐，从虚拟中透见真实，同时杜绝传统匿名社交监管困难、审查不严、隐私泄露弊病 软硬件联动：通过团队成员自主研发的智能机器人，实现将网络空间扩展至现实世界的效果。

3 技术方案

3.1 数据集

数据由两列组成，以制表符（'t'）分隔，第一列是情绪分类的类别（0表示消极；1表示中性；2表示积极），第二列是以空格分词的中文文本，如下示例，文件为 utf8 编码。

label text_a 0 谁 骂人 了 ？ 我 从来 不 骂人 ， 我 骂 的 都 不是 人 ， 你 是 人 吗 ？ 1 我 有事 等会儿 就 回来 和 你 聊 2 我 见到 你 很高兴 谢谢 你 帮 我

我们使用一份已标注的、经过分词预处理的机器人聊天数据集，其目录结构如下：

. ├── train.tsv # 训练集 ├── dev.tsv # 验证集 ├── test.tsv # 测试集 ├── infer.tsv # 待预测数据 ├── vocab.txt # 词典

3.2 具体技术实现

可以使用的模型类型如下表所示，只需要更换train_parameters中的"model_type"参数即可。

模型名 train_parameters BOW "model_type": "bow_net" CNN "model_type": "cnn_net" GRU "model_type": "gru_net" LSTM "model_type": "lstm_net" Bi-LSTM "model_type": "bilstm_net" TEXTCNN "model_type": "textcnn_net"

3.2.1 BOW模型

BOW（Bag Of Words）模型是个在自然语言处理和信息检索(IR)下被简化的表达模型。该模型是一个非序列模型，使用基本的全连接结构，忽略掉文本的语法和语序等要素，将其仅仅看作是若干个词汇的集合，文档中每个单词的出现都是独立的。BoW使用一组无序的单词(words)来表达一段文字或一个文档。

3.2.2 CNN模型

CNN（Convolutional Neural Networks），是一个基础的序列模型，能处理变长序列输入，提取局部区域之内的特征。

3.2.3 GRU模型

GRU（Gated Recurrent Unit），序列模型，能够较好地解决序列文本中长距离依赖的问题。

3.2.4 LSTM模型

LSTM（Long Short Term Memory）模型是一种特殊的RNN模型，是为了解决RNN模型梯度弥散的问题而提出的，并且能够较好地解决序列文本中长距离依赖的问题。

3.2.5 Bi-LSTM模型

Bi-LSTM（Bidirectional Long Short Term Memory），序列模型，采用双向LSTM结构，更好地捕获句子中的语义特征。

3.2.6 TEXTCNN模型

TEXTCNN模型将卷积神经网络CNN应用到文本分类任务，利用多个不同size的kernel来提取句子中的关键信息（类似于多窗口大小的n-gram），从而能够更好地捕捉局部相关性。

3.3 实现步骤

In [1]

# 代码的整体结构如下 # |---data/ # 训练数据 # |---logs/ # 日志文件 # |---pretrained_model/ # 预训练参数 # |---train_model/ # 训练过程中保存的参数，分散形式 # |---infer_model/ # 固化后的推理模型，参数为分散形式 # |---work/ # | |---model.py # 模型结构 # | |---utils.py # 日志，模型所需工具函数相关 # |---config.py # 配置文件，强烈建议先看看配置文件 # |---train.py # |---freeze.py # |---eval.py # |---infer.py

3.3.1 解压训练数据

In [2]

!cd data/data9740/ && unzip -qo 对话情绪识别.zip !cd data/data9740/ && mv data/* . !cd data/data9740/ && mv train.tsv train.txt !cd data/data9740/ && mv test.tsv test.txt !cd data/data9740/ && mv infer.tsv infer.txt !cd data/data9740/ && mv dev.tsv dev.txt

3.3.2 BOW模型的训练与预测

3.3.2.1 训练模型

更改train_parameters中"model_type"为"bow_net"

In [3]

!python train.py

2021-08-02 23:29:04,470-INFO: start train text classification, train params: {'class_dim': 3, 'sample_count': 9656, 'label_dict': {'消极': 0, '中性': 1, '积极': 2}, 'data_dir': 'data/data9740', 'vocabulary': 'vocab.txt', 'train_file_list': 'train.txt', 'label_file': 'label_list.txt', 'save_model_dir': './train_model', 'save_freeze_dir': './infer_model', 'pretrained_model_dir': './pretrained_model', 'continue_train': False, 'pretrained': False, 'model_type': 'bow_net', 'num_epochs': 12, 'train_batch_size': 128, 'use_gpu': True, 'sample_frequency': 10, 'rsm_strategy': {'learning_rate': 0.001, 'lr_epochs': [6, 8, 10], 'lr_decay': [1, 0.5, 0.25, 0.1]}, 'momentum_strategy': {'learning_rate': 0.001, 'lr_epochs': [6, 8, 10], 'lr_decay': [1, 0.5, 0.25, 0.1]}, 'sgd_strategy': {'learning_rate': 0.001, 'lr_epochs': [6, 8, 10], 'lr_decay': [1, 0.5, 0.25, 0.1]}, 'adam_strategy': {'learning_rate': 0.001}} 2021-08-02 23:29:04,470 - [8518] [line:133] - INFO: start train text classification, train params: {'class_dim': 3, 'sample_count': 9656, 'label_dict': {'消极': 0, '中性': 1, '积极': 2}, 'data_dir': 'data/data9740', 'vocabulary': 'vocab.txt', 'train_file_list': 'train.txt', 'label_file': 'label_list.txt', 'save_model_dir': './train_model', 'save_freeze_dir': './infer_model', 'pretrained_model_dir': './pretrained_model', 'continue_train': False, 'pretrained': False, 'model_type': 'bow_net', 'num_epochs': 12, 'train_batch_size': 128, 'use_gpu': True, 'sample_frequency': 10, 'rsm_strategy': {'learning_rate': 0.001, 'lr_epochs': [6, 8, 10], 'lr_decay': [1, 0.5, 0.25, 0.1]}, 'momentum_strategy': {'learning_rate': 0.001, 'lr_epochs': [6, 8, 10], 'lr_decay': [1, 0.5, 0.25, 0.1]}, 'sgd_strategy': {'learning_rate': 0.001, 'lr_epochs': [6, 8, 10], 'lr_decay': [1, 0.5, 0.25, 0.1]}, 'adam_strategy': {'learning_rate': 0.001}} 2021-08-02 23:29:04,470-INFO: create place, use gpu: True 2021-08-02 23:29:04,470 - [8518] [line:134] - INFO: create place, use gpu: True 2021-08-02 23:29:04,901-INFO: define input data tensor 2021-08-02 23:29:04,901 - [8518] [line:139] - INFO: define input data tensor 2021-08-02 23:29:04,901-INFO: build custom reader 2021-08-02 23:29:04,901 - [8518] [line:144] - INFO: build custom reader 2021-08-02 23:29:04,923-INFO: build network bow_net 2021-08-02 23:29:04,923 - [8518] [line:147] - INFO: build network bow_net 2021-08-02 23:29:04,937-INFO: build optimizer 2021-08-02 23:29:04,937 - [8518] [line:164] - INFO: build optimizer W0802 23:29:04.954095 8518 device_context.cc:252] Please NOTE: device: 0, CUDA Capability: 70, Driver API Version: 10.1, Runtime API Version: 9.0 W0802 23:29:04.958819 8518 device_context.cc:260] device: 0, cuDNN Version: 7.6. 2021-08-02 23:29:18,714-INFO: current pass: 0, start read text 2021-08-02 23:29:18,714 - [8518] [line:179] - INFO: current pass: 0, start read text 2021-08-02 23:29:18,754-INFO: pass 0, batch 0, accuracy:0.3203125 loss:1.10135817527771 time:0.04 sec 2021-08-02 23:29:18,754 - [8518] [line:200] - INFO: pass 0, batch 0, accuracy:0.3203125 loss:1.10135817527771 time:0.04 sec 2021-08-02 23:29:18,968-INFO: pass 0, batch 10, accuracy:0.734375 loss:1.0824909210205078 time:0.02 sec 2021-08-02 23:29:18,968 - [8518] [line:200] - INFO: pass 0, batch 10, accuracy:0.734375 loss:1.0824909210205078 time:0.02 sec 2021-08-02 23:29:19,186-INFO: pass 0, batch 20, accuracy:0.7421875 loss:1.0645222663879395 time:0.02 sec 2021-08-02 23:29:19,186 - [8518] [line:200] - INFO: pass 0, batch 20, accuracy:0.7421875 loss:1.0645222663879395 time:0.02 sec 2021-08-02 23:29:19,399-INFO: pass 0, batch 30, accuracy:0.7890625 loss:1.0455031394958496 time:0.02 sec 2021-08-02 23:29:19,399 - [8518] [line:200] - INFO: pass 0, batch 30, accuracy:0.7890625 loss:1.0455031394958496 time:0.02 sec 2021-08-02 23:29:19,612-INFO: pass 0, batch 40, accuracy:0.75 loss:1.0346649885177612 time:0.02 sec 2021-08-02 23:29:19,612 - [8518] [line:200] - INFO: pass 0, batch 40, accuracy:0.75 loss:1.0346649885177612 time:0.02 sec 2021-08-02 23:29:19,824-INFO: pass 0, batch 50, accuracy:0.6640625 loss:1.0345054864883423 time:0.02 sec 2021-08-02 23:29:19,824 - [8518] [line:200] - INFO: pass 0, batch 50, accuracy:0.6640625 loss:1.0345054864883423 time:0.02 sec 2021-08-02 23:29:20,036-INFO: pass 0, batch 60, accuracy:0.640625 loss:1.0276448726654053 time:0.02 sec 2021-08-02 23:29:20,036 - [8518] [line:200] - INFO: pass 0, batch 60, accuracy:0.640625 loss:1.0276448726654053 time:0.02 sec 2021-08-02 23:29:20,257-INFO: pass 0, batch 70, accuracy:0.71875 loss:1.0030725002288818 time:0.02 sec 2021-08-02 23:29:20,257 - [8518] [line:200] - INFO: pass 0, batch 70, accuracy:0.71875 loss:1.0030725002288818 time:0.02 sec 2021-08-02 23:29:20,363-INFO: pass 0 train result, current pass mean accuracy:0.7188640103528374 loss:1.0402885181339163 2021-08-02 23:29:20,363 - [8518] [line:207] - INFO: pass 0 train result, current pass mean accuracy:0.7188640103528374 loss:1.0402885181339163 2021-08-02 23:29:20,363-INFO: temp save pass 0 train result, current best pass mean accuracy:0.7188640103528374 2021-08-02 23:29:20,363 - [8518] [line:209] - INFO: temp save pass 0 train result, current best pass mean accuracy:0.7188640103528374 2021-08-02 23:29:20,976-INFO: current pass: 1, start read text 2021-08-02 23:29:20,976 - [8518] [line:179] - INFO: current pass: 1, start read text 2021-08-02 23:29:20,996-INFO: pass 1, batch 0, accuracy:0.703125 loss:0.996056318283081 time:0.02 sec 2021-08-02 23:29:20,996 - [8518] [line:200] - INFO: pass 1, batch 0, accuracy:0.703125 loss:0.996056318283081 time:0.02 sec 2021-08-02 23:29:21,217-INFO: pass 1, batch 10, accuracy:0.7421875 loss:0.9736344814300537 time:0.02 sec 2021-08-02 23:29:21,217 - [8518] [line:200] - INFO: pass 1, batch 10, accuracy:0.7421875 loss:0.9736344814300537 time:0.02 sec 2021-08-02 23:29:21,569-INFO: pass 1, batch 20, accuracy:0.75 loss:0.9586015343666077 time:0.02 sec 2021-08-02 23:29:21,569 - [8518] [line:200] - INFO: pass 1, batch 20, accuracy:0.75 loss:0.9586015343666077 time:0.02 sec 2021-08-02 23:29:21,795-INFO: pass 1, batch 30, accuracy:0.703125 loss:0.9604085683822632 time:0.02 sec 2021-08-02 23:29:21,795 - [8518] [line:200] - INFO: pass 1, batch 30, accuracy:0.703125 loss:0.9604085683822632 time:0.02 sec 2021-08-02 23:29:22,006-INFO: pass 1, batch 40, accuracy:0.7421875 loss:0.9418747425079346 time:0.02 sec 2021-08-02 23:29:22,006 - [8518] [line:200] - INFO: pass 1, batch 40, accuracy:0.7421875 loss:0.9418747425079346 time:0.02 sec 2021-08-02 23:29:22,234-INFO: pass 1, batch 50, accuracy:0.78125 loss:0.9146920442581177 time:0.02 sec 2021-08-02 23:29:22,234 - [8518] [line:200] - INFO: pass 1, batch 50, accuracy:0.78125 loss:0.9146920442581177 time:0.02 sec 2021-08-02 23:29:22,446-INFO: pass 1, batch 60, accuracy:0.7421875 loss:0.9218395948410034 time:0.02 sec 2021-08-02 23:29:22,446 - [8518] [line:200] - INFO: pass 1, batch 60, accuracy:0.7421875 loss:0.9218395948410034 time:0.02 sec 2021-08-02 23:29:22,659-INFO: pass 1, batch 70, accuracy:0.7265625 loss:0.9175072312355042 time:0.02 sec 2021-08-02 23:29:22,659 - [8518] [line:200] - INFO: pass 1, batch 70, accuracy:0.7265625 loss:0.9175072312355042 time:0.02 sec 2021-08-02 23:29:22,773-INFO: pass 1 train result, current pass mean accuracy:0.7448676738299822 loss:0.94479905853146 2021-08-02 23:29:22,773 - [8518] [line:207] - INFO: pass 1 train result, current pass mean accuracy:0.7448676738299822 loss:0.94479905853146 2021-08-02 23:29:22,773-INFO: temp save pass 1 train result, current best pass mean accuracy:0.7448676738299822 2021-08-02 23:29:22,773 - [8518] [line:209] - INFO: temp save pass 1 train result, current best pass mean accuracy:0.7448676738299822 2021-08-02 23:29:23,395-INFO: current pass: 2, start read text 2021-08-02 23:29:23,395 - [8518] [line:179] - INFO: current pass: 2, start read text 2021-08-02 23:29:23,416-INFO: pass 2, batch 0, accuracy:0.734375 loss:0.9117376804351807 time:0.02 sec 2021-08-02 23:29:23,416 - [8518] [line:200] - INFO: pass 2, batch 0, accuracy:0.734375 loss:0.9117376804351807 time:0.02 sec 2021-08-02 23:29:23,627-INFO: pass 2, batch 10, accuracy:0.765625 loss:0.877050518989563 time:0.02 sec 2021-08-02 23:29:23,627 - [8518] [line:200] - INFO: pass 2, batch 10, accuracy:0.765625 loss:0.877050518989563 time:0.02 sec 2021-08-02 23:29:23,841-INFO: pass 2, batch 20, accuracy:0.6796875 loss:0.9222185611724854 time:0.02 sec 2021-08-02 23:29:23,841 - [8518] [line:200] - INFO: pass 2, batch 20, accuracy:0.6796875 loss:0.9222185611724854 time:0.02 sec 2021-08-02 23:29:24,055-INFO: pass 2, batch 30, accuracy:0.7578125 loss:0.8722472786903381 time:0.02 sec 2021-08-02 23:29:24,055 - [8518] [line:200] - INFO: pass 2, batch 30, accuracy:0.7578125 loss:0.8722472786903381 time:0.02 sec 2021-08-02 23:29:24,265-INFO: pass 2, batch 40, accuracy:0.8046875 loss:0.8403973579406738 time:0.02 sec 2021-08-02 23:29:24,265 - [8518] [line:200] - INFO: pass 2, batch 40, accuracy:0.8046875 loss:0.8403973579406738 time:0.02 sec 2021-08-02 23:29:24,477-INFO: pass 2, batch 50, accuracy:0.75 loss:0.8672002553939819 time:0.02 sec 2021-08-02 23:29:24,477 - [8518] [line:200] - INFO: pass 2, batch 50, accuracy:0.75 loss:0.8672002553939819 time:0.02 sec 2021-08-02 23:29:24,689-INFO: pass 2, batch 60, accuracy:0.796875 loss:0.8254350423812866 time:0.02 sec 2021-08-02 23:29:24,689 - [8518] [line:200] - INFO: pass 2, batch 60, accuracy:0.796875 loss:0.8254350423812866 time:0.02 sec 2021-08-02 23:29:24,904-INFO: pass 2, batch 70, accuracy:0.6875 loss:0.8920384049415588 time:0.02 sec 2021-08-02 23:29:24,904 - [8518] [line:200] - INFO: pass 2, batch 70, accuracy:0.6875 loss:0.8920384049415588 time:0.02 sec 2021-08-02 23:29:25,011-INFO: pass 2 train result, current pass mean accuracy:0.7445947967077556 loss:0.8793819091821972 2021-08-02 23:29:25,011 - [8518] [line:207] - INFO: pass 2 train result, current pass mean accuracy:0.7445947967077556 loss:0.8793819091821972 2021-08-02 23:29:25,011-INFO: current pass: 3, start read text 2021-08-02 23:29:25,011 - [8518] [line:179] - INFO: current pass: 3, start read text 2021-08-02 23:29:25,032-INFO: pass 3, batch 0, accuracy:0.734375 loss:0.8585522174835205 time:0.02 sec 2021-08-02 23:29:25,032 - [8518] [line:200] - INFO: pass 3, batch 0, accuracy:0.734375 loss:0.8585522174835205 time:0.02 sec 2021-08-02 23:29:25,251-INFO: pass 3, batch 10, accuracy:0.671875 loss:0.8995742797851562 time:0.02 sec 2021-08-02 23:29:25,251 - [8518] [line:200] - INFO: pass 3, batch 10, accuracy:0.671875 loss:0.8995742797851562 time:0.02 sec 2021-08-02 23:29:25,465-INFO: pass 3, batch 20, accuracy:0.734375 loss:0.8539764881134033 time:0.02 sec 2021-08-02 23:29:25,465 - [8518] [line:200] - INFO: pass 3, batch 20, accuracy:0.734375 loss:0.8539764881134033 time:0.02 sec 2021-08-02 23:29:25,678-INFO: pass 3, batch 30, accuracy:0.7421875 loss:0.8352509140968323 time:0.02 sec 2021-08-02 23:29:25,678 - [8518] [line:200] - INFO: pass 3, batch 30, accuracy:0.7421875 loss:0.8352509140968323 time:0.02 sec 2021-08-02 23:29:25,884-INFO: pass 3, batch 40, accuracy:0.7265625 loss:0.8421309590339661 time:0.02 sec 2021-08-02 23:29:25,884 - [8518] [line:200] - INFO: pass 3, batch 40, accuracy:0.7265625 loss:0.8421309590339661 time:0.02 sec 2021-08-02 23:29:26,094-INFO: pass 3, batch 50, accuracy:0.734375 loss:0.8350303769111633 time:0.02 sec 2021-08-02 23:29:26,094 - [8518] [line:200] - INFO: pass 3, batch 50, accuracy:0.734375 loss:0.8350303769111633 time:0.02 sec 2021-08-02 23:29:26,313-INFO: pass 3, batch 60, accuracy:0.703125 loss:0.8594977855682373 time:0.02 sec 2021-08-02 23:29:26,313 - [8518] [line:200] - INFO: pass 3, batch 60, accuracy:0.703125 loss:0.8594977855682373 time:0.02 sec 2021-08-02 23:29:26,527-INFO: pass 3, batch 70, accuracy:0.7265625 loss:0.8272154927253723 time:0.02 sec 2021-08-02 23:29:26,527 - [8518] [line:200] - INFO: pass 3, batch 70, accuracy:0.7265625 loss:0.8272154927253723 time:0.02 sec 2021-08-02 23:29:26,633-INFO: pass 3 train result, current pass mean accuracy:0.7440490432475743 loss:0.8344153157974544 2021-08-02 23:29:26,633 - [8518] [line:207] - INFO: pass 3 train result, current pass mean accuracy:0.7440490432475743 loss:0.8344153157974544 2021-08-02 23:29:26,633-INFO: current pass: 4, start read text 2021-08-02 23:29:26,633 - [8518] [line:179] - INFO: current pass: 4, start read text 2021-08-02 23:29:26,654-INFO: pass 4, batch 0, accuracy:0.7265625 loss:0.8458027839660645 time:0.02 sec 2021-08-02 23:29:26,654 - [8518] [line:200] - INFO: pass 4, batch 0, accuracy:0.7265625 loss:0.8458027839660645 time:0.02 sec 2021-08-02 23:29:26,874-INFO: pass 4, batch 10, accuracy:0.7578125 loss:0.7846702337265015 time:0.03 sec 2021-08-02 23:29:26,874 - [8518] [line:200] - INFO: pass 4, batch 10, accuracy:0.7578125 loss:0.7846702337265015 time:0.03 sec 2021-08-02 23:29:27,088-INFO: pass 4, batch 20, accuracy:0.7890625 loss:0.7644598484039307 time:0.02 sec 2021-08-02 23:29:27,088 - [8518] [line:200] - INFO: pass 4, batch 20, accuracy:0.7890625 loss:0.7644598484039307 time:0.02 sec 2021-08-02 23:29:27,300-INFO: pass 4, batch 30, accuracy:0.7578125 loss:0.8010584712028503 time:0.02 sec 2021-08-02 23:29:27,300 - [8518] [line:200] - INFO: pass 4, batch 30, accuracy:0.7578125 loss:0.8010584712028503 time:0.02 sec 2021-08-02 23:29:27,513-INFO: pass 4, batch 40, accuracy:0.7890625 loss:0.7610439658164978 time:0.02 sec 2021-08-02 23:29:27,513 - [8518] [line:200] - INFO: pass 4, batch 40, accuracy:0.7890625 loss:0.7610439658164978 time:0.02 sec 2021-08-02 23:29:27,727-INFO: pass 4, batch 50, accuracy:0.765625 loss:0.7683327794075012 time:0.02 sec 2021-08-02 23:29:27,727 - [8518] [line:200] - INFO: pass 4, batch 50, accuracy:0.765625 loss:0.7683327794075012 time:0.02 sec 2021-08-02 23:29:27,940-INFO: pass 4, batch 60, accuracy:0.7734375 loss:0.7692136168479919 time:0.02 sec 2021-08-02 23:29:27,940 - [8518] [line:200] - INFO: pass 4, batch 60, accuracy:0.7734375 loss:0.7692136168479919 time:0.02 sec 2021-08-02 23:29:28,159-INFO: pass 4, batch 70, accuracy:0.765625 loss:0.7770060300827026 time:0.02 sec 2021-08-02 23:29:28,159 - [8518] [line:200] - INFO: pass 4, batch 70, accuracy:0.765625 loss:0.7770060300827026 time:0.02 sec 2021-08-02 23:29:28,266-INFO: pass 4 train result, current pass mean accuracy:0.7445947967077556 loss:0.8022725950730475 2021-08-02 23:29:28,266 - [8518] [line:207] - INFO: pass 4 train result, current pass mean accuracy:0.7445947967077556 loss:0.8022725950730475 2021-08-02 23:29:28,266-INFO: current pass: 5, start read text 2021-08-02 23:29:28,266 - [8518] [line:179] - INFO: current pass: 5, start read text 2021-08-02 23:29:28,288-INFO: pass 5, batch 0, accuracy:0.7734375 loss:0.7529339790344238 time:0.02 sec 2021-08-02 23:29:28,288 - [8518] [line:200] - INFO: pass 5, batch 0, accuracy:0.7734375 loss:0.7529339790344238 time:0.02 sec 2021-08-02 23:29:28,500-INFO: pass 5, batch 10, accuracy:0.7265625 loss:0.8127733469009399 time:0.02 sec 2021-08-02 23:29:28,500 - [8518] [line:200] - INFO: pass 5, batch 10, accuracy:0.7265625 loss:0.8127733469009399 time:0.02 sec 2021-08-02 23:29:28,720-INFO: pass 5, batch 20, accuracy:0.7734375 loss:0.7375451326370239 time:0.02 sec 2021-08-02 23:29:28,720 - [8518] [line:200] - INFO: pass 5, batch 20, accuracy:0.7734375 loss:0.7375451326370239 time:0.02 sec 2021-08-02 23:29:28,934-INFO: pass 5, batch 30, accuracy:0.7421875 loss:0.7844938039779663 time:0.02 sec 2021-08-02 23:29:28,934 - [8518] [line:200] - INFO: pass 5, batch 30, accuracy:0.7421875 loss:0.7844938039779663 time:0.02 sec 2021-08-02 23:29:29,146-INFO: pass 5, batch 40, accuracy:0.796875 loss:0.7200489640235901 time:0.02 sec 2021-08-02 23:29:29,146 - [8518] [line:200] - INFO: pass 5, batch 40, accuracy:0.796875 loss:0.7200489640235901 time:0.02 sec 2021-08-02 23:29:29,366-INFO: pass 5, batch 50, accuracy:0.734375 loss:0.7850679755210876 time:0.02 sec 2021-08-02 23:29:29,366 - [8518] [line:200] - INFO: pass 5, batch 50, accuracy:0.734375 loss:0.7850679755210876 time:0.02 sec 2021-08-02 23:29:29,580-INFO: pass 5, batch 60, accuracy:0.734375 loss:0.7857416868209839 time:0.02 sec 2021-08-02 23:29:29,580 - [8518] [line:200] - INFO: pass 5, batch 60, accuracy:0.734375 loss:0.7857416868209839 time:0.02 sec 2021-08-02 23:29:29,792-INFO: pass 5, batch 70, accuracy:0.78125 loss:0.7242230772972107 time:0.02 sec 2021-08-02 23:29:29,792 - [8518] [line:200] - INFO: pass 5, batch 70, accuracy:0.78125 loss:0.7242230772972107 time:0.02 sec 2021-08-02 23:29:29,905-INFO: pass 5 train result, current pass mean accuracy:0.7441854814165517 loss:0.7799161029489416 2021-08-02 23:29:29,905 - [8518] [line:207] - INFO: pass 5 train result, current pass mean accuracy:0.7441854814165517 loss:0.7799161029489416 2021-08-02 23:29:29,905-INFO: current pass: 6, start read text 2021-08-02 23:29:29,905 - [8518] [line:179] - INFO: current pass: 6, start read text 2021-08-02 23:29:29,927-INFO: pass 6, batch 0, accuracy:0.7421875 loss:0.7838683128356934 time:0.02 sec 2021-08-02 23:29:29,927 - [8518] [line:200] - INFO: pass 6, batch 0, accuracy:0.7421875 loss:0.7838683128356934 time:0.02 sec 2021-08-02 23:29:30,139-INFO: pass 6, batch 10, accuracy:0.7265625 loss:0.7745072245597839 time:0.02 sec 2021-08-02 23:29:30,139 - [8518] [line:200] - INFO: pass 6, batch 10, accuracy:0.7265625 loss:0.7745072245597839 time:0.02 sec 2021-08-02 23:29:30,352-INFO: pass 6, batch 20, accuracy:0.7578125 loss:0.7590588331222534 time:0.02 sec 2021-08-02 23:29:30,352 - [8518] [line:200] - INFO: pass 6, batch 20, accuracy:0.7578125 loss:0.7590588331222534 time:0.02 sec 2021-08-02 23:29:30,572-INFO: pass 6, batch 30, accuracy:0.7578125 loss:0.7634695768356323 time:0.02 sec 2021-08-02 23:29:30,572 - [8518] [line:200] - INFO: pass 6, batch 30, accuracy:0.7578125 loss:0.7634695768356323 time:0.02 sec 2021-08-02 23:29:30,785-INFO: pass 6, batch 40, accuracy:0.6875 loss:0.8138336539268494 time:0.02 sec 2021-08-02 23:29:30,785 - [8518] [line:200] - INFO: pass 6, batch 40, accuracy:0.6875 loss:0.8138336539268494 time:0.02 sec 2021-08-02 23:29:31,001-INFO: pass 6, batch 50, accuracy:0.6875 loss:0.8206539750099182 time:0.03 sec 2021-08-02 23:29:31,001 - [8518] [line:200] - INFO: pass 6, batch 50, accuracy:0.6875 loss:0.8206539750099182 time:0.03 sec 2021-08-02 23:29:31,212-INFO: pass 6, batch 60, accuracy:0.7109375 loss:0.8077903985977173 time:0.02 sec 2021-08-02 23:29:31,212 - [8518] [line:200] - INFO: pass 6, batch 60, accuracy:0.7109375 loss:0.8077903985977173 time:0.02 sec 2021-08-02 23:29:31,421-INFO: pass 6, batch 70, accuracy:0.734375 loss:0.7780964374542236 time:0.02 sec 2021-08-02 23:29:31,421 - [8518] [line:200] - INFO: pass 6, batch 70, accuracy:0.734375 loss:0.7780964374542236 time:0.02 sec 2021-08-02 23:29:31,528-INFO: pass 6 train result, current pass mean accuracy:0.7444583585387782 loss:0.766914031223247 2021-08-02 23:29:31,528 - [8518] [line:207] - INFO: pass 6 train result, current pass mean accuracy:0.7444583585387782 loss:0.766914031223247 2021-08-02 23:29:31,528-INFO: current pass: 7, start read text 2021-08-02 23:29:31,528 - [8518] [line:179] - INFO: current pass: 7, start read text 2021-08-02 23:29:31,550-INFO: pass 7, batch 0, accuracy:0.7421875 loss:0.7668405771255493 time:0.02 sec 2021-08-02 23:29:31,550 - [8518] [line:200] - INFO: pass 7, batch 0, accuracy:0.7421875 loss:0.7668405771255493 time:0.02 sec 2021-08-02 23:29:31,770-INFO: pass 7, batch 10, accuracy:0.7421875 loss:0.7647318840026855 time:0.02 sec 2021-08-02 23:29:31,770 - [8518] [line:200] - INFO: pass 7, batch 10, accuracy:0.7421875 loss:0.7647318840026855 time:0.02 sec 2021-08-02 23:29:31,981-INFO: pass 7, batch 20, accuracy:0.8359375 loss:0.6733522415161133 time:0.02 sec 2021-08-02 23:29:31,981 - [8518] [line:200] - INFO: pass 7, batch 20, accuracy:0.8359375 loss:0.6733522415161133 time:0.02 sec 2021-08-02 23:29:32,203-INFO: pass 7, batch 30, accuracy:0.703125 loss:0.7983409762382507 time:0.02 sec 2021-08-02 23:29:32,203 - [8518] [line:200] - INFO: pass 7, batch 30, accuracy:0.703125 loss:0.7983409762382507 time:0.02 sec 2021-08-02 23:29:32,416-INFO: pass 7, batch 40, accuracy:0.8046875 loss:0.6874650716781616 time:0.02 sec 2021-08-02 23:29:32,416 - [8518] [line:200] - INFO: pass 7, batch 40, accuracy:0.8046875 loss:0.6874650716781616 time:0.02 sec 2021-08-02 23:29:32,628-INFO: pass 7, batch 50, accuracy:0.765625 loss:0.7372406721115112 time:0.02 sec 2021-08-02 23:29:32,628 - [8518] [line:200] - INFO: pass 7, batch 50, accuracy:0.765625 loss:0.7372406721115112 time:0.02 sec 2021-08-02 23:29:32,849-INFO: pass 7, batch 60, accuracy:0.6875 loss:0.8232687711715698 time:0.02 sec 2021-08-02 23:29:32,849 - [8518] [line:200] - INFO: pass 7, batch 60, accuracy:0.6875 loss:0.8232687711715698 time:0.02 sec 2021-08-02 23:29:33,062-INFO: pass 7, batch 70, accuracy:0.734375 loss:0.7785210013389587 time:0.02 sec 2021-08-02 23:29:33,062 - [8518] [line:200] - INFO: pass 7, batch 70, accuracy:0.734375 loss:0.7785210013389587 time:0.02 sec 2021-08-02 23:29:33,167-INFO: pass 7 train result, current pass mean accuracy:0.7445947967077556 loss:0.7595870102706709 2021-08-02 23:29:33,167 - [8518] [line:207] - INFO: pass 7 train result, current pass mean accuracy:0.7445947967077556 loss:0.7595870102706709 2021-08-02 23:29:33,168-INFO: current pass: 8, start read text 2021-08-02 23:29:33,168 - [8518] [line:179] - INFO: current pass: 8, start read text 2021-08-02 23:29:33,188-INFO: pass 8, batch 0, accuracy:0.703125 loss:0.8140039443969727 time:0.02 sec 2021-08-02 23:29:33,188 - [8518] [line:200] - INFO: pass 8, batch 0, accuracy:0.703125 loss:0.8140039443969727 time:0.02 sec 2021-08-02 23:29:33,410-INFO: pass 8, batch 10, accuracy:0.7265625 loss:0.7693229913711548 time:0.02 sec 2021-08-02 23:29:33,410 - [8518] [line:200] - INFO: pass 8, batch 10, accuracy:0.7265625 loss:0.7693229913711548 time:0.02 sec 2021-08-02 23:29:33,622-INFO: pass 8, batch 20, accuracy:0.734375 loss:0.7694175243377686 time:0.02 sec 2021-08-02 23:29:33,622 - [8518] [line:200] - INFO: pass 8, batch 20, accuracy:0.734375 loss:0.7694175243377686 time:0.02 sec 2021-08-02 23:29:33,835-INFO: pass 8, batch 30, accuracy:0.7109375 loss:0.8087602853775024 time:0.02 sec 2021-08-02 23:29:33,835 - [8518] [line:200] - INFO: pass 8, batch 30, accuracy:0.7109375 loss:0.8087602853775024 time:0.02 sec 2021-08-02 23:29:34,108-INFO: pass 8, batch 40, accuracy:0.7265625 loss:0.7804495096206665 time:0.02 sec 2021-08-02 23:29:34,108 - [8518] [line:200] - INFO: pass 8, batch 40, accuracy:0.7265625 loss:0.7804495096206665 time:0.02 sec 2021-08-02 23:29:34,322-INFO: pass 8, batch 50, accuracy:0.8125 loss:0.6808887720108032 time:0.02 sec 2021-08-02 23:29:34,322 - [8518] [line:200] - INFO: pass 8, batch 50, accuracy:0.8125 loss:0.6808887720108032 time:0.02 sec 2021-08-02 23:29:34,535-INFO: pass 8, batch 60, accuracy:0.8046875 loss:0.6881036758422852 time:0.02 sec 2021-08-02 23:29:34,535 - [8518] [line:200] - INFO: pass 8, batch 60, accuracy:0.8046875 loss:0.6881036758422852 time:0.02 sec 2021-08-02 23:29:34,754-INFO: pass 8, batch 70, accuracy:0.765625 loss:0.7288857102394104 time:0.02 sec 2021-08-02 23:29:34,754 - [8518] [line:200] - INFO: pass 8, batch 70, accuracy:0.765625 loss:0.7288857102394104 time:0.02 sec 2021-08-02 23:29:34,861-INFO: pass 8 train result, current pass mean accuracy:0.7443219195855292 loss:0.7553345608083826 2021-08-02 23:29:34,861 - [8518] [line:207] - INFO: pass 8 train result, current pass mean accuracy:0.7443219195855292 loss:0.7553345608083826 2021-08-02 23:29:34,861-INFO: current pass: 9, start read text 2021-08-02 23:29:34,861 - [8518] [line:179] - INFO: current pass: 9, start read text 2021-08-02 23:29:34,883-INFO: pass 9, batch 0, accuracy:0.75 loss:0.7523658871650696 time:0.02 sec 2021-08-02 23:29:34,883 - [8518] [line:200] - INFO: pass 9, batch 0, accuracy:0.75 loss:0.7523658871650696 time:0.02 sec 2021-08-02 23:29:35,094-INFO: pass 9, batch 10, accuracy:0.78125 loss:0.7139319777488708 time:0.02 sec 2021-08-02 23:29:35,094 - [8518] [line:200] - INFO: pass 9, batch 10, accuracy:0.78125 loss:0.7139319777488708 time:0.02 sec 2021-08-02 23:29:35,306-INFO: pass 9, batch 20, accuracy:0.71875 loss:0.7850377559661865 time:0.02 sec 2021-08-02 23:29:35,306 - [8518] [line:200] - INFO: pass 9, batch 20, accuracy:0.71875 loss:0.7850377559661865 time:0.02 sec 2021-08-02 23:29:35,514-INFO: pass 9, batch 30, accuracy:0.7109375 loss:0.7887594699859619 time:0.02 sec 2021-08-02 23:29:35,514 - [8518] [line:200] - INFO: pass 9, batch 30, accuracy:0.7109375 loss:0.7887594699859619 time:0.02 sec 2021-08-02 23:29:35,724-INFO: pass 9, batch 40, accuracy:0.7578125 loss:0.7470149397850037 time:0.02 sec 2021-08-02 23:29:35,724 - [8518] [line:200] - INFO: pass 9, batch 40, accuracy:0.7578125 loss:0.7470149397850037 time:0.02 sec 2021-08-02 23:29:35,938-INFO: pass 9, batch 50, accuracy:0.734375 loss:0.7686177492141724 time:0.02 sec 2021-08-02 23:29:35,938 - [8518] [line:200] - INFO: pass 9, batch 50, accuracy:0.734375 loss:0.7686177492141724 time:0.02 sec 2021-08-02 23:29:36,151-INFO: pass 9, batch 60, accuracy:0.671875 loss:0.8167787194252014 time:0.02 sec 2021-08-02 23:29:36,151 - [8518] [line:200] - INFO: pass 9, batch 60, accuracy:0.671875 loss:0.8167787194252014 time:0.02 sec 2021-08-02 23:29:36,363-INFO: pass 9, batch 70, accuracy:0.8125 loss:0.6641634106636047 time:0.02 sec 2021-08-02 23:29:36,363 - [8518] [line:200] - INFO: pass 9, batch 70, accuracy:0.8125 loss:0.6641634106636047 time:0.02 sec 2021-08-02 23:29:36,477-INFO: pass 9 train result, current pass mean accuracy:0.7443219195855292 loss:0.7519115727198752 2021-08-02 23:29:36,477 - [8518] [line:207] - INFO: pass 9 train result, current pass mean accuracy:0.7443219195855292 loss:0.7519115727198752 2021-08-02 23:29:36,477-INFO: current pass: 10, start read text 2021-08-02 23:29:36,477 - [8518] [line:179] - INFO: current pass: 10, start read text 2021-08-02 23:29:36,499-INFO: pass 10, batch 0, accuracy:0.78125 loss:0.713808000087738 time:0.02 sec 2021-08-02 23:29:36,499 - [8518] [line:200] - INFO: pass 10, batch 0, accuracy:0.78125 loss:0.713808000087738 time:0.02 sec 2021-08-02 23:29:36,710-INFO: pass 10, batch 10, accuracy:0.765625 loss:0.7260519862174988 time:0.02 sec 2021-08-02 23:29:36,710 - [8518] [line:200] - INFO: pass 10, batch 10, accuracy:0.765625 loss:0.7260519862174988 time:0.02 sec 2021-08-02 23:29:36,923-INFO: pass 10, batch 20, accuracy:0.7734375 loss:0.7242304086685181 time:0.02 sec 2021-08-02 23:29:36,923 - [8518] [line:200] - INFO: pass 10, batch 20, accuracy:0.7734375 loss:0.7242304086685181 time:0.02 sec 2021-08-02 23:29:37,137-INFO: pass 10, batch 30, accuracy:0.7109375 loss:0.781160831451416 time:0.02 sec 2021-08-02 23:29:37,137 - [8518] [line:200] - INFO: pass 10, batch 30, accuracy:0.7109375 loss:0.781160831451416 time:0.02 sec 2021-08-02 23:29:37,350-INFO: pass 10, batch 40, accuracy:0.78125 loss:0.6947404742240906 time:0.02 sec 2021-08-02 23:29:37,350 - [8518] [line:200] - INFO: pass 10, batch 40, accuracy:0.78125 loss:0.6947404742240906 time:0.02 sec 2021-08-02 23:29:37,569-INFO: pass 10, batch 50, accuracy:0.7578125 loss:0.7260289192199707 time:0.03 sec 2021-08-02 23:29:37,569 - [8518] [line:200] - INFO: pass 10, batch 50, accuracy:0.7578125 loss:0.7260289192199707 time:0.03 sec 2021-08-02 23:29:37,783-INFO: pass 10, batch 60, accuracy:0.765625 loss:0.7251702547073364 time:0.02 sec 2021-08-02 23:29:37,783 - [8518] [line:200] - INFO: pass 10, batch 60, accuracy:0.765625 loss:0.7251702547073364 time:0.02 sec 2021-08-02 23:29:37,995-INFO: pass 10, batch 70, accuracy:0.75 loss:0.7569742798805237 time:0.02 sec 2021-08-02 23:29:37,995 - [8518] [line:200] - INFO: pass 10, batch 70, accuracy:0.75 loss:0.7569742798805237 time:0.02 sec 2021-08-02 23:29:38,101-INFO: pass 10 train result, current pass mean accuracy:0.7448676738299822 loss:0.7496904409245441 2021-08-02 23:29:38,101 - [8518] [line:207] - INFO: pass 10 train result, current pass mean accuracy:0.7448676738299822 loss:0.7496904409245441 2021-08-02 23:29:38,101-INFO: current pass: 11, start read text 2021-08-02 23:29:38,101 - [8518] [line:179] - INFO: current pass: 11, start read text 2021-08-02 23:29:38,122-INFO: pass 11, batch 0, accuracy:0.71875 loss:0.7894006967544556 time:0.02 sec 2021-08-02 23:29:38,122 - [8518] [line:200] - INFO: pass 11, batch 0, accuracy:0.71875 loss:0.7894006967544556 time:0.02 sec 2021-08-02 23:29:38,343-INFO: pass 11, batch 10, accuracy:0.75 loss:0.7604379057884216 time:0.02 sec 2021-08-02 23:29:38,343 - [8518] [line:200] - INFO: pass 11, batch 10, accuracy:0.75 loss:0.7604379057884216 time:0.02 sec 2021-08-02 23:29:38,554-INFO: pass 11, batch 20, accuracy:0.7265625 loss:0.7730419039726257 time:0.02 sec 2021-08-02 23:29:38,554 - [8518] [line:200] - INFO: pass 11, batch 20, accuracy:0.7265625 loss:0.7730419039726257 time:0.02 sec 2021-08-02 23:29:38,774-INFO: pass 11, batch 30, accuracy:0.7265625 loss:0.7723853588104248 time:0.02 sec 2021-08-02 23:29:38,774 - [8518] [line:200] - INFO: pass 11, batch 30, accuracy:0.7265625 loss:0.7723853588104248 time:0.02 sec 2021-08-02 23:29:38,988-INFO: pass 11, batch 40, accuracy:0.875 loss:0.594655454158783 time:0.02 sec 2021-08-02 23:29:38,988 - [8518] [line:200] - INFO: pass 11, batch 40, accuracy:0.875 loss:0.594655454158783 time:0.02 sec 2021-08-02 23:29:39,199-INFO: pass 11, batch 50, accuracy:0.8203125 loss:0.6593010425567627 time:0.02 sec 2021-08-02 23:29:39,199 - [8518] [line:200] - INFO: pass 11, batch 50, accuracy:0.8203125 loss:0.6593010425567627 time:0.02 sec 2021-08-02 23:29:39,419-INFO: pass 11, batch 60, accuracy:0.75 loss:0.7294450998306274 time:0.02 sec 2021-08-02 23:29:39,419 - [8518] [line:200] - INFO: pass 11, batch 60, accuracy:0.75 loss:0.7294450998306274 time:0.02 sec 2021-08-02 23:29:39,633-INFO: pass 11, batch 70, accuracy:0.7109375 loss:0.7864516973495483 time:0.02 sec 2021-08-02 23:29:39,633 - [8518] [line:200] - INFO: pass 11, batch 70, accuracy:0.7109375 loss:0.7864516973495483 time:0.02 sec 2021-08-02 23:29:39,739-INFO: pass 11 train result, current pass mean accuracy:0.7437761661253477 loss:0.7498400564256468 2021-08-02 23:29:39,739 - [8518] [line:207] - INFO: pass 11 train result, current pass mean accuracy:0.7437761661253477 loss:0.7498400564256468 2021-08-02 23:29:39,739-INFO: training till last epcho, end training 2021-08-02 23:29:39,739 - [8518] [line:214] - INFO: training till last epcho, end training

3.3.2.2 固化训练参数

In [4]

!python freeze.py

2021-08-02 23:29:42,403-INFO: build network bow_net to freeze 2021-08-02 23:29:42,403 - [8613] [line:46] - INFO: build network bow_net to freeze 2021-08-02 23:29:43,766-INFO: freeze bow_net end 2021-08-02 23:29:43,766 - [8613] [line:66] - INFO: freeze bow_net end

3.3.2.3 验证训练结果

In [5]

!python eval.py

W0802 23:29:45.443478 8679 device_context.cc:252] Please NOTE: device: 0, CUDA Capability: 70, Driver API Version: 10.1, Runtime API Version: 9.0 W0802 23:29:45.448276 8679 device_context.cc:260] device: 0, cuDNN Version: 7.6. total eval count:1036 cost time:7.02 sec predict accuracy:0.750965250965251

3.3.2.4 使用模型进行预测

In [6]

!python infer.py

W0802 23:30:08.974885 8784 device_context.cc:252] Please NOTE: device: 0, CUDA Capability: 70, Driver API Version: 10.1, Runtime API Version: 9.0 W0802 23:30:08.979648 8784 device_context.cc:260] device: 0, cuDNN Version: 7.6. infer text:我 就是 喜欢 你的 知性 大方 温柔 infer result:1

3.3.3 CNN模型的训练与预测

3.3.3.1 训练模型

更改train_parameters中"model_type"为"cnn_net"

In [7]

!python train.py

2021-08-02 23:32:29,776-INFO: start train text classification, train params: {'class_dim': 3, 'sample_count': 9656, 'label_dict': {'消极': 0, '中性': 1, '积极': 2}, 'data_dir': 'data/data9740', 'vocabulary': 'vocab.txt', 'train_file_list': 'train.txt', 'label_file': 'label_list.txt', 'save_model_dir': './train_model', 'save_freeze_dir': './infer_model', 'pretrained_model_dir': './pretrained_model', 'continue_train': False, 'pretrained': False, 'model_type': 'cnn_net', 'num_epochs': 12, 'train_batch_size': 128, 'use_gpu': True, 'sample_frequency': 10, 'rsm_strategy': {'learning_rate': 0.001, 'lr_epochs': [6, 8, 10], 'lr_decay': [1, 0.5, 0.25, 0.1]}, 'momentum_strategy': {'learning_rate': 0.001, 'lr_epochs': [6, 8, 10], 'lr_decay': [1, 0.5, 0.25, 0.1]}, 'sgd_strategy': {'learning_rate': 0.001, 'lr_epochs': [6, 8, 10], 'lr_decay': [1, 0.5, 0.25, 0.1]}, 'adam_strategy': {'learning_rate': 0.001}} 2021-08-02 23:32:29,776 - [8970] [line:133] - INFO: start train text classification, train params: {'class_dim': 3, 'sample_count': 9656, 'label_dict': {'消极': 0, '中性': 1, '积极': 2}, 'data_dir': 'data/data9740', 'vocabulary': 'vocab.txt', 'train_file_list': 'train.txt', 'label_file': 'label_list.txt', 'save_model_dir': './train_model', 'save_freeze_dir': './infer_model', 'pretrained_model_dir': './pretrained_model', 'continue_train': False, 'pretrained': False, 'model_type': 'cnn_net', 'num_epochs': 12, 'train_batch_size': 128, 'use_gpu': True, 'sample_frequency': 10, 'rsm_strategy': {'learning_rate': 0.001, 'lr_epochs': [6, 8, 10], 'lr_decay': [1, 0.5, 0.25, 0.1]}, 'momentum_strategy': {'learning_rate': 0.001, 'lr_epochs': [6, 8, 10], 'lr_decay': [1, 0.5, 0.25, 0.1]}, 'sgd_strategy': {'learning_rate': 0.001, 'lr_epochs': [6, 8, 10], 'lr_decay': [1, 0.5, 0.25, 0.1]}, 'adam_strategy': {'learning_rate': 0.001}} 2021-08-02 23:32:29,776-INFO: create place, use gpu: True 2021-08-02 23:32:29,776 - [8970] [line:134] - INFO: create place, use gpu: True 2021-08-02 23:32:30,208-INFO: define input data tensor 2021-08-02 23:32:30,208 - [8970] [line:139] - INFO: define input data tensor 2021-08-02 23:32:30,209-INFO: build custom reader 2021-08-02 23:32:30,209 - [8970] [line:144] - INFO: build custom reader 2021-08-02 23:32:30,235-INFO: build network cnn_net 2021-08-02 23:32:30,235 - [8970] [line:147] - INFO: build network cnn_net 2021-08-02 23:32:30,248-INFO: build optimizer 2021-08-02 23:32:30,248 - [8970] [line:164] - INFO: build optimizer W0802 23:32:30.265537 8970 device_context.cc:252] Please NOTE: device: 0, CUDA Capability: 70, Driver API Version: 10.1, Runtime API Version: 9.0 W0802 23:32:30.270296 8970 device_context.cc:260] device: 0, cuDNN Version: 7.6. 2021-08-02 23:32:44,077-INFO: current pass: 0, start read text 2021-08-02 23:32:44,077 - [8970] [line:179] - INFO: current pass: 0, start read text 2021-08-02 23:32:44,119-INFO: pass 0, batch 0, accuracy:0.09375 loss:1.1013975143432617 time:0.04 sec 2021-08-02 23:32:44,119 - [8970] [line:200] - INFO: pass 0, batch 0, accuracy:0.09375 loss:1.1013975143432617 time:0.04 sec 2021-08-02 23:32:44,342-INFO: pass 0, batch 10, accuracy:0.734375 loss:1.0878925323486328 time:0.02 sec 2021-08-02 23:32:44,342 - [8970] [line:200] - INFO: pass 0, batch 10, accuracy:0.734375 loss:1.0878925323486328 time:0.02 sec 2021-08-02 23:32:44,571-INFO: pass 0, batch 20, accuracy:0.78125 loss:1.0729808807373047 time:0.02 sec 2021-08-02 23:32:44,571 - [8970] [line:200] - INFO: pass 0, batch 20, accuracy:0.78125 loss:1.0729808807373047 time:0.02 sec 2021-08-02 23:32:44,796-INFO: pass 0, batch 30, accuracy:0.7265625 loss:1.064940094947815 time:0.02 sec 2021-08-02 23:32:44,796 - [8970] [line:200] - INFO: pass 0, batch 30, accuracy:0.7265625 loss:1.064940094947815 time:0.02 sec 2021-08-02 23:32:45,019-INFO: pass 0, batch 40, accuracy:0.765625 loss:1.048965334892273 time:0.02 sec 2021-08-02 23:32:45,019 - [8970] [line:200] - INFO: pass 0, batch 40, accuracy:0.765625 loss:1.048965334892273 time:0.02 sec 2021-08-02 23:32:45,249-INFO: pass 0, batch 50, accuracy:0.75 loss:1.0398499965667725 time:0.02 sec 2021-08-02 23:32:45,249 - [8970] [line:200] - INFO: pass 0, batch 50, accuracy:0.75 loss:1.0398499965667725 time:0.02 sec 2021-08-02 23:32:45,473-INFO: pass 0, batch 60, accuracy:0.7265625 loss:1.0328619480133057 time:0.02 sec 2021-08-02 23:32:45,473 - [8970] [line:200] - INFO: pass 0, batch 60, accuracy:0.7265625 loss:1.0328619480133057 time:0.02 sec 2021-08-02 23:32:45,697-INFO: pass 0, batch 70, accuracy:0.6796875 loss:1.0320944786071777 time:0.02 sec 2021-08-02 23:32:45,697 - [8970] [line:200] - INFO: pass 0, batch 70, accuracy:0.6796875 loss:1.0320944786071777 time:0.02 sec 2021-08-02 23:32:45,798-INFO: pass 0 train result, current pass mean accuracy:0.7156735948826137 loss:1.0554037188228809 2021-08-02 23:32:45,798 - [8970] [line:207] - INFO: pass 0 train result, current pass mean accuracy:0.7156735948826137 loss:1.0554037188228809 2021-08-02 23:32:45,799-INFO: temp save pass 0 train result, current best pass mean accuracy:0.7156735948826137 2021-08-02 23:32:45,799 - [8970] [line:209] - INFO: temp save pass 0 train result, current best pass mean accuracy:0.7156735948826137 2021-08-02 23:32:46,412-INFO: current pass: 1, start read text 2021-08-02 23:32:46,412 - [8970] [line:179] - INFO: current pass: 1, start read text 2021-08-02 23:32:46,434-INFO: pass 1, batch 0, accuracy:0.8046875 loss:0.9993318319320679 time:0.02 sec 2021-08-02 23:32:46,434 - [8970] [line:200] - INFO: pass 1, batch 0, accuracy:0.8046875 loss:0.9993318319320679 time:0.02 sec 2021-08-02 23:32:46,656-INFO: pass 1, batch 10, accuracy:0.7890625 loss:0.9915479421615601 time:0.02 sec 2021-08-02 23:32:46,656 - [8970] [line:200] - INFO: pass 1, batch 10, accuracy:0.7890625 loss:0.9915479421615601 time:0.02 sec 2021-08-02 23:32:46,877-INFO: pass 1, batch 20, accuracy:0.78125 loss:0.9827027916908264 time:0.02 sec 2021-08-02 23:32:46,877 - [8970] [line:200] - INFO: pass 1, batch 20, accuracy:0.78125 loss:0.9827027916908264 time:0.02 sec 2021-08-02 23:32:47,101-INFO: pass 1, batch 30, accuracy:0.765625 loss:0.9777587652206421 time:0.02 sec 2021-08-02 23:32:47,101 - [8970] [line:200] - INFO: pass 1, batch 30, accuracy:0.765625 loss:0.9777587652206421 time:0.02 sec 2021-08-02 23:32:47,326-INFO: pass 1, batch 40, accuracy:0.8046875 loss:0.954832136631012 time:0.02 sec 2021-08-02 23:32:47,326 - [8970] [line:200] - INFO: pass 1, batch 40, accuracy:0.8046875 loss:0.954832136631012 time:0.02 sec 2021-08-02 23:32:47,539-INFO: pass 1, batch 50, accuracy:0.7421875 loss:0.9663257002830505 time:0.02 sec 2021-08-02 23:32:47,539 - [8970] [line:200] - INFO: pass 1, batch 50, accuracy:0.7421875 loss:0.9663257002830505 time:0.02 sec 2021-08-02 23:32:47,771-INFO: pass 1, batch 60, accuracy:0.78125 loss:0.9437062740325928 time:0.02 sec 2021-08-02 23:32:47,771 - [8970] [line:200] - INFO: pass 1, batch 60, accuracy:0.78125 loss:0.9437062740325928 time:0.02 sec 2021-08-02 23:32:47,994-INFO: pass 1, batch 70, accuracy:0.7109375 loss:0.9628456830978394 time:0.02 sec 2021-08-02 23:32:47,994 - [8970] [line:200] - INFO: pass 1, batch 70, accuracy:0.7109375 loss:0.9628456830978394 time:0.02 sec 2021-08-02 23:32:48,104-INFO: pass 1 train result, current pass mean accuracy:0.7440490432475743 loss:0.9780148252060539 2021-08-02 23:32:48,104 - [8970] [line:207] - INFO: pass 1 train result, current pass mean accuracy:0.7440490432475743 loss:0.9780148252060539 2021-08-02 23:32:48,104-INFO: temp save pass 1 train result, current best pass mean accuracy:0.7440490432475743 2021-08-02 23:32:48,104 - [8970] [line:209] - INFO: temp save pass 1 train result, current best pass mean accuracy:0.7440490432475743 2021-08-02 23:32:48,717-INFO: current pass: 2, start read text 2021-08-02 23:32:48,717 - [8970] [line:179] - INFO: current pass: 2, start read text 2021-08-02 23:32:48,740-INFO: pass 2, batch 0, accuracy:0.765625 loss:0.93613600730896 time:0.02 sec 2021-08-02 23:32:48,740 - [8970] [line:200] - INFO: pass 2, batch 0, accuracy:0.765625 loss:0.93613600730896 time:0.02 sec 2021-08-02 23:32:48,972-INFO: pass 2, batch 10, accuracy:0.71875 loss:0.9484882354736328 time:0.02 sec 2021-08-02 23:32:48,972 - [8970] [line:200] - INFO: pass 2, batch 10, accuracy:0.71875 loss:0.9484882354736328 time:0.02 sec 2021-08-02 23:32:49,194-INFO: pass 2, batch 20, accuracy:0.7890625 loss:0.907993495464325 time:0.02 sec 2021-08-02 23:32:49,194 - [8970] [line:200] - INFO: pass 2, batch 20, accuracy:0.7890625 loss:0.907993495464325 time:0.02 sec 2021-08-02 23:32:49,412-INFO: pass 2, batch 30, accuracy:0.78125 loss:0.9043691158294678 time:0.02 sec 2021-08-02 23:32:49,412 - [8970] [line:200] - INFO: pass 2, batch 30, accuracy:0.78125 loss:0.9043691158294678 time:0.02 sec 2021-08-02 23:32:49,638-INFO: pass 2, batch 40, accuracy:0.7421875 loss:0.9172401428222656 time:0.02 sec 2021-08-02 23:32:49,638 - [8970] [line:200] - INFO: pass 2, batch 40, accuracy:0.7421875 loss:0.9172401428222656 time:0.02 sec 2021-08-02 23:32:49,862-INFO: pass 2, batch 50, accuracy:0.78125 loss:0.8896579146385193 time:0.02 sec 2021-08-02 23:32:49,862 - [8970] [line:200] - INFO: pass 2, batch 50, accuracy:0.78125 loss:0.8896579146385193 time:0.02 sec 2021-08-02 23:32:50,087-INFO: pass 2, batch 60, accuracy:0.7421875 loss:0.9037520289421082 time:0.02 sec 2021-08-02 23:32:50,087 - [8970] [line:200] - INFO: pass 2, batch 60, accuracy:0.7421875 loss:0.9037520289421082 time:0.02 sec 2021-08-02 23:32:50,315-INFO: pass 2, batch 70, accuracy:0.734375 loss:0.9033712148666382 time:0.02 sec 2021-08-02 23:32:50,315 - [8970] [line:200] - INFO: pass 2, batch 70, accuracy:0.734375 loss:0.9033712148666382 time:0.02 sec 2021-08-02 23:32:50,427-INFO: pass 2 train result, current pass mean accuracy:0.7443219195855292 loss:0.9186339794020904 2021-08-02 23:32:50,427 - [8970] [line:207] - INFO: pass 2 train result, current pass mean accuracy:0.7443219195855292 loss:0.9186339794020904 2021-08-02 23:32:50,427-INFO: temp save pass 2 train result, current best pass mean accuracy:0.7443219195855292 2021-08-02 23:32:50,427 - [8970] [line:209] - INFO: temp save pass 2 train result, current best pass mean accuracy:0.7443219195855292 2021-08-02 23:32:51,034-INFO: current pass: 3, start read text 2021-08-02 23:32:51,034 - [8970] [line:179] - INFO: current pass: 3, start read text 2021-08-02 23:32:51,058-INFO: pass 3, batch 0, accuracy:0.796875 loss:0.8632824420928955 time:0.02 sec 2021-08-02 23:32:51,058 - [8970] [line:200] - INFO: pass 3, batch 0, accuracy:0.796875 loss:0.8632824420928955 time:0.02 sec 2021-08-02 23:32:51,270-INFO: pass 3, batch 10, accuracy:0.7421875 loss:0.8891596794128418 time:0.02 sec 2021-08-02 23:32:51,270 - [8970] [line:200] - INFO: pass 3, batch 10, accuracy:0.7421875 loss:0.8891596794128418 time:0.02 sec 2021-08-02 23:32:51,502-INFO: pass 3, batch 20, accuracy:0.734375 loss:0.88825523853302 time:0.02 sec 2021-08-02 23:32:51,502 - [8970] [line:200] - INFO: pass 3, batch 20, accuracy:0.734375 loss:0.88825523853302 time:0.02 sec 2021-08-02 23:32:51,725-INFO: pass 3, batch 30, accuracy:0.765625 loss:0.8633568286895752 time:0.02 sec 2021-08-02 23:32:51,725 - [8970] [line:200] - INFO: pass 3, batch 30, accuracy:0.765625 loss:0.8633568286895752 time:0.02 sec 2021-08-02 23:32:51,948-INFO: pass 3, batch 40, accuracy:0.78125 loss:0.8450338840484619 time:0.02 sec 2021-08-02 23:32:51,948 - [8970] [line:200] - INFO: pass 3, batch 40, accuracy:0.78125 loss:0.8450338840484619 time:0.02 sec 2021-08-02 23:32:52,180-INFO: pass 3, batch 50, accuracy:0.8046875 loss:0.824402391910553 time:0.02 sec 2021-08-02 23:32:52,180 - [8970] [line:200] - INFO: pass 3, batch 50, accuracy:0.8046875 loss:0.824402391910553 time:0.02 sec 2021-08-02 23:32:52,396-INFO: pass 3, batch 60, accuracy:0.6484375 loss:0.9306120872497559 time:0.02 sec 2021-08-02 23:32:52,396 - [8970] [line:200] - INFO: pass 3, batch 60, accuracy:0.6484375 loss:0.9306120872497559 time:0.02 sec 2021-08-02 23:32:52,617-INFO: pass 3, batch 70, accuracy:0.6875 loss:0.8973121643066406 time:0.02 sec 2021-08-02 23:32:52,617 - [8970] [line:200] - INFO: pass 3, batch 70, accuracy:0.6875 loss:0.8973121643066406 time:0.02 sec 2021-08-02 23:32:52,735-INFO: pass 3 train result, current pass mean accuracy:0.7452769883369145 loss:0.8727092397840399 2021-08-02 23:32:52,735 - [8970] [line:207] - INFO: pass 3 train result, current pass mean accuracy:0.7452769883369145 loss:0.8727092397840399 2021-08-02 23:32:52,735-INFO: temp save pass 3 train result, current best pass mean accuracy:0.7452769883369145 2021-08-02 23:32:52,735 - [8970] [line:209] - INFO: temp save pass 3 train result, current best pass mean accuracy:0.7452769883369145 2021-08-02 23:32:53,367-INFO: current pass: 4, start read text 2021-08-02 23:32:53,367 - [8970] [line:179] - INFO: current pass: 4, start read text 2021-08-02 23:32:53,390-INFO: pass 4, batch 0, accuracy:0.78125 loss:0.826328694820404 time:0.02 sec 2021-08-02 23:32:53,390 - [8970] [line:200] - INFO: pass 4, batch 0, accuracy:0.78125 loss:0.826328694820404 time:0.02 sec 2021-08-02 23:32:53,615-INFO: pass 4, batch 10, accuracy:0.71875 loss:0.8700000047683716 time:0.02 sec 2021-08-02 23:32:53,615 - [8970] [line:200] - INFO: pass 4, batch 10, accuracy:0.71875 loss:0.8700000047683716 time:0.02 sec 2021-08-02 23:32:53,838-INFO: pass 4, batch 20, accuracy:0.703125 loss:0.8794984817504883 time:0.02 sec 2021-08-02 23:32:53,838 - [8970] [line:200] - INFO: pass 4, batch 20, accuracy:0.703125 loss:0.8794984817504883 time:0.02 sec 2021-08-02 23:32:54,068-INFO: pass 4, batch 30, accuracy:0.703125 loss:0.8750355839729309 time:0.02 sec 2021-08-02 23:32:54,068 - [8970] [line:200] - INFO: pass 4, batch 30, accuracy:0.703125 loss:0.8750355839729309 time:0.02 sec 2021-08-02 23:32:54,281-INFO: pass 4, batch 40, accuracy:0.7578125 loss:0.8268516063690186 time:0.02 sec 2021-08-02 23:32:54,281 - [8970] [line:200] - INFO: pass 4, batch 40, accuracy:0.7578125 loss:0.8268516063690186 time:0.02 sec 2021-08-02 23:32:54,505-INFO: pass 4, batch 50, accuracy:0.71875 loss:0.8547161221504211 time:0.02 sec 2021-08-02 23:32:54,505 - [8970] [line:200] - INFO: pass 4, batch 50, accuracy:0.71875 loss:0.8547161221504211 time:0.02 sec 2021-08-02 23:32:54,730-INFO: pass 4, batch 60, accuracy:0.6796875 loss:0.88545823097229 time:0.02 sec 2021-08-02 23:32:54,730 - [8970] [line:200] - INFO: pass 4, batch 60, accuracy:0.6796875 loss:0.88545823097229 time:0.02 sec 2021-08-02 23:32:54,953-INFO: pass 4, batch 70, accuracy:0.734375 loss:0.8349552750587463 time:0.02 sec 2021-08-02 23:32:54,953 - [8970] [line:200] - INFO: pass 4, batch 70, accuracy:0.734375 loss:0.8349552750587463 time:0.02 sec 2021-08-02 23:32:55,062-INFO: pass 4 train result, current pass mean accuracy:0.7448676738299822 loss:0.8384564895378915 2021-08-02 23:32:55,062 - [8970] [line:207] - INFO: pass 4 train result, current pass mean accuracy:0.7448676738299822 loss:0.8384564895378915 2021-08-02 23:32:55,062-INFO: current pass: 5, start read text 2021-08-02 23:32:55,062 - [8970] [line:179] - INFO: current pass: 5, start read text 2021-08-02 23:32:55,085-INFO: pass 5, batch 0, accuracy:0.7890625 loss:0.7858906984329224 time:0.02 sec 2021-08-02 23:32:55,085 - [8970] [line:200] - INFO: pass 5, batch 0, accuracy:0.7890625 loss:0.7858906984329224 time:0.02 sec 2021-08-02 23:32:55,317-INFO: pass 5, batch 10, accuracy:0.75 loss:0.8153935670852661 time:0.02 sec 2021-08-02 23:32:55,317 - [8970] [line:200] - INFO: pass 5, batch 10, accuracy:0.75 loss:0.8153935670852661 time:0.02 sec 2021-08-02 23:32:55,537-INFO: pass 5, batch 20, accuracy:0.6796875 loss:0.8773574829101562 time:0.02 sec 2021-08-02 23:32:55,537 - [8970] [line:200] - INFO: pass 5, batch 20, accuracy:0.6796875 loss:0.8773574829101562 time:0.02 sec 2021-08-02 23:32:55,760-INFO: pass 5, batch 30, accuracy:0.75 loss:0.8097986578941345 time:0.02 sec 2021-08-02 23:32:55,760 - [8970] [line:200] - INFO: pass 5, batch 30, accuracy:0.75 loss:0.8097986578941345 time:0.02 sec 2021-08-02 23:32:55,992-INFO: pass 5, batch 40, accuracy:0.78125 loss:0.7779437303543091 time:0.02 sec 2021-08-02 23:32:55,992 - [8970] [line:200] - INFO: pass 5, batch 40, accuracy:0.78125 loss:0.7779437303543091 time:0.02 sec 2021-08-02 23:32:56,214-INFO: pass 5, batch 50, accuracy:0.7421875 loss:0.8110630512237549 time:0.02 sec 2021-08-02 23:32:56,214 - [8970] [line:200] - INFO: pass 5, batch 50, accuracy:0.7421875 loss:0.8110630512237549 time:0.02 sec 2021-08-02 23:32:56,438-INFO: pass 5, batch 60, accuracy:0.71875 loss:0.8321855068206787 time:0.02 sec 2021-08-02 23:32:56,438 - [8970] [line:200] - INFO: pass 5, batch 60, accuracy:0.71875 loss:0.8321855068206787 time:0.02 sec 2021-08-02 23:32:56,660-INFO: pass 5, batch 70, accuracy:0.8359375 loss:0.7149022817611694 time:0.02 sec 2021-08-02 23:32:56,660 - [8970] [line:200] - INFO: pass 5, batch 70, accuracy:0.8359375 loss:0.7149022817611694 time:0.02 sec 2021-08-02 23:32:56,772-INFO: pass 5 train result, current pass mean accuracy:0.7441854814165517 loss:0.8130515500118858 2021-08-02 23:32:56,772 - [8970] [line:207] - INFO: pass 5 train result, current pass mean accuracy:0.7441854814165517 loss:0.8130515500118858 2021-08-02 23:32:56,772-INFO: current pass: 6, start read text 2021-08-02 23:32:56,772 - [8970] [line:179] - INFO: current pass: 6, start read text 2021-08-02 23:32:56,795-INFO: pass 6, batch 0, accuracy:0.75 loss:0.7998899221420288 time:0.02 sec 2021-08-02 23:32:56,795 - [8970] [line:200] - INFO: pass 6, batch 0, accuracy:0.75 loss:0.7998899221420288 time:0.02 sec 2021-08-02 23:32:57,025-INFO: pass 6, batch 10, accuracy:0.6953125 loss:0.8514322638511658 time:0.02 sec 2021-08-02 23:32:57,025 - [8970] [line:200] - INFO: pass 6, batch 10, accuracy:0.6953125 loss:0.8514322638511658 time:0.02 sec 2021-08-02 23:32:57,252-INFO: pass 6, batch 20, accuracy:0.7421875 loss:0.7990052700042725 time:0.02 sec 2021-08-02 23:32:57,252 - [8970] [line:200] - INFO: pass 6, batch 20, accuracy:0.7421875 loss:0.7990052700042725 time:0.02 sec 2021-08-02 23:32:57,470-INFO: pass 6, batch 30, accuracy:0.671875 loss:0.8762463331222534 time:0.02 sec 2021-08-02 23:32:57,470 - [8970] [line:200] - INFO: pass 6, batch 30, accuracy:0.671875 loss:0.8762463331222534 time:0.02 sec 2021-08-02 23:32:57,693-INFO: pass 6, batch 40, accuracy:0.7109375 loss:0.8294878005981445 time:0.02 sec 2021-08-02 23:32:57,693 - [8970] [line:200] - INFO: pass 6, batch 40, accuracy:0.7109375 loss:0.8294878005981445 time:0.02 sec 2021-08-02 23:32:57,924-INFO: pass 6, batch 50, accuracy:0.671875 loss:0.8709583282470703 time:0.02 sec 2021-08-02 23:32:57,924 - [8970] [line:200] - INFO: pass 6, batch 50, accuracy:0.671875 loss:0.8709583282470703 time:0.02 sec 2021-08-02 23:32:58,148-INFO: pass 6, batch 60, accuracy:0.78125 loss:0.7583836317062378 time:0.02 sec 2021-08-02 23:32:58,148 - [8970] [line:200] - INFO: pass 6, batch 60, accuracy:0.78125 loss:0.7583836317062378 time:0.02 sec 2021-08-02 23:32:58,370-INFO: pass 6, batch 70, accuracy:0.703125 loss:0.8358422517776489 time:0.02 sec 2021-08-02 23:32:58,370 - [8970] [line:200] - INFO: pass 6, batch 70, accuracy:0.703125 loss:0.8358422517776489 time:0.02 sec 2021-08-02 23:32:58,482-INFO: pass 6 train result, current pass mean accuracy:0.7443219195855292 loss:0.7982935097656751 2021-08-02 23:32:58,482 - [8970] [line:207] - INFO: pass 6 train result, current pass mean accuracy:0.7443219195855292 loss:0.7982935097656751 2021-08-02 23:32:58,482-INFO: current pass: 7, start read text 2021-08-02 23:32:58,482 - [8970] [line:179] - INFO: current pass: 7, start read text 2021-08-02 23:32:58,505-INFO: pass 7, batch 0, accuracy:0.7421875 loss:0.7975772619247437 time:0.02 sec 2021-08-02 23:32:58,505 - [8970] [line:200] - INFO: pass 7, batch 0, accuracy:0.7421875 loss:0.7975772619247437 time:0.02 sec 2021-08-02 23:32:58,728-INFO: pass 7, batch 10, accuracy:0.828125 loss:0.7014524936676025 time:0.02 sec 2021-08-02 23:32:58,728 - [8970] [line:200] - INFO: pass 7, batch 10, accuracy:0.828125 loss:0.7014524936676025 time:0.02 sec 2021-08-02 23:32:58,950-INFO: pass 7, batch 20, accuracy:0.7890625 loss:0.7436310052871704 time:0.02 sec 2021-08-02 23:32:58,950 - [8970] [line:200] - INFO: pass 7, batch 20, accuracy:0.7890625 loss:0.7436310052871704 time:0.02 sec 2021-08-02 23:32:59,172-INFO: pass 7, batch 30, accuracy:0.7578125 loss:0.7746716737747192 time:0.02 sec 2021-08-02 23:32:59,172 - [8970] [line:200] - INFO: pass 7, batch 30, accuracy:0.7578125 loss:0.7746716737747192 time:0.02 sec 2021-08-02 23:32:59,395-INFO: pass 7, batch 40, accuracy:0.7578125 loss:0.7764407396316528 time:0.02 sec 2021-08-02 23:32:59,395 - [8970] [line:200] - INFO: pass 7, batch 40, accuracy:0.7578125 loss:0.7764407396316528 time:0.02 sec 2021-08-02 23:32:59,620-INFO: pass 7, batch 50, accuracy:0.75 loss:0.7831007242202759 time:0.02 sec 2021-08-02 23:32:59,620 - [8970] [line:200] - INFO: pass 7, batch 50, accuracy:0.75 loss:0.7831007242202759 time:0.02 sec 2021-08-02 23:32:59,843-INFO: pass 7, batch 60, accuracy:0.78125 loss:0.7463842034339905 time:0.02 sec 2021-08-02 23:32:59,843 - [8970] [line:200] - INFO: pass 7, batch 60, accuracy:0.78125 loss:0.7463842034339905 time:0.02 sec 2021-08-02 23:33:00,065-INFO: pass 7, batch 70, accuracy:0.7890625 loss:0.7395037412643433 time:0.02 sec 2021-08-02 23:33:00,065 - [8970] [line:200] - INFO: pass 7, batch 70, accuracy:0.7890625 loss:0.7395037412643433 time:0.02 sec 2021-08-02 23:33:00,177-INFO: pass 7 train result, current pass mean accuracy:0.7440490432475743 loss:0.7900420278310776 2021-08-02 23:33:00,177 - [8970] [line:207] - INFO: pass 7 train result, current pass mean accuracy:0.7440490432475743 loss:0.7900420278310776 2021-08-02 23:33:00,178-INFO: current pass: 8, start read text 2021-08-02 23:33:00,178 - [8970] [line:179] - INFO: current pass: 8, start read text 2021-08-02 23:33:00,200-INFO: pass 8, batch 0, accuracy:0.7265625 loss:0.8045200109481812 time:0.02 sec 2021-08-02 23:33:00,200 - [8970] [line:200] - INFO: pass 8, batch 0, accuracy:0.7265625 loss:0.8045200109481812 time:0.02 sec 2021-08-02 23:33:00,424-INFO: pass 8, batch 10, accuracy:0.796875 loss:0.732032299041748 time:0.02 sec 2021-08-02 23:33:00,424 - [8970] [line:200] - INFO: pass 8, batch 10, accuracy:0.796875 loss:0.732032299041748 time:0.02 sec 2021-08-02 23:33:00,648-INFO: pass 8, batch 20, accuracy:0.7109375 loss:0.828032910823822 time:0.02 sec 2021-08-02 23:33:00,648 - [8970] [line:200] - INFO: pass 8, batch 20, accuracy:0.7109375 loss:0.828032910823822 time:0.02 sec 2021-08-02 23:33:00,873-INFO: pass 8, batch 30, accuracy:0.734375 loss:0.7989935874938965 time:0.02 sec 2021-08-02 23:33:00,873 - [8970] [line:200] - INFO: pass 8, batch 30, accuracy:0.734375 loss:0.7989935874938965 time:0.02 sec 2021-08-02 23:33:01,105-INFO: pass 8, batch 40, accuracy:0.703125 loss:0.8305236101150513 time:0.02 sec 2021-08-02 23:33:01,105 - [8970] [line:200] - INFO: pass 8, batch 40, accuracy:0.703125 loss:0.8305236101150513 time:0.02 sec 2021-08-02 23:33:01,329-INFO: pass 8, batch 50, accuracy:0.8125 loss:0.7073854207992554 time:0.02 sec 2021-08-02 23:33:01,329 - [8970] [line:200] - INFO: pass 8, batch 50, accuracy:0.8125 loss:0.7073854207992554 time:0.02 sec 2021-08-02 23:33:01,551-INFO: pass 8, batch 60, accuracy:0.7265625 loss:0.7988113164901733 time:0.02 sec 2021-08-02 23:33:01,551 - [8970] [line:200] - INFO: pass 8, batch 60, accuracy:0.7265625 loss:0.7988113164901733 time:0.02 sec 2021-08-02 23:33:01,783-INFO: pass 8, batch 70, accuracy:0.75 loss:0.7754559516906738 time:0.02 sec 2021-08-02 23:33:01,783 - [8970] [line:200] - INFO: pass 8, batch 70, accuracy:0.75 loss:0.7754559516906738 time:0.02 sec 2021-08-02 23:33:01,894-INFO: pass 8 train result, current pass mean accuracy:0.7447312348767331 loss:0.7839354660950208 2021-08-02 23:33:01,894 - [8970] [line:207] - INFO: pass 8 train result, current pass mean accuracy:0.7447312348767331 loss:0.7839354660950208 2021-08-02 23:33:01,895-INFO: current pass: 9, start read text 2021-08-02 23:33:01,895 - [8970] [line:179] - INFO: current pass: 9, start read text 2021-08-02 23:33:01,916-INFO: pass 9, batch 0, accuracy:0.75 loss:0.7755324840545654 time:0.02 sec 2021-08-02 23:33:01,916 - [8970] [line:200] - INFO: pass 9, batch 0, accuracy:0.75 loss:0.7755324840545654 time:0.02 sec 2021-08-02 23:33:02,139-INFO: pass 9, batch 10, accuracy:0.7578125 loss:0.7688612937927246 time:0.02 sec 2021-08-02 23:33:02,139 - [8970] [line:200] - INFO: pass 9, batch 10, accuracy:0.7578125 loss:0.7688612937927246 time:0.02 sec 2021-08-02 23:33:02,371-INFO: pass 9, batch 20, accuracy:0.7734375 loss:0.7474617958068848 time:0.03 sec 2021-08-02 23:33:02,371 - [8970] [line:200] - INFO: pass 9, batch 20, accuracy:0.7734375 loss:0.7474617958068848 time:0.03 sec 2021-08-02 23:33:02,590-INFO: pass 9, batch 30, accuracy:0.7734375 loss:0.7492794394493103 time:0.02 sec 2021-08-02 23:33:02,590 - [8970] [line:200] - INFO: pass 9, batch 30, accuracy:0.7734375 loss:0.7492794394493103 time:0.02 sec 2021-08-02 23:33:02,808-INFO: pass 9, batch 40, accuracy:0.7890625 loss:0.7318198084831238 time:0.02 sec 2021-08-02 23:33:02,808 - [8970] [line:200] - INFO: pass 9, batch 40, accuracy:0.7890625 loss:0.7318198084831238 time:0.02 sec 2021-08-02 23:33:03,032-INFO: pass 9, batch 50, accuracy:0.796875 loss:0.7194355130195618 time:0.02 sec 2021-08-02 23:33:03,032 - [8970] [line:200] - INFO: pass 9, batch 50, accuracy:0.796875 loss:0.7194355130195618 time:0.02 sec 2021-08-02 23:33:03,308-INFO: pass 9, batch 60, accuracy:0.7890625 loss:0.7339265942573547 time:0.02 sec 2021-08-02 23:33:03,308 - [8970] [line:200] - INFO: pass 9, batch 60, accuracy:0.7890625 loss:0.7339265942573547 time:0.02 sec 2021-08-02 23:33:03,530-INFO: pass 9, batch 70, accuracy:0.6640625 loss:0.8642477989196777 time:0.02 sec 2021-08-02 23:33:03,530 - [8970] [line:200] - INFO: pass 9, batch 70, accuracy:0.6640625 loss:0.8642477989196777 time:0.02 sec 2021-08-02 23:33:03,642-INFO: pass 9 train result, current pass mean accuracy:0.7450041119989596 loss:0.7800869918183276 2021-08-02 23:33:03,642 - [8970] [line:207] - INFO: pass 9 train result, current pass mean accuracy:0.7450041119989596 loss:0.7800869918183276 2021-08-02 23:33:03,642-INFO: current pass: 10, start read text 2021-08-02 23:33:03,642 - [8970] [line:179] - INFO: current pass: 10, start read text 2021-08-02 23:33:03,665-INFO: pass 10, batch 0, accuracy:0.78125 loss:0.7379545569419861 time:0.02 sec 2021-08-02 23:33:03,665 - [8970] [line:200] - INFO: pass 10, batch 0, accuracy:0.78125 loss:0.7379545569419861 time:0.02 sec 2021-08-02 23:33:03,889-INFO: pass 10, batch 10, accuracy:0.71875 loss:0.8092842698097229 time:0.02 sec 2021-08-02 23:33:03,889 - [8970] [line:200] - INFO: pass 10, batch 10, accuracy:0.71875 loss:0.8092842698097229 time:0.02 sec 2021-08-02 23:33:04,109-INFO: pass 10, batch 20, accuracy:0.7734375 loss:0.7497470378875732 time:0.02 sec 2021-08-02 23:33:04,109 - [8970] [line:200] - INFO: pass 10, batch 20, accuracy:0.7734375 loss:0.7497470378875732 time:0.02 sec 2021-08-02 23:33:04,333-INFO: pass 10, batch 30, accuracy:0.7734375 loss:0.7431987524032593 time:0.02 sec 2021-08-02 23:33:04,333 - [8970] [line:200] - INFO: pass 10, batch 30, accuracy:0.7734375 loss:0.7431987524032593 time:0.02 sec 2021-08-02 23:33:04,559-INFO: pass 10, batch 40, accuracy:0.7734375 loss:0.7450143694877625 time:0.02 sec 2021-08-02 23:33:04,559 - [8970] [line:200] - INFO: pass 10, batch 40, accuracy:0.7734375 loss:0.7450143694877625 time:0.02 sec 2021-08-02 23:33:04,781-INFO: pass 10, batch 50, accuracy:0.734375 loss:0.790243923664093 time:0.02 sec 2021-08-02 23:33:04,781 - [8970] [line:200] - INFO: pass 10, batch 50, accuracy:0.734375 loss:0.790243923664093 time:0.02 sec 2021-08-02 23:33:05,012-INFO: pass 10, batch 60, accuracy:0.796875 loss:0.717749834060669 time:0.03 sec 2021-08-02 23:33:05,012 - [8970] [line:200] - INFO: pass 10, batch 60, accuracy:0.796875 loss:0.717749834060669 time:0.03 sec 2021-08-02 23:33:05,235-INFO: pass 10, batch 70, accuracy:0.765625 loss:0.7550437450408936 time:0.02 sec 2021-08-02 23:33:05,235 - [8970] [line:200] - INFO: pass 10, batch 70, accuracy:0.765625 loss:0.7550437450408936 time:0.02 sec 2021-08-02 23:33:05,346-INFO: pass 10 train result, current pass mean accuracy:0.7447312348767331 loss:0.7783224480716806 2021-08-02 23:33:05,346 - [8970] [line:207] - INFO: pass 10 train result, current pass mean accuracy:0.7447312348767331 loss:0.7783224480716806 2021-08-02 23:33:05,347-INFO: current pass: 11, start read text 2021-08-02 23:33:05,347 - [8970] [line:179] - INFO: current pass: 11, start read text 2021-08-02 23:33:05,370-INFO: pass 11, batch 0, accuracy:0.8203125 loss:0.6896326541900635 time:0.02 sec 2021-08-02 23:33:05,370 - [8970] [line:200] - INFO: pass 11, batch 0, accuracy:0.8203125 loss:0.6896326541900635 time:0.02 sec 2021-08-02 23:33:05,593-INFO: pass 11, batch 10, accuracy:0.703125 loss:0.8248531818389893 time:0.02 sec 2021-08-02 23:33:05,593 - [8970] [line:200] - INFO: pass 11, batch 10, accuracy:0.703125 loss:0.8248531818389893 time:0.02 sec 2021-08-02 23:33:05,823-INFO: pass 11, batch 20, accuracy:0.75 loss:0.7687795758247375 time:0.02 sec 2021-08-02 23:33:05,823 - [8970] [line:200] - INFO: pass 11, batch 20, accuracy:0.75 loss:0.7687795758247375 time:0.02 sec 2021-08-02 23:33:06,048-INFO: pass 11, batch 30, accuracy:0.671875 loss:0.8578289151191711 time:0.02 sec 2021-08-02 23:33:06,048 - [8970] [line:200] - INFO: pass 11, batch 30, accuracy:0.671875 loss:0.8578289151191711 time:0.02 sec 2021-08-02 23:33:06,267-INFO: pass 11, batch 40, accuracy:0.765625 loss:0.7530752420425415 time:0.02 sec 2021-08-02 23:33:06,267 - [8970] [line:200] - INFO: pass 11, batch 40, accuracy:0.765625 loss:0.7530752420425415 time:0.02 sec 2021-08-02 23:33:06,493-INFO: pass 11, batch 50, accuracy:0.765625 loss:0.749035120010376 time:0.02 sec 2021-08-02 23:33:06,493 - [8970] [line:200] - INFO: pass 11, batch 50, accuracy:0.765625 loss:0.749035120010376 time:0.02 sec 2021-08-02 23:33:06,716-INFO: pass 11, batch 60, accuracy:0.6953125 loss:0.8319502472877502 time:0.02 sec 2021-08-02 23:33:06,716 - [8970] [line:200] - INFO: pass 11, batch 60, accuracy:0.6953125 loss:0.8319502472877502 time:0.02 sec 2021-08-02 23:33:06,939-INFO: pass 11, batch 70, accuracy:0.703125 loss:0.8242783546447754 time:0.02 sec 2021-08-02 23:33:06,939 - [8970] [line:200] - INFO: pass 11, batch 70, accuracy:0.703125 loss:0.8242783546447754 time:0.02 sec 2021-08-02 23:33:07,051-INFO: pass 11 train result, current pass mean accuracy:0.7450041119989596 loss:0.7767221496293419 2021-08-02 23:33:07,051 - [8970] [line:207] - INFO: pass 11 train result, current pass mean accuracy:0.7450041119989596 loss:0.7767221496293419 2021-08-02 23:33:07,051-INFO: training till last epcho, end training 2021-08-02 23:33:07,051 - [8970] [line:214] - INFO: training till last epcho, end training

3.3.3.2 固化训练参数

In [8]

!python freeze.py

2021-08-02 23:34:42,387-INFO: build network cnn_net to freeze 2021-08-02 23:34:42,387 - [9163] [line:46] - INFO: build network cnn_net to freeze 2021-08-02 23:34:43,464-INFO: freeze cnn_net end 2021-08-02 23:34:43,464 - [9163] [line:66] - INFO: freeze cnn_net end

3.3.3.3 验证训练结果

In [9]

!python eval.py

W0802 23:35:19.363926 9263 device_context.cc:252] Please NOTE: device: 0, CUDA Capability: 70, Driver API Version: 10.1, Runtime API Version: 9.0 W0802 23:35:19.368824 9263 device_context.cc:260] device: 0, cuDNN Version: 7.6. total eval count:1036 cost time:7.04 sec predict accuracy:0.750965250965251

3.3.3.4 使用模型进行预测

In [10]

!python infer.py

W0802 23:40:42.673434 9762 device_context.cc:252] Please NOTE: device: 0, CUDA Capability: 70, Driver API Version: 10.1, Runtime API Version: 9.0 W0802 23:40:42.678300 9762 device_context.cc:260] device: 0, cuDNN Version: 7.6. infer text:我 就是 喜欢 你的 知性 大方 温柔 infer result:1

3.3.4 GRU模型的训练与预测

3.3.4.1 训练模型

更改train_parameters中"model_type"为"gru_net"

In [11]

!python train.py

2021-08-02 23:42:27,973-INFO: start train text classification, train params: {'class_dim': 3, 'sample_count': 9656, 'label_dict': {'消极': 0, '中性': 1, '积极': 2}, 'data_dir': 'data/data9740', 'vocabulary': 'vocab.txt', 'train_file_list': 'train.txt', 'label_file': 'label_list.txt', 'save_model_dir': './train_model', 'save_freeze_dir': './infer_model', 'pretrained_model_dir': './pretrained_model', 'continue_train': False, 'pretrained': False, 'model_type': 'gru_net', 'num_epochs': 12, 'train_batch_size': 128, 'use_gpu': True, 'sample_frequency': 10, 'rsm_strategy': {'learning_rate': 0.001, 'lr_epochs': [6, 8, 10], 'lr_decay': [1, 0.5, 0.25, 0.1]}, 'momentum_strategy': {'learning_rate': 0.001, 'lr_epochs': [6, 8, 10], 'lr_decay': [1, 0.5, 0.25, 0.1]}, 'sgd_strategy': {'learning_rate': 0.001, 'lr_epochs': [6, 8, 10], 'lr_decay': [1, 0.5, 0.25, 0.1]}, 'adam_strategy': {'learning_rate': 0.001}} 2021-08-02 23:42:27,973 - [9865] [line:133] - INFO: start train text classification, train params: {'class_dim': 3, 'sample_count': 9656, 'label_dict': {'消极': 0, '中性': 1, '积极': 2}, 'data_dir': 'data/data9740', 'vocabulary': 'vocab.txt', 'train_file_list': 'train.txt', 'label_file': 'label_list.txt', 'save_model_dir': './train_model', 'save_freeze_dir': './infer_model', 'pretrained_model_dir': './pretrained_model', 'continue_train': False, 'pretrained': False, 'model_type': 'gru_net', 'num_epochs': 12, 'train_batch_size': 128, 'use_gpu': True, 'sample_frequency': 10, 'rsm_strategy': {'learning_rate': 0.001, 'lr_epochs': [6, 8, 10], 'lr_decay': [1, 0.5, 0.25, 0.1]}, 'momentum_strategy': {'learning_rate': 0.001, 'lr_epochs': [6, 8, 10], 'lr_decay': [1, 0.5, 0.25, 0.1]}, 'sgd_strategy': {'learning_rate': 0.001, 'lr_epochs': [6, 8, 10], 'lr_decay': [1, 0.5, 0.25, 0.1]}, 'adam_strategy': {'learning_rate': 0.001}} 2021-08-02 23:42:27,973-INFO: create place, use gpu: True 2021-08-02 23:42:27,973 - [9865] [line:134] - INFO: create place, use gpu: True 2021-08-02 23:42:28,397-INFO: define input data tensor 2021-08-02 23:42:28,397 - [9865] [line:139] - INFO: define input data tensor 2021-08-02 23:42:28,398-INFO: build custom reader 2021-08-02 23:42:28,398 - [9865] [line:144] - INFO: build custom reader 2021-08-02 23:42:28,420-INFO: build network gru_net 2021-08-02 23:42:28,420 - [9865] [line:147] - INFO: build network gru_net 2021-08-02 23:42:28,434-INFO: build optimizer 2021-08-02 23:42:28,434 - [9865] [line:164] - INFO: build optimizer W0802 23:42:28.453099 9865 device_context.cc:252] Please NOTE: device: 0, CUDA Capability: 70, Driver API Version: 10.1, Runtime API Version: 9.0 W0802 23:42:28.457787 9865 device_context.cc:260] device: 0, cuDNN Version: 7.6. 2021-08-02 23:42:42,256-INFO: current pass: 0, start read text 2021-08-02 23:42:42,256 - [9865] [line:179] - INFO: current pass: 0, start read text 2021-08-02 23:42:42,304-INFO: pass 0, batch 0, accuracy:0.1640625 loss:1.098846197128296 time:0.05 sec 2021-08-02 23:42:42,304 - [9865] [line:200] - INFO: pass 0, batch 0, accuracy:0.1640625 loss:1.098846197128296 time:0.05 sec 2021-08-02 23:42:42,549-INFO: pass 0, batch 10, accuracy:0.6875 loss:1.0858808755874634 time:0.02 sec 2021-08-02 23:42:42,549 - [9865] [line:200] - INFO: pass 0, batch 10, accuracy:0.6875 loss:1.0858808755874634 time:0.02 sec 2021-08-02 23:42:42,794-INFO: pass 0, batch 20, accuracy:0.765625 loss:1.0678585767745972 time:0.02 sec 2021-08-02 23:42:42,794 - [9865] [line:200] - INFO: pass 0, batch 20, accuracy:0.765625 loss:1.0678585767745972 time:0.02 sec 2021-08-02 23:42:43,050-INFO: pass 0, batch 30, accuracy:0.734375 loss:1.0562174320220947 time:0.03 sec 2021-08-02 23:42:43,050 - [9865] [line:200] - INFO: pass 0, batch 30, accuracy:0.734375 loss:1.0562174320220947 time:0.03 sec 2021-08-02 23:42:43,271-INFO: pass 0, batch 40, accuracy:0.703125 loss:1.0461890697479248 time:0.02 sec 2021-08-02 23:42:43,271 - [9865] [line:200] - INFO: pass 0, batch 40, accuracy:0.703125 loss:1.0461890697479248 time:0.02 sec 2021-08-02 23:42:43,508-INFO: pass 0, batch 50, accuracy:0.703125 loss:1.0345451831817627 time:0.02 sec 2021-08-02 23:42:43,508 - [9865] [line:200] - INFO: pass 0, batch 50, accuracy:0.703125 loss:1.0345451831817627 time:0.02 sec 2021-08-02 23:42:43,747-INFO: pass 0, batch 60, accuracy:0.7890625 loss:1.0082533359527588 time:0.02 sec 2021-08-02 23:42:43,747 - [9865] [line:200] - INFO: pass 0, batch 60, accuracy:0.7890625 loss:1.0082533359527588 time:0.02 sec 2021-08-02 23:42:44,000-INFO: pass 0, batch 70, accuracy:0.7578125 loss:1.0000827312469482 time:0.03 sec 2021-08-02 23:42:44,000 - [9865] [line:200] - INFO: pass 0, batch 70, accuracy:0.7578125 loss:1.0000827312469482 time:0.03 sec 2021-08-02 23:42:44,115-INFO: pass 0 train result, current pass mean accuracy:0.7342479064276344 loss:1.0470415134178965 2021-08-02 23:42:44,115 - [9865] [line:207] - INFO: pass 0 train result, current pass mean accuracy:0.7342479064276344 loss:1.0470415134178965 2021-08-02 23:42:44,115-INFO: temp save pass 0 train result, current best pass mean accuracy:0.7342479064276344 2021-08-02 23:42:44,115 - [9865] [line:209] - INFO: temp save pass 0 train result, current best pass mean accuracy:0.7342479064276344 2021-08-02 23:42:44,725-INFO: current pass: 1, start read text 2021-08-02 23:42:44,725 - [9865] [line:179] - INFO: current pass: 1, start read text 2021-08-02 23:42:44,749-INFO: pass 1, batch 0, accuracy:0.7421875 loss:1.0000437498092651 time:0.02 sec 2021-08-02 23:42:44,749 - [9865] [line:200] - INFO: pass 1, batch 0, accuracy:0.7421875 loss:1.0000437498092651 time:0.02 sec 2021-08-02 23:42:45,005-INFO: pass 1, batch 10, accuracy:0.75 loss:0.9847410917282104 time:0.02 sec 2021-08-02 23:42:45,005 - [9865] [line:200] - INFO: pass 1, batch 10, accuracy:0.75 loss:0.9847410917282104 time:0.02 sec 2021-08-02 23:42:45,249-INFO: pass 1, batch 20, accuracy:0.75 loss:0.9784713983535767 time:0.02 sec 2021-08-02 23:42:45,249 - [9865] [line:200] - INFO: pass 1, batch 20, accuracy:0.75 loss:0.9784713983535767 time:0.02 sec 2021-08-02 23:42:45,508-INFO: pass 1, batch 30, accuracy:0.7421875 loss:0.9676991701126099 time:0.03 sec 2021-08-02 23:42:45,508 - [9865] [line:200] - INFO: pass 1, batch 30, accuracy:0.7421875 loss:0.9676991701126099 time:0.03 sec 2021-08-02 23:42:45,762-INFO: pass 1, batch 40, accuracy:0.7734375 loss:0.943374752998352 time:0.02 sec 2021-08-02 23:42:45,762 - [9865] [line:200] - INFO: pass 1, batch 40, accuracy:0.7734375 loss:0.943374752998352 time:0.02 sec 2021-08-02 23:42:46,002-INFO: pass 1, batch 50, accuracy:0.7109375 loss:0.9600652456283569 time:0.03 sec 2021-08-02 23:42:46,002 - [9865] [line:200] - INFO: pass 1, batch 50, accuracy:0.7109375 loss:0.9600652456283569 time:0.03 sec 2021-08-02 23:42:46,238-INFO: pass 1, batch 60, accuracy:0.7109375 loss:0.9512149691581726 time:0.02 sec 2021-08-02 23:42:46,238 - [9865] [line:200] - INFO: pass 1, batch 60, accuracy:0.7109375 loss:0.9512149691581726 time:0.02 sec 2021-08-02 23:42:46,462-INFO: pass 1, batch 70, accuracy:0.6796875 loss:0.9635403156280518 time:0.02 sec 2021-08-02 23:42:46,462 - [9865] [line:200] - INFO: pass 1, batch 70, accuracy:0.6796875 loss:0.9635403156280518 time:0.02 sec 2021-08-02 23:42:46,592-INFO: pass 1 train result, current pass mean accuracy:0.7445947967077556 loss:0.961847015117344 2021-08-02 23:42:46,592 - [9865] [line:207] - INFO: pass 1 train result, current pass mean accuracy:0.7445947967077556 loss:0.961847015117344 2021-08-02 23:42:46,592-INFO: temp save pass 1 train result, current best pass mean accuracy:0.7445947967077556 2021-08-02 23:42:46,592 - [9865] [line:209] - INFO: temp save pass 1 train result, current best pass mean accuracy:0.7445947967077556 2021-08-02 23:42:47,206-INFO: current pass: 2, start read text 2021-08-02 23:42:47,206 - [9865] [line:179] - INFO: current pass: 2, start read text 2021-08-02 23:42:47,236-INFO: pass 2, batch 0, accuracy:0.7578125 loss:0.9222443103790283 time:0.03 sec 2021-08-02 23:42:47,236 - [9865] [line:200] - INFO: pass 2, batch 0, accuracy:0.7578125 loss:0.9222443103790283 time:0.03 sec 2021-08-02 23:42:47,486-INFO: pass 2, batch 10, accuracy:0.7578125 loss:0.9142199754714966 time:0.02 sec 2021-08-02 23:42:47,486 - [9865] [line:200] - INFO: pass 2, batch 10, accuracy:0.7578125 loss:0.9142199754714966 time:0.02 sec 2021-08-02 23:42:47,730-INFO: pass 2, batch 20, accuracy:0.7578125 loss:0.9075149893760681 time:0.02 sec 2021-08-02 23:42:47,730 - [9865] [line:200] - INFO: pass 2, batch 20, accuracy:0.7578125 loss:0.9075149893760681 time:0.02 sec 2021-08-02 23:42:47,998-INFO: pass 2, batch 30, accuracy:0.7265625 loss:0.9142878651618958 time:0.02 sec 2021-08-02 23:42:47,998 - [9865] [line:200] - INFO: pass 2, batch 30, accuracy:0.7265625 loss:0.9142878651618958 time:0.02 sec 2021-08-02 23:42:48,234-INFO: pass 2, batch 40, accuracy:0.703125 loss:0.9199079275131226 time:0.02 sec 2021-08-02 23:42:48,234 - [9865] [line:200] - INFO: pass 2, batch 40, accuracy:0.703125 loss:0.9199079275131226 time:0.02 sec 2021-08-02 23:42:48,468-INFO: pass 2, batch 50, accuracy:0.7578125 loss:0.8813605308532715 time:0.02 sec 2021-08-02 23:42:48,468 - [9865] [line:200] - INFO: pass 2, batch 50, accuracy:0.7578125 loss:0.8813605308532715 time:0.02 sec 2021-08-02 23:42:48,706-INFO: pass 2, batch 60, accuracy:0.78125 loss:0.8650902509689331 time:0.02 sec 2021-08-02 23:42:48,706 - [9865] [line:200] - INFO: pass 2, batch 60, accuracy:0.78125 loss:0.8650902509689331 time:0.02 sec 2021-08-02 23:42:48,944-INFO: pass 2, batch 70, accuracy:0.6953125 loss:0.9035236239433289 time:0.03 sec 2021-08-02 23:42:48,944 - [9865] [line:200] - INFO: pass 2, batch 70, accuracy:0.6953125 loss:0.9035236239433289 time:0.03 sec 2021-08-02 23:42:49,057-INFO: pass 2 train result, current pass mean accuracy:0.7448676738299822 loss:0.899472389566271 2021-08-02 23:42:49,057 - [9865] [line:207] - INFO: pass 2 train result, current pass mean accuracy:0.7448676738299822 loss:0.899472389566271 2021-08-02 23:42:49,058-INFO: temp save pass 2 train result, current best pass mean accuracy:0.7448676738299822 2021-08-02 23:42:49,058 - [9865] [line:209] - INFO: temp save pass 2 train result, current best pass mean accuracy:0.7448676738299822 2021-08-02 23:42:49,675-INFO: current pass: 3, start read text 2021-08-02 23:42:49,675 - [9865] [line:179] - INFO: current pass: 3, start read text 2021-08-02 23:42:49,699-INFO: pass 3, batch 0, accuracy:0.703125 loss:0.8954682350158691 time:0.02 sec 2021-08-02 23:42:49,699 - [9865] [line:200] - INFO: pass 3, batch 0, accuracy:0.703125 loss:0.8954682350158691 time:0.02 sec 2021-08-02 23:42:49,948-INFO: pass 3, batch 10, accuracy:0.7578125 loss:0.860008955001831 time:0.03 sec 2021-08-02 23:42:49,948 - [9865] [line:200] - INFO: pass 3, batch 10, accuracy:0.7578125 loss:0.860008955001831 time:0.03 sec 2021-08-02 23:42:50,201-INFO: pass 3, batch 20, accuracy:0.7421875 loss:0.8643391132354736 time:0.02 sec 2021-08-02 23:42:50,201 - [9865] [line:200] - INFO: pass 3, batch 20, accuracy:0.7421875 loss:0.8643391132354736 time:0.02 sec 2021-08-02 23:42:50,435-INFO: pass 3, batch 30, accuracy:0.7890625 loss:0.8279826641082764 time:0.02 sec 2021-08-02 23:42:50,435 - [9865] [line:200] - INFO: pass 3, batch 30, accuracy:0.7890625 loss:0.8279826641082764 time:0.02 sec 2021-08-02 23:42:50,671-INFO: pass 3, batch 40, accuracy:0.7421875 loss:0.8540748357772827 time:0.02 sec 2021-08-02 23:42:50,671 - [9865] [line:200] - INFO: pass 3, batch 40, accuracy:0.7421875 loss:0.8540748357772827 time:0.02 sec 2021-08-02 23:42:50,915-INFO: pass 3, batch 50, accuracy:0.765625 loss:0.8309152722358704 time:0.02 sec 2021-08-02 23:42:50,915 - [9865] [line:200] - INFO: pass 3, batch 50, accuracy:0.765625 loss:0.8309152722358704 time:0.02 sec 2021-08-02 23:42:51,177-INFO: pass 3, batch 60, accuracy:0.703125 loss:0.8847692608833313 time:0.03 sec 2021-08-02 23:42:51,177 - [9865] [line:200] - INFO: pass 3, batch 60, accuracy:0.703125 loss:0.8847692608833313 time:0.03 sec 2021-08-02 23:42:51,413-INFO: pass 3, batch 70, accuracy:0.71875 loss:0.8622555136680603 time:0.02 sec 2021-08-02 23:42:51,413 - [9865] [line:200] - INFO: pass 3, batch 70, accuracy:0.71875 loss:0.8622555136680603 time:0.02 sec 2021-08-02 23:42:51,534-INFO: pass 3 train result, current pass mean accuracy:0.7445947967077556 loss:0.8547408769005224 2021-08-02 23:42:51,534 - [9865] [line:207] - INFO: pass 3 train result, current pass mean accuracy:0.7445947967077556 loss:0.8547408769005224 2021-08-02 23:42:51,535-INFO: current pass: 4, start read text 2021-08-02 23:42:51,535 - [9865] [line:179] - INFO: current pass: 4, start read text 2021-08-02 23:42:51,558-INFO: pass 4, batch 0, accuracy:0.7265625 loss:0.845553994178772 time:0.02 sec 2021-08-02 23:42:51,558 - [9865] [line:200] - INFO: pass 4, batch 0, accuracy:0.7265625 loss:0.845553994178772 time:0.02 sec 2021-08-02 23:42:51,810-INFO: pass 4, batch 10, accuracy:0.71875 loss:0.850944995880127 time:0.02 sec 2021-08-02 23:42:51,810 - [9865] [line:200] - INFO: pass 4, batch 10, accuracy:0.71875 loss:0.850944995880127 time:0.02 sec 2021-08-02 23:42:52,041-INFO: pass 4, batch 20, accuracy:0.7109375 loss:0.8596910834312439 time:0.02 sec 2021-08-02 23:42:52,041 - [9865] [line:200] - INFO: pass 4, batch 20, accuracy:0.7109375 loss:0.8596910834312439 time:0.02 sec 2021-08-02 23:42:52,281-INFO: pass 4, batch 30, accuracy:0.7578125 loss:0.8184800148010254 time:0.02 sec 2021-08-02 23:42:52,281 - [9865] [line:200] - INFO: pass 4, batch 30, accuracy:0.7578125 loss:0.8184800148010254 time:0.02 sec 2021-08-02 23:42:52,519-INFO: pass 4, batch 40, accuracy:0.7578125 loss:0.8150058388710022 time:0.02 sec 2021-08-02 23:42:52,519 - [9865] [line:200] - INFO: pass 4, batch 40, accuracy:0.7578125 loss:0.8150058388710022 time:0.02 sec 2021-08-02 23:42:52,772-INFO: pass 4, batch 50, accuracy:0.7578125 loss:0.8035138249397278 time:0.03 sec 2021-08-02 23:42:52,772 - [9865] [line:200] - INFO: pass 4, batch 50, accuracy:0.7578125 loss:0.8035138249397278 time:0.03 sec 2021-08-02 23:42:53,018-INFO: pass 4, batch 60, accuracy:0.7421875 loss:0.8203713893890381 time:0.03 sec 2021-08-02 23:42:53,018 - [9865] [line:200] - INFO: pass 4, batch 60, accuracy:0.7421875 loss:0.8203713893890381 time:0.03 sec 2021-08-02 23:42:53,256-INFO: pass 4, batch 70, accuracy:0.75 loss:0.8069104552268982 time:0.03 sec 2021-08-02 23:42:53,256 - [9865] [line:200] - INFO: pass 4, batch 70, accuracy:0.75 loss:0.8069104552268982 time:0.03 sec 2021-08-02 23:42:53,379-INFO: pass 4 train result, current pass mean accuracy:0.745140550167937 loss:0.8219439818670875 2021-08-02 23:42:53,379 - [9865] [line:207] - INFO: pass 4 train result, current pass mean accuracy:0.745140550167937 loss:0.8219439818670875 2021-08-02 23:42:53,379-INFO: temp save pass 4 train result, current best pass mean accuracy:0.745140550167937 2021-08-02 23:42:53,379 - [9865] [line:209] - INFO: temp save pass 4 train result, current best pass mean accuracy:0.745140550167937 2021-08-02 23:42:54,390-INFO: current pass: 5, start read text 2021-08-02 23:42:54,390 - [9865] [line:179] - INFO: current pass: 5, start read text 2021-08-02 23:42:54,413-INFO: pass 5, batch 0, accuracy:0.7265625 loss:0.8307363986968994 time:0.02 sec 2021-08-02 23:42:54,413 - [9865] [line:200] - INFO: pass 5, batch 0, accuracy:0.7265625 loss:0.8307363986968994 time:0.02 sec 2021-08-02 23:42:54,666-INFO: pass 5, batch 10, accuracy:0.7578125 loss:0.8029578328132629 time:0.03 sec 2021-08-02 23:42:54,666 - [9865] [line:200] - INFO: pass 5, batch 10, accuracy:0.7578125 loss:0.8029578328132629 time:0.03 sec 2021-08-02 23:42:54,895-INFO: pass 5, batch 20, accuracy:0.7890625 loss:0.7628553509712219 time:0.02 sec 2021-08-02 23:42:54,895 - [9865] [line:200] - INFO: pass 5, batch 20, accuracy:0.7890625 loss:0.7628553509712219 time:0.02 sec 2021-08-02 23:42:55,131-INFO: pass 5, batch 30, accuracy:0.75 loss:0.7868269681930542 time:0.02 sec 2021-08-02 23:42:55,131 - [9865] [line:200] - INFO: pass 5, batch 30, accuracy:0.75 loss:0.7868269681930542 time:0.02 sec 2021-08-02 23:42:55,376-INFO: pass 5, batch 40, accuracy:0.7421875 loss:0.8057217001914978 time:0.02 sec 2021-08-02 23:42:55,376 - [9865] [line:200] - INFO: pass 5, batch 40, accuracy:0.7421875 loss:0.8057217001914978 time:0.02 sec 2021-08-02 23:42:55,620-INFO: pass 5, batch 50, accuracy:0.7734375 loss:0.7623082399368286 time:0.02 sec 2021-08-02 23:42:55,620 - [9865] [line:200] - INFO: pass 5, batch 50, accuracy:0.7734375 loss:0.7623082399368286 time:0.02 sec 2021-08-02 23:42:55,873-INFO: pass 5, batch 60, accuracy:0.6640625 loss:0.8758686780929565 time:0.02 sec 2021-08-02 23:42:55,873 - [9865] [line:200] - INFO: pass 5, batch 60, accuracy:0.6640625 loss:0.8758686780929565 time:0.02 sec 2021-08-02 23:42:56,113-INFO: pass 5, batch 70, accuracy:0.7578125 loss:0.7838821411132812 time:0.02 sec 2021-08-02 23:42:56,113 - [9865] [line:200] - INFO: pass 5, batch 70, accuracy:0.7578125 loss:0.7838821411132812 time:0.02 sec 2021-08-02 23:42:56,235-INFO: pass 5 train result, current pass mean accuracy:0.7441854814165517 loss:0.7995447736037405 2021-08-02 23:42:56,235 - [9865] [line:207] - INFO: pass 5 train result, current pass mean accuracy:0.7441854814165517 loss:0.7995447736037405 2021-08-02 23:42:56,235-INFO: current pass: 6, start read text 2021-08-02 23:42:56,235 - [9865] [line:179] - INFO: current pass: 6, start read text 2021-08-02 23:42:56,259-INFO: pass 6, batch 0, accuracy:0.7421875 loss:0.7975800037384033 time:0.02 sec 2021-08-02 23:42:56,259 - [9865] [line:200] - INFO: pass 6, batch 0, accuracy:0.7421875 loss:0.7975800037384033 time:0.02 sec 2021-08-02 23:42:56,512-INFO: pass 6, batch 10, accuracy:0.7578125 loss:0.7741910815238953 time:0.02 sec 2021-08-02 23:42:56,512 - [9865] [line:200] - INFO: pass 6, batch 10, accuracy:0.7578125 loss:0.7741910815238953 time:0.02 sec 2021-08-02 23:42:56,756-INFO: pass 6, batch 20, accuracy:0.7734375 loss:0.7690607905387878 time:0.03 sec 2021-08-02 23:42:56,756 - [9865] [line:200] - INFO: pass 6, batch 20, accuracy:0.7734375 loss:0.7690607905387878 time:0.03 sec 2021-08-02 23:42:57,000-INFO: pass 6, batch 30, accuracy:0.765625 loss:0.7575628757476807 time:0.02 sec 2021-08-02 23:42:57,000 - [9865] [line:200] - INFO: pass 6, batch 30, accuracy:0.765625 loss:0.7575628757476807 time:0.02 sec 2021-08-02 23:42:57,233-INFO: pass 6, batch 40, accuracy:0.671875 loss:0.8592793941497803 time:0.02 sec 2021-08-02 23:42:57,233 - [9865] [line:200] - INFO: pass 6, batch 40, accuracy:0.671875 loss:0.8592793941497803 time:0.02 sec 2021-08-02 23:42:57,456-INFO: pass 6, batch 50, accuracy:0.7109375 loss:0.8231767416000366 time:0.02 sec 2021-08-02 23:42:57,456 - [9865] [line:200] - INFO: pass 6, batch 50, accuracy:0.7109375 loss:0.8231767416000366 time:0.02 sec 2021-08-02 23:42:57,707-INFO: pass 6, batch 60, accuracy:0.7578125 loss:0.7802766561508179 time:0.02 sec 2021-08-02 23:42:57,707 - [9865] [line:200] - INFO: pass 6, batch 60, accuracy:0.7578125 loss:0.7802766561508179 time:0.02 sec 2021-08-02 23:42:57,953-INFO: pass 6, batch 70, accuracy:0.7421875 loss:0.7809707522392273 time:0.02 sec 2021-08-02 23:42:57,953 - [9865] [line:200] - INFO: pass 6, batch 70, accuracy:0.7421875 loss:0.7809707522392273 time:0.02 sec 2021-08-02 23:42:58,074-INFO: pass 6 train result, current pass mean accuracy:0.745140550167937 loss:0.7857619398518613 2021-08-02 23:42:58,074 - [9865] [line:207] - INFO: pass 6 train result, current pass mean accuracy:0.745140550167937 loss:0.7857619398518613 2021-08-02 23:42:58,074-INFO: current pass: 7, start read text 2021-08-02 23:42:58,074 - [9865] [line:179] - INFO: current pass: 7, start read text 2021-08-02 23:42:58,104-INFO: pass 7, batch 0, accuracy:0.703125 loss:0.831833004951477 time:0.03 sec 2021-08-02 23:42:58,104 - [9865] [line:200] - INFO: pass 7, batch 0, accuracy:0.703125 loss:0.831833004951477 time:0.03 sec 2021-08-02 23:42:58,356-INFO: pass 7, batch 10, accuracy:0.7265625 loss:0.8091002106666565 time:0.02 sec 2021-08-02 23:42:58,356 - [9865] [line:200] - INFO: pass 7, batch 10, accuracy:0.7265625 loss:0.8091002106666565 time:0.02 sec 2021-08-02 23:42:58,615-INFO: pass 7, batch 20, accuracy:0.734375 loss:0.7975572943687439 time:0.02 sec 2021-08-02 23:42:58,615 - [9865] [line:200] - INFO: pass 7, batch 20, accuracy:0.734375 loss:0.7975572943687439 time:0.02 sec 2021-08-02 23:42:58,859-INFO: pass 7, batch 30, accuracy:0.7578125 loss:0.7585275769233704 time:0.03 sec 2021-08-02 23:42:58,859 - [9865] [line:200] - INFO: pass 7, batch 30, accuracy:0.7578125 loss:0.7585275769233704 time:0.03 sec 2021-08-02 23:42:59,087-INFO: pass 7, batch 40, accuracy:0.796875 loss:0.7215265035629272 time:0.02 sec 2021-08-02 23:42:59,087 - [9865] [line:200] - INFO: pass 7, batch 40, accuracy:0.796875 loss:0.7215265035629272 time:0.02 sec 2021-08-02 23:42:59,321-INFO: pass 7, batch 50, accuracy:0.75 loss:0.7717796564102173 time:0.02 sec 2021-08-02 23:42:59,321 - [9865] [line:200] - INFO: pass 7, batch 50, accuracy:0.75 loss:0.7717796564102173 time:0.02 sec 2021-08-02 23:42:59,557-INFO: pass 7, batch 60, accuracy:0.7890625 loss:0.7347252368927002 time:0.02 sec 2021-08-02 23:42:59,557 - [9865] [line:200] - INFO: pass 7, batch 60, accuracy:0.7890625 loss:0.7347252368927002 time:0.02 sec 2021-08-02 23:42:59,824-INFO: pass 7, batch 70, accuracy:0.7421875 loss:0.7753323912620544 time:0.02 sec 2021-08-02 23:42:59,824 - [9865] [line:200] - INFO: pass 7, batch 70, accuracy:0.7421875 loss:0.7753323912620544 time:0.02 sec 2021-08-02 23:42:59,945-INFO: pass 7 train result, current pass mean accuracy:0.7450041119989596 loss:0.7787333052409323 2021-08-02 23:42:59,945 - [9865] [line:207] - INFO: pass 7 train result, current pass mean accuracy:0.7450041119989596 loss:0.7787333052409323 2021-08-02 23:42:59,945-INFO: current pass: 8, start read text 2021-08-02 23:42:59,945 - [9865] [line:179] - INFO: current pass: 8, start read text 2021-08-02 23:42:59,967-INFO: pass 8, batch 0, accuracy:0.7578125 loss:0.7613320350646973 time:0.02 sec 2021-08-02 23:42:59,967 - [9865] [line:200] - INFO: pass 8, batch 0, accuracy:0.7578125 loss:0.7613320350646973 time:0.02 sec 2021-08-02 23:43:00,203-INFO: pass 8, batch 10, accuracy:0.6796875 loss:0.8612288236618042 time:0.02 sec 2021-08-02 23:43:00,203 - [9865] [line:200] - INFO: pass 8, batch 10, accuracy:0.6796875 loss:0.8612288236618042 time:0.02 sec 2021-08-02 23:43:00,469-INFO: pass 8, batch 20, accuracy:0.7109375 loss:0.8205891847610474 time:0.03 sec 2021-08-02 23:43:00,469 - [9865] [line:200] - INFO: pass 8, batch 20, accuracy:0.7109375 loss:0.8205891847610474 time:0.03 sec 2021-08-02 23:43:00,713-INFO: pass 8, batch 30, accuracy:0.6953125 loss:0.8218110203742981 time:0.03 sec 2021-08-02 23:43:00,713 - [9865] [line:200] - INFO: pass 8, batch 30, accuracy:0.6953125 loss:0.8218110203742981 time:0.03 sec 2021-08-02 23:43:00,936-INFO: pass 8, batch 40, accuracy:0.71875 loss:0.8052453398704529 time:0.02 sec 2021-08-02 23:43:00,936 - [9865] [line:200] - INFO: pass 8, batch 40, accuracy:0.71875 loss:0.8052453398704529 time:0.02 sec 2021-08-02 23:43:01,202-INFO: pass 8, batch 50, accuracy:0.765625 loss:0.7568838596343994 time:0.02 sec 2021-08-02 23:43:01,202 - [9865] [line:200] - INFO: pass 8, batch 50, accuracy:0.765625 loss:0.7568838596343994 time:0.02 sec 2021-08-02 23:43:01,477-INFO: pass 8, batch 60, accuracy:0.71875 loss:0.8064470291137695 time:0.02 sec 2021-08-02 23:43:01,477 - [9865] [line:200] - INFO: pass 8, batch 60, accuracy:0.71875 loss:0.8064470291137695 time:0.02 sec 2021-08-02 23:43:01,722-INFO: pass 8, batch 70, accuracy:0.7734375 loss:0.741539478302002 time:0.02 sec 2021-08-02 23:43:01,722 - [9865] [line:200] - INFO: pass 8, batch 70, accuracy:0.7734375 loss:0.741539478302002 time:0.02 sec 2021-08-02 23:43:01,844-INFO: pass 8 train result, current pass mean accuracy:0.7450041119989596 loss:0.7742127031087875 2021-08-02 23:43:01,844 - [9865] [line:207] - INFO: pass 8 train result, current pass mean accuracy:0.7450041119989596 loss:0.7742127031087875 2021-08-02 23:43:01,844-INFO: current pass: 9, start read text 2021-08-02 23:43:01,844 - [9865] [line:179] - INFO: current pass: 9, start read text 2021-08-02 23:43:01,866-INFO: pass 9, batch 0, accuracy:0.78125 loss:0.7313366532325745 time:0.02 sec 2021-08-02 23:43:01,866 - [9865] [line:200] - INFO: pass 9, batch 0, accuracy:0.78125 loss:0.7313366532325745 time:0.02 sec 2021-08-02 23:43:02,111-INFO: pass 9, batch 10, accuracy:0.75 loss:0.7556494474411011 time:0.02 sec 2021-08-02 23:43:02,111 - [9865] [line:200] - INFO: pass 9, batch 10, accuracy:0.75 loss:0.7556494474411011 time:0.02 sec 2021-08-02 23:43:02,343-INFO: pass 9, batch 20, accuracy:0.734375 loss:0.7892246246337891 time:0.03 sec 2021-08-02 23:43:02,343 - [9865] [line:200] - INFO: pass 9, batch 20, accuracy:0.734375 loss:0.7892246246337891 time:0.03 sec 2021-08-02 23:43:02,597-INFO: pass 9, batch 30, accuracy:0.6796875 loss:0.8557205200195312 time:0.03 sec 2021-08-02 23:43:02,597 - [9865] [line:200] - INFO: pass 9, batch 30, accuracy:0.6796875 loss:0.8557205200195312 time:0.03 sec 2021-08-02 23:43:02,842-INFO: pass 9, batch 40, accuracy:0.6953125 loss:0.8161095380783081 time:0.02 sec 2021-08-02 23:43:02,842 - [9865] [line:200] - INFO: pass 9, batch 40, accuracy:0.6953125 loss:0.8161095380783081 time:0.02 sec 2021-08-02 23:43:03,078-INFO: pass 9, batch 50, accuracy:0.6796875 loss:0.8485594987869263 time:0.02 sec 2021-08-02 23:43:03,078 - [9865] [line:200] - INFO: pass 9, batch 50, accuracy:0.6796875 loss:0.8485594987869263 time:0.02 sec 2021-08-02 23:43:03,330-INFO: pass 9, batch 60, accuracy:0.765625 loss:0.7638888359069824 time:0.02 sec 2021-08-02 23:43:03,330 - [9865] [line:200] - INFO: pass 9, batch 60, accuracy:0.765625 loss:0.7638888359069824 time:0.02 sec 2021-08-02 23:43:03,563-INFO: pass 9, batch 70, accuracy:0.7578125 loss:0.7412906885147095 time:0.02 sec 2021-08-02 23:43:03,563 - [9865] [line:200] - INFO: pass 9, batch 70, accuracy:0.7578125 loss:0.7412906885147095 time:0.02 sec 2021-08-02 23:43:03,693-INFO: pass 9 train result, current pass mean accuracy:0.7445947967077556 loss:0.7718612896768671 2021-08-02 23:43:03,693 - [9865] [line:207] - INFO: pass 9 train result, current pass mean accuracy:0.7445947967077556 loss:0.7718612896768671 2021-08-02 23:43:03,693-INFO: current pass: 10, start read text 2021-08-02 23:43:03,693 - [9865] [line:179] - INFO: current pass: 10, start read text 2021-08-02 23:43:03,723-INFO: pass 10, batch 0, accuracy:0.734375 loss:0.7869415283203125 time:0.03 sec 2021-08-02 23:43:03,723 - [9865] [line:200] - INFO: pass 10, batch 0, accuracy:0.734375 loss:0.7869415283203125 time:0.03 sec 2021-08-02 23:43:03,967-INFO: pass 10, batch 10, accuracy:0.7890625 loss:0.7224206924438477 time:0.02 sec 2021-08-02 23:43:03,967 - [9865] [line:200] - INFO: pass 10, batch 10, accuracy:0.7890625 loss:0.7224206924438477 time:0.02 sec 2021-08-02 23:43:04,212-INFO: pass 10, batch 20, accuracy:0.7578125 loss:0.7613909244537354 time:0.02 sec 2021-08-02 23:43:04,212 - [9865] [line:200] - INFO: pass 10, batch 20, accuracy:0.7578125 loss:0.7613909244537354 time:0.02 sec 2021-08-02 23:43:04,455-INFO: pass 10, batch 30, accuracy:0.7734375 loss:0.74172043800354 time:0.03 sec 2021-08-02 23:43:04,455 - [9865] [line:200] - INFO: pass 10, batch 30, accuracy:0.7734375 loss:0.74172043800354 time:0.03 sec 2021-08-02 23:43:04,721-INFO: pass 10, batch 40, accuracy:0.6796875 loss:0.8513548374176025 time:0.02 sec 2021-08-02 23:43:04,721 - [9865] [line:200] - INFO: pass 10, batch 40, accuracy:0.6796875 loss:0.8513548374176025 time:0.02 sec 2021-08-02 23:43:04,958-INFO: pass 10, batch 50, accuracy:0.8203125 loss:0.6757145524024963 time:0.02 sec 2021-08-02 23:43:04,958 - [9865] [line:200] - INFO: pass 10, batch 50, accuracy:0.8203125 loss:0.6757145524024963 time:0.02 sec 2021-08-02 23:43:05,202-INFO: pass 10, batch 60, accuracy:0.7265625 loss:0.7822446227073669 time:0.02 sec 2021-08-02 23:43:05,202 - [9865] [line:200] - INFO: pass 10, batch 60, accuracy:0.7265625 loss:0.7822446227073669 time:0.02 sec 2021-08-02 23:43:05,438-INFO: pass 10, batch 70, accuracy:0.7734375 loss:0.73652184009552 time:0.02 sec 2021-08-02 23:43:05,438 - [9865] [line:200] - INFO: pass 10, batch 70, accuracy:0.7734375 loss:0.73652184009552 time:0.02 sec 2021-08-02 23:43:05,568-INFO: pass 10 train result, current pass mean accuracy:0.7454134272901636 loss:0.7690599082331908 2021-08-02 23:43:05,568 - [9865] [line:207] - INFO: pass 10 train result, current pass mean accuracy:0.7454134272901636 loss:0.7690599082331908 2021-08-02 23:43:05,569-INFO: temp save pass 10 train result, current best pass mean accuracy:0.7454134272901636 2021-08-02 23:43:05,569 - [9865] [line:209] - INFO: temp save pass 10 train result, current best pass mean accuracy:0.7454134272901636 2021-08-02 23:43:06,188-INFO: current pass: 11, start read text 2021-08-02 23:43:06,188 - [9865] [line:179] - INFO: current pass: 11, start read text 2021-08-02 23:43:06,209-INFO: pass 11, batch 0, accuracy:0.7578125 loss:0.7463846802711487 time:0.02 sec 2021-08-02 23:43:06,209 - [9865] [line:200] - INFO: pass 11, batch 0, accuracy:0.7578125 loss:0.7463846802711487 time:0.02 sec 2021-08-02 23:43:06,440-INFO: pass 11, batch 10, accuracy:0.734375 loss:0.7693502902984619 time:0.02 sec 2021-08-02 23:43:06,440 - [9865] [line:200] - INFO: pass 11, batch 10, accuracy:0.734375 loss:0.7693502902984619 time:0.02 sec 2021-08-02 23:43:06,677-INFO: pass 11, batch 20, accuracy:0.6953125 loss:0.8175010681152344 time:0.02 sec 2021-08-02 23:43:06,677 - [9865] [line:200] - INFO: pass 11, batch 20, accuracy:0.6953125 loss:0.8175010681152344 time:0.02 sec 2021-08-02 23:43:06,927-INFO: pass 11, batch 30, accuracy:0.6875 loss:0.8315010070800781 time:0.02 sec 2021-08-02 23:43:06,927 - [9865] [line:200] - INFO: pass 11, batch 30, accuracy:0.6875 loss:0.8315010070800781 time:0.02 sec 2021-08-02 23:43:07,196-INFO: pass 11, batch 40, accuracy:0.6875 loss:0.8320574760437012 time:0.03 sec 2021-08-02 23:43:07,196 - [9865] [line:200] - INFO: pass 11, batch 40, accuracy:0.6875 loss:0.8320574760437012 time:0.03 sec 2021-08-02 23:43:07,425-INFO: pass 11, batch 50, accuracy:0.75 loss:0.7554004788398743 time:0.02 sec 2021-08-02 23:43:07,425 - [9865] [line:200] - INFO: pass 11, batch 50, accuracy:0.75 loss:0.7554004788398743 time:0.02 sec 2021-08-02 23:43:07,669-INFO: pass 11, batch 60, accuracy:0.7578125 loss:0.7624839544296265 time:0.02 sec 2021-08-02 23:43:07,669 - [9865] [line:200] - INFO: pass 11, batch 60, accuracy:0.7578125 loss:0.7624839544296265 time:0.02 sec 2021-08-02 23:43:07,912-INFO: pass 11, batch 70, accuracy:0.734375 loss:0.779876708984375 time:0.02 sec 2021-08-02 23:43:07,912 - [9865] [line:200] - INFO: pass 11, batch 70, accuracy:0.734375 loss:0.779876708984375 time:0.02 sec 2021-08-02 23:43:08,036-INFO: pass 11 train result, current pass mean accuracy:0.7458227425813675 loss:0.7677009952695746 2021-08-02 23:43:08,036 - [9865] [line:207] - INFO: pass 11 train result, current pass mean accuracy:0.7458227425813675 loss:0.7677009952695746 2021-08-02 23:43:08,036-INFO: temp save pass 11 train result, current best pass mean accuracy:0.7458227425813675 2021-08-02 23:43:08,036 - [9865] [line:209] - INFO: temp save pass 11 train result, current best pass mean accuracy:0.7458227425813675 2021-08-02 23:43:08,644-INFO: training till last epcho, end training 2021-08-02 23:43:08,644 - [9865] [line:214] - INFO: training till last epcho, end training

3.3.4.2 固化训练参数

In [12]

!python freeze.py

2021-08-02 23:44:00,009-INFO: build network gru_net to freeze 2021-08-02 23:44:00,009 - [10008] [line:46] - INFO: build network gru_net to freeze 2021-08-02 23:44:01,099-INFO: freeze gru_net end 2021-08-02 23:44:01,099 - [10008] [line:66] - INFO: freeze gru_net end

3.3.4.3 验证训练结果

In [13]

!python eval.py

W0802 23:44:07.763453 10064 device_context.cc:252] Please NOTE: device: 0, CUDA Capability: 70, Driver API Version: 10.1, Runtime API Version: 9.0 W0802 23:44:07.768394 10064 device_context.cc:260] device: 0, cuDNN Version: 7.6. total eval count:1036 cost time:7.16 sec predict accuracy:0.750965250965251

3.3.4.4 使用模型进行预测

In [14]

!python infer.py

W0802 23:44:53.459877 10131 device_context.cc:252] Please NOTE: device: 0, CUDA Capability: 70, Driver API Version: 10.1, Runtime API Version: 9.0 W0802 23:44:53.464712 10131 device_context.cc:260] device: 0, cuDNN Version: 7.6. infer text:我 就是 喜欢 你的 知性 大方 温柔 infer result:1

3.3.5 LSTM模型

3.3.5.1 训练模型

更改train_parameters中"model_type"为"lstm_net"

In [15]

!python train.py

2021-08-02 23:45:30,782-INFO: start train text classification, train params: {'class_dim': 3, 'sample_count': 9656, 'label_dict': {'消极': 0, '中性': 1, '积极': 2}, 'data_dir': 'data/data9740', 'vocabulary': 'vocab.txt', 'train_file_list': 'train.txt', 'label_file': 'label_list.txt', 'save_model_dir': './train_model', 'save_freeze_dir': './infer_model', 'pretrained_model_dir': './pretrained_model', 'continue_train': False, 'pretrained': False, 'model_type': 'lstm_net', 'num_epochs': 12, 'train_batch_size': 128, 'use_gpu': True, 'sample_frequency': 10, 'rsm_strategy': {'learning_rate': 0.001, 'lr_epochs': [6, 8, 10], 'lr_decay': [1, 0.5, 0.25, 0.1]}, 'momentum_strategy': {'learning_rate': 0.001, 'lr_epochs': [6, 8, 10], 'lr_decay': [1, 0.5, 0.25, 0.1]}, 'sgd_strategy': {'learning_rate': 0.001, 'lr_epochs': [6, 8, 10], 'lr_decay': [1, 0.5, 0.25, 0.1]}, 'adam_strategy': {'learning_rate': 0.001}} 2021-08-02 23:45:30,782 - [10189] [line:133] - INFO: start train text classification, train params: {'class_dim': 3, 'sample_count': 9656, 'label_dict': {'消极': 0, '中性': 1, '积极': 2}, 'data_dir': 'data/data9740', 'vocabulary': 'vocab.txt', 'train_file_list': 'train.txt', 'label_file': 'label_list.txt', 'save_model_dir': './train_model', 'save_freeze_dir': './infer_model', 'pretrained_model_dir': './pretrained_model', 'continue_train': False, 'pretrained': False, 'model_type': 'lstm_net', 'num_epochs': 12, 'train_batch_size': 128, 'use_gpu': True, 'sample_frequency': 10, 'rsm_strategy': {'learning_rate': 0.001, 'lr_epochs': [6, 8, 10], 'lr_decay': [1, 0.5, 0.25, 0.1]}, 'momentum_strategy': {'learning_rate': 0.001, 'lr_epochs': [6, 8, 10], 'lr_decay': [1, 0.5, 0.25, 0.1]}, 'sgd_strategy': {'learning_rate': 0.001, 'lr_epochs': [6, 8, 10], 'lr_decay': [1, 0.5, 0.25, 0.1]}, 'adam_strategy': {'learning_rate': 0.001}} 2021-08-02 23:45:30,782-INFO: create place, use gpu: True 2021-08-02 23:45:30,782 - [10189] [line:134] - INFO: create place, use gpu: True 2021-08-02 23:45:31,213-INFO: define input data tensor 2021-08-02 23:45:31,213 - [10189] [line:139] - INFO: define input data tensor 2021-08-02 23:45:31,214-INFO: build custom reader 2021-08-02 23:45:31,214 - [10189] [line:144] - INFO: build custom reader 2021-08-02 23:45:31,235-INFO: build network lstm_net 2021-08-02 23:45:31,235 - [10189] [line:147] - INFO: build network lstm_net 2021-08-02 23:45:31,250-INFO: build optimizer 2021-08-02 23:45:31,250 - [10189] [line:164] - INFO: build optimizer W0802 23:45:31.268640 10189 device_context.cc:252] Please NOTE: device: 0, CUDA Capability: 70, Driver API Version: 10.1, Runtime API Version: 9.0 W0802 23:45:31.273340 10189 device_context.cc:260] device: 0, cuDNN Version: 7.6. 2021-08-02 23:45:45,093-INFO: current pass: 0, start read text 2021-08-02 23:45:45,093 - [10189] [line:179] - INFO: current pass: 0, start read text 2021-08-02 23:45:45,141-INFO: pass 0, batch 0, accuracy:0.21875 loss:1.0986751317977905 time:0.04 sec 2021-08-02 23:45:45,141 - [10189] [line:200] - INFO: pass 0, batch 0, accuracy:0.21875 loss:1.0986751317977905 time:0.04 sec 2021-08-02 23:45:45,362-INFO: pass 0, batch 10, accuracy:0.7265625 loss:1.0889456272125244 time:0.02 sec 2021-08-02 23:45:45,362 - [10189] [line:200] - INFO: pass 0, batch 10, accuracy:0.7265625 loss:1.0889456272125244 time:0.02 sec 2021-08-02 23:45:45,579-INFO: pass 0, batch 20, accuracy:0.734375 loss:1.0790202617645264 time:0.02 sec 2021-08-02 23:45:45,579 - [10189] [line:200] - INFO: pass 0, batch 20, accuracy:0.734375 loss:1.0790202617645264 time:0.02 sec 2021-08-02 23:45:45,805-INFO: pass 0, batch 30, accuracy:0.7734375 loss:1.0675158500671387 time:0.02 sec 2021-08-02 23:45:45,805 - [10189] [line:200] - INFO: pass 0, batch 30, accuracy:0.7734375 loss:1.0675158500671387 time:0.02 sec 2021-08-02 23:45:46,027-INFO: pass 0, batch 40, accuracy:0.7421875 loss:1.0600175857543945 time:0.02 sec 2021-08-02 23:45:46,027 - [10189] [line:200] - INFO: pass 0, batch 40, accuracy:0.7421875 loss:1.0600175857543945 time:0.02 sec 2021-08-02 23:45:46,251-INFO: pass 0, batch 50, accuracy:0.75 loss:1.0503298044204712 time:0.02 sec 2021-08-02 23:45:46,251 - [10189] [line:200] - INFO: pass 0, batch 50, accuracy:0.75 loss:1.0503298044204712 time:0.02 sec 2021-08-02 23:45:46,485-INFO: pass 0, batch 60, accuracy:0.7109375 loss:1.0472145080566406 time:0.02 sec 2021-08-02 23:45:46,485 - [10189] [line:200] - INFO: pass 0, batch 60, accuracy:0.7109375 loss:1.0472145080566406 time:0.02 sec 2021-08-02 23:45:46,708-INFO: pass 0, batch 70, accuracy:0.703125 loss:1.0409847497940063 time:0.02 sec 2021-08-02 23:45:46,708 - [10189] [line:200] - INFO: pass 0, batch 70, accuracy:0.703125 loss:1.0409847497940063 time:0.02 sec 2021-08-02 23:45:46,820-INFO: pass 0 train result, current pass mean accuracy:0.7376027962094859 loss:1.062960325103057 2021-08-02 23:45:46,820 - [10189] [line:207] - INFO: pass 0 train result, current pass mean accuracy:0.7376027962094859 loss:1.062960325103057 2021-08-02 23:45:46,820-INFO: temp save pass 0 train result, current best pass mean accuracy:0.7376027962094859 2021-08-02 23:45:46,820 - [10189] [line:209] - INFO: temp save pass 0 train result, current best pass mean accuracy:0.7376027962094859 2021-08-02 23:45:47,480-INFO: current pass: 1, start read text 2021-08-02 23:45:47,480 - [10189] [line:179] - INFO: current pass: 1, start read text 2021-08-02 23:45:47,508-INFO: pass 1, batch 0, accuracy:0.7265625 loss:1.0324511528015137 time:0.03 sec 2021-08-02 23:45:47,508 - [10189] [line:200] - INFO: pass 1, batch 0, accuracy:0.7265625 loss:1.0324511528015137 time:0.03 sec 2021-08-02 23:45:47,732-INFO: pass 1, batch 10, accuracy:0.8125 loss:1.0077426433563232 time:0.02 sec 2021-08-02 23:45:47,732 - [10189] [line:200] - INFO: pass 1, batch 10, accuracy:0.8125 loss:1.0077426433563232 time:0.02 sec 2021-08-02 23:45:47,956-INFO: pass 1, batch 20, accuracy:0.7421875 loss:1.0145409107208252 time:0.02 sec 2021-08-02 23:45:47,956 - [10189] [line:200] - INFO: pass 1, batch 20, accuracy:0.7421875 loss:1.0145409107208252 time:0.02 sec 2021-08-02 23:45:48,178-INFO: pass 1, batch 30, accuracy:0.7421875 loss:1.0069184303283691 time:0.02 sec 2021-08-02 23:45:48,178 - [10189] [line:200] - INFO: pass 1, batch 30, accuracy:0.7421875 loss:1.0069184303283691 time:0.02 sec 2021-08-02 23:45:48,400-INFO: pass 1, batch 40, accuracy:0.7265625 loss:1.0005524158477783 time:0.02 sec 2021-08-02 23:45:48,400 - [10189] [line:200] - INFO: pass 1, batch 40, accuracy:0.7265625 loss:1.0005524158477783 time:0.02 sec 2021-08-02 23:45:48,625-INFO: pass 1, batch 50, accuracy:0.7109375 loss:0.9986919164657593 time:0.02 sec 2021-08-02 23:45:48,625 - [10189] [line:200] - INFO: pass 1, batch 50, accuracy:0.7109375 loss:0.9986919164657593 time:0.02 sec 2021-08-02 23:45:48,849-INFO: pass 1, batch 60, accuracy:0.7265625 loss:0.9898288249969482 time:0.02 sec 2021-08-02 23:45:48,849 - [10189] [line:200] - INFO: pass 1, batch 60, accuracy:0.7265625 loss:0.9898288249969482 time:0.02 sec 2021-08-02 23:45:49,082-INFO: pass 1, batch 70, accuracy:0.7265625 loss:0.983740508556366 time:0.02 sec 2021-08-02 23:45:49,082 - [10189] [line:200] - INFO: pass 1, batch 70, accuracy:0.7265625 loss:0.983740508556366 time:0.02 sec 2021-08-02 23:45:49,192-INFO: pass 1 train result, current pass mean accuracy:0.7445947967077556 loss:1.00030901165385 2021-08-02 23:45:49,192 - [10189] [line:207] - INFO: pass 1 train result, current pass mean accuracy:0.7445947967077556 loss:1.00030901165385 2021-08-02 23:45:49,192-INFO: temp save pass 1 train result, current best pass mean accuracy:0.7445947967077556 2021-08-02 23:45:49,192 - [10189] [line:209] - INFO: temp save pass 1 train result, current best pass mean accuracy:0.7445947967077556 2021-08-02 23:45:49,825-INFO: current pass: 2, start read text 2021-08-02 23:45:49,825 - [10189] [line:179] - INFO: current pass: 2, start read text 2021-08-02 23:45:49,849-INFO: pass 2, batch 0, accuracy:0.765625 loss:0.9668731689453125 time:0.02 sec 2021-08-02 23:45:49,849 - [10189] [line:200] - INFO: pass 2, batch 0, accuracy:0.765625 loss:0.9668731689453125 time:0.02 sec 2021-08-02 23:45:50,073-INFO: pass 2, batch 10, accuracy:0.796875 loss:0.9485573172569275 time:0.02 sec 2021-08-02 23:45:50,073 - [10189] [line:200] - INFO: pass 2, batch 10, accuracy:0.796875 loss:0.9485573172569275 time:0.02 sec 2021-08-02 23:45:50,302-INFO: pass 2, batch 20, accuracy:0.6875 loss:0.9810523986816406 time:0.03 sec 2021-08-02 23:45:50,302 - [10189] [line:200] - INFO: pass 2, batch 20, accuracy:0.6875 loss:0.9810523986816406 time:0.03 sec 2021-08-02 23:45:50,528-INFO: pass 2, batch 30, accuracy:0.7890625 loss:0.9362427592277527 time:0.02 sec 2021-08-02 23:45:50,528 - [10189] [line:200] - INFO: pass 2, batch 30, accuracy:0.7890625 loss:0.9362427592277527 time:0.02 sec 2021-08-02 23:45:50,752-INFO: pass 2, batch 40, accuracy:0.7578125 loss:0.9429517984390259 time:0.02 sec 2021-08-02 23:45:50,752 - [10189] [line:200] - INFO: pass 2, batch 40, accuracy:0.7578125 loss:0.9429517984390259 time:0.02 sec 2021-08-02 23:45:50,975-INFO: pass 2, batch 50, accuracy:0.7421875 loss:0.9435610771179199 time:0.02 sec 2021-08-02 23:45:50,975 - [10189] [line:200] - INFO: pass 2, batch 50, accuracy:0.7421875 loss:0.9435610771179199 time:0.02 sec 2021-08-02 23:45:51,199-INFO: pass 2, batch 60, accuracy:0.7265625 loss:0.9423719644546509 time:0.02 sec 2021-08-02 23:45:51,199 - [10189] [line:200] - INFO: pass 2, batch 60, accuracy:0.7265625 loss:0.9423719644546509 time:0.02 sec 2021-08-02 23:45:51,425-INFO: pass 2, batch 70, accuracy:0.6953125 loss:0.9496077299118042 time:0.02 sec 2021-08-02 23:45:51,425 - [10189] [line:200] - INFO: pass 2, batch 70, accuracy:0.6953125 loss:0.9496077299118042 time:0.02 sec 2021-08-02 23:45:51,535-INFO: pass 2 train result, current pass mean accuracy:0.7445947967077556 loss:0.9494063187586633 2021-08-02 23:45:51,535 - [10189] [line:207] - INFO: pass 2 train result, current pass mean accuracy:0.7445947967077556 loss:0.9494063187586633 2021-08-02 23:45:51,536-INFO: current pass: 3, start read text 2021-08-02 23:45:51,536 - [10189] [line:179] - INFO: current pass: 3, start read text 2021-08-02 23:45:51,559-INFO: pass 3, batch 0, accuracy:0.6796875 loss:0.9555757641792297 time:0.02 sec 2021-08-02 23:45:51,559 - [10189] [line:200] - INFO: pass 3, batch 0, accuracy:0.6796875 loss:0.9555757641792297 time:0.02 sec 2021-08-02 23:45:51,787-INFO: pass 3, batch 10, accuracy:0.7109375 loss:0.9358402490615845 time:0.02 sec 2021-08-02 23:45:51,787 - [10189] [line:200] - INFO: pass 3, batch 10, accuracy:0.7109375 loss:0.9358402490615845 time:0.02 sec 2021-08-02 23:45:52,011-INFO: pass 3, batch 20, accuracy:0.7578125 loss:0.9086047410964966 time:0.02 sec 2021-08-02 23:45:52,011 - [10189] [line:200] - INFO: pass 3, batch 20, accuracy:0.7578125 loss:0.9086047410964966 time:0.02 sec 2021-08-02 23:45:52,236-INFO: pass 3, batch 30, accuracy:0.7265625 loss:0.9203474521636963 time:0.02 sec 2021-08-02 23:45:52,236 - [10189] [line:200] - INFO: pass 3, batch 30, accuracy:0.7265625 loss:0.9203474521636963 time:0.02 sec 2021-08-02 23:45:52,466-INFO: pass 3, batch 40, accuracy:0.7890625 loss:0.8803378939628601 time:0.02 sec 2021-08-02 23:45:52,466 - [10189] [line:200] - INFO: pass 3, batch 40, accuracy:0.7890625 loss:0.8803378939628601 time:0.02 sec 2021-08-02 23:45:52,690-INFO: pass 3, batch 50, accuracy:0.765625 loss:0.8890682458877563 time:0.02 sec 2021-08-02 23:45:52,690 - [10189] [line:200] - INFO: pass 3, batch 50, accuracy:0.765625 loss:0.8890682458877563 time:0.02 sec 2021-08-02 23:45:52,911-INFO: pass 3, batch 60, accuracy:0.7578125 loss:0.8886351585388184 time:0.02 sec 2021-08-02 23:45:52,911 - [10189] [line:200] - INFO: pass 3, batch 60, accuracy:0.7578125 loss:0.8886351585388184 time:0.02 sec 2021-08-02 23:45:53,134-INFO: pass 3, batch 70, accuracy:0.703125 loss:0.9183830618858337 time:0.02 sec 2021-08-02 23:45:53,134 - [10189] [line:200] - INFO: pass 3, batch 70, accuracy:0.703125 loss:0.9183830618858337 time:0.02 sec 2021-08-02 23:45:53,240-INFO: pass 3 train result, current pass mean accuracy:0.7443219195855292 loss:0.9082369718112444 2021-08-02 23:45:53,240 - [10189] [line:207] - INFO: pass 3 train result, current pass mean accuracy:0.7443219195855292 loss:0.9082369718112444 2021-08-02 23:45:53,240-INFO: current pass: 4, start read text 2021-08-02 23:45:53,240 - [10189] [line:179] - INFO: current pass: 4, start read text 2021-08-02 23:45:53,263-INFO: pass 4, batch 0, accuracy:0.7578125 loss:0.8805445432662964 time:0.02 sec 2021-08-02 23:45:53,263 - [10189] [line:200] - INFO: pass 4, batch 0, accuracy:0.7578125 loss:0.8805445432662964 time:0.02 sec 2021-08-02 23:45:53,486-INFO: pass 4, batch 10, accuracy:0.7578125 loss:0.879603385925293 time:0.02 sec 2021-08-02 23:45:53,486 - [10189] [line:200] - INFO: pass 4, batch 10, accuracy:0.7578125 loss:0.879603385925293 time:0.02 sec 2021-08-02 23:45:53,709-INFO: pass 4, batch 20, accuracy:0.71875 loss:0.8979922533035278 time:0.02 sec 2021-08-02 23:45:53,709 - [10189] [line:200] - INFO: pass 4, batch 20, accuracy:0.71875 loss:0.8979922533035278 time:0.02 sec 2021-08-02 23:45:53,935-INFO: pass 4, batch 30, accuracy:0.71875 loss:0.8960754871368408 time:0.02 sec 2021-08-02 23:45:53,935 - [10189] [line:200] - INFO: pass 4, batch 30, accuracy:0.71875 loss:0.8960754871368408 time:0.02 sec 2021-08-02 23:45:54,159-INFO: pass 4, batch 40, accuracy:0.7578125 loss:0.8642590045928955 time:0.02 sec 2021-08-02 23:45:54,159 - [10189] [line:200] - INFO: pass 4, batch 40, accuracy:0.7578125 loss:0.8642590045928955 time:0.02 sec 2021-08-02 23:45:54,381-INFO: pass 4, batch 50, accuracy:0.7734375 loss:0.8500072956085205 time:0.02 sec 2021-08-02 23:45:54,381 - [10189] [line:200] - INFO: pass 4, batch 50, accuracy:0.7734375 loss:0.8500072956085205 time:0.02 sec 2021-08-02 23:45:54,608-INFO: pass 4, batch 60, accuracy:0.734375 loss:0.8738957643508911 time:0.02 sec 2021-08-02 23:45:54,608 - [10189] [line:200] - INFO: pass 4, batch 60, accuracy:0.734375 loss:0.8738957643508911 time:0.02 sec 2021-08-02 23:45:54,828-INFO: pass 4, batch 70, accuracy:0.7421875 loss:0.864019513130188 time:0.02 sec 2021-08-02 23:45:54,828 - [10189] [line:200] - INFO: pass 4, batch 70, accuracy:0.7421875 loss:0.864019513130188 time:0.02 sec 2021-08-02 23:45:54,941-INFO: pass 4 train result, current pass mean accuracy:0.7439126042943252 loss:0.8751089682704524 2021-08-02 23:45:54,941 - [10189] [line:207] - INFO: pass 4 train result, current pass mean accuracy:0.7439126042943252 loss:0.8751089682704524 2021-08-02 23:45:54,941-INFO: current pass: 5, start read text 2021-08-02 23:45:54,941 - [10189] [line:179] - INFO: current pass: 5, start read text 2021-08-02 23:45:54,964-INFO: pass 5, batch 0, accuracy:0.7265625 loss:0.8747999668121338 time:0.02 sec 2021-08-02 23:45:54,964 - [10189] [line:200] - INFO: pass 5, batch 0, accuracy:0.7265625 loss:0.8747999668121338 time:0.02 sec 2021-08-02 23:45:55,186-INFO: pass 5, batch 10, accuracy:0.7734375 loss:0.8366762399673462 time:0.02 sec 2021-08-02 23:45:55,186 - [10189] [line:200] - INFO: pass 5, batch 10, accuracy:0.7734375 loss:0.8366762399673462 time:0.02 sec 2021-08-02 23:45:55,418-INFO: pass 5, batch 20, accuracy:0.71875 loss:0.870479941368103 time:0.02 sec 2021-08-02 23:45:55,418 - [10189] [line:200] - INFO: pass 5, batch 20, accuracy:0.71875 loss:0.870479941368103 time:0.02 sec 2021-08-02 23:45:55,644-INFO: pass 5, batch 30, accuracy:0.7109375 loss:0.8715454936027527 time:0.02 sec 2021-08-02 23:45:55,644 - [10189] [line:200] - INFO: pass 5, batch 30, accuracy:0.7109375 loss:0.8715454936027527 time:0.02 sec 2021-08-02 23:45:55,866-INFO: pass 5, batch 40, accuracy:0.796875 loss:0.8111903667449951 time:0.02 sec 2021-08-02 23:45:55,866 - [10189] [line:200] - INFO: pass 5, batch 40, accuracy:0.796875 loss:0.8111903667449951 time:0.02 sec 2021-08-02 23:45:56,087-INFO: pass 5, batch 50, accuracy:0.671875 loss:0.8962817788124084 time:0.02 sec 2021-08-02 23:45:56,087 - [10189] [line:200] - INFO: pass 5, batch 50, accuracy:0.671875 loss:0.8962817788124084 time:0.02 sec 2021-08-02 23:45:56,313-INFO: pass 5, batch 60, accuracy:0.75 loss:0.8358069658279419 time:0.02 sec 2021-08-02 23:45:56,313 - [10189] [line:200] - INFO: pass 5, batch 60, accuracy:0.75 loss:0.8358069658279419 time:0.02 sec 2021-08-02 23:45:56,537-INFO: pass 5, batch 70, accuracy:0.765625 loss:0.8253481984138489 time:0.02 sec 2021-08-02 23:45:56,537 - [10189] [line:200] - INFO: pass 5, batch 70, accuracy:0.765625 loss:0.8253481984138489 time:0.02 sec 2021-08-02 23:45:56,649-INFO: pass 5 train result, current pass mean accuracy:0.7448676738299822 loss:0.8475426709965656 2021-08-02 23:45:56,649 - [10189] [line:207] - INFO: pass 5 train result, current pass mean accuracy:0.7448676738299822 loss:0.8475426709965656 2021-08-02 23:45:56,649-INFO: temp save pass 5 train result, current best pass mean accuracy:0.7448676738299822 2021-08-02 23:45:56,649 - [10189] [line:209] - INFO: temp save pass 5 train result, current best pass mean accuracy:0.7448676738299822 2021-08-02 23:45:57,281-INFO: current pass: 6, start read text 2021-08-02 23:45:57,281 - [10189] [line:179] - INFO: current pass: 6, start read text 2021-08-02 23:45:57,303-INFO: pass 6, batch 0, accuracy:0.7421875 loss:0.8358213305473328 time:0.02 sec 2021-08-02 23:45:57,303 - [10189] [line:200] - INFO: pass 6, batch 0, accuracy:0.7421875 loss:0.8358213305473328 time:0.02 sec 2021-08-02 23:45:57,520-INFO: pass 6, batch 10, accuracy:0.7734375 loss:0.808340847492218 time:0.02 sec 2021-08-02 23:45:57,520 - [10189] [line:200] - INFO: pass 6, batch 10, accuracy:0.7734375 loss:0.808340847492218 time:0.02 sec 2021-08-02 23:45:57,744-INFO: pass 6, batch 20, accuracy:0.7890625 loss:0.7933476567268372 time:0.02 sec 2021-08-02 23:45:57,744 - [10189] [line:200] - INFO: pass 6, batch 20, accuracy:0.7890625 loss:0.7933476567268372 time:0.02 sec 2021-08-02 23:45:57,965-INFO: pass 6, batch 30, accuracy:0.75 loss:0.8284498453140259 time:0.02 sec 2021-08-02 23:45:57,965 - [10189] [line:200] - INFO: pass 6, batch 30, accuracy:0.75 loss:0.8284498453140259 time:0.02 sec 2021-08-02 23:45:58,198-INFO: pass 6, batch 40, accuracy:0.765625 loss:0.8153940439224243 time:0.02 sec 2021-08-02 23:45:58,198 - [10189] [line:200] - INFO: pass 6, batch 40, accuracy:0.765625 loss:0.8153940439224243 time:0.02 sec 2021-08-02 23:45:58,422-INFO: pass 6, batch 50, accuracy:0.6796875 loss:0.8849328756332397 time:0.02 sec 2021-08-02 23:45:58,422 - [10189] [line:200] - INFO: pass 6, batch 50, accuracy:0.6796875 loss:0.8849328756332397 time:0.02 sec 2021-08-02 23:45:58,647-INFO: pass 6, batch 60, accuracy:0.7734375 loss:0.8036515712738037 time:0.02 sec 2021-08-02 23:45:58,647 - [10189] [line:200] - INFO: pass 6, batch 60, accuracy:0.7734375 loss:0.8036515712738037 time:0.02 sec 2021-08-02 23:45:58,870-INFO: pass 6, batch 70, accuracy:0.734375 loss:0.829437255859375 time:0.02 sec 2021-08-02 23:45:58,870 - [10189] [line:200] - INFO: pass 6, batch 70, accuracy:0.734375 loss:0.829437255859375 time:0.02 sec 2021-08-02 23:45:58,979-INFO: pass 6 train result, current pass mean accuracy:0.7448676738299822 loss:0.831356417191656 2021-08-02 23:45:58,979 - [10189] [line:207] - INFO: pass 6 train result, current pass mean accuracy:0.7448676738299822 loss:0.831356417191656 2021-08-02 23:45:58,979-INFO: current pass: 7, start read text 2021-08-02 23:45:58,979 - [10189] [line:179] - INFO: current pass: 7, start read text 2021-08-02 23:45:59,001-INFO: pass 7, batch 0, accuracy:0.6953125 loss:0.868660569190979 time:0.02 sec 2021-08-02 23:45:59,001 - [10189] [line:200] - INFO: pass 7, batch 0, accuracy:0.6953125 loss:0.868660569190979 time:0.02 sec 2021-08-02 23:45:59,220-INFO: pass 7, batch 10, accuracy:0.7734375 loss:0.7964760065078735 time:0.02 sec 2021-08-02 23:45:59,220 - [10189] [line:200] - INFO: pass 7, batch 10, accuracy:0.7734375 loss:0.7964760065078735 time:0.02 sec 2021-08-02 23:45:59,442-INFO: pass 7, batch 20, accuracy:0.7265625 loss:0.8406094312667847 time:0.02 sec 2021-08-02 23:45:59,442 - [10189] [line:200] - INFO: pass 7, batch 20, accuracy:0.7265625 loss:0.8406094312667847 time:0.02 sec 2021-08-02 23:45:59,666-INFO: pass 7, batch 30, accuracy:0.7578125 loss:0.8141382932662964 time:0.02 sec 2021-08-02 23:45:59,666 - [10189] [line:200] - INFO: pass 7, batch 30, accuracy:0.7578125 loss:0.8141382932662964 time:0.02 sec 2021-08-02 23:45:59,889-INFO: pass 7, batch 40, accuracy:0.7109375 loss:0.8532629609107971 time:0.02 sec 2021-08-02 23:45:59,889 - [10189] [line:200] - INFO: pass 7, batch 40, accuracy:0.7109375 loss:0.8532629609107971 time:0.02 sec 2021-08-02 23:46:00,112-INFO: pass 7, batch 50, accuracy:0.7890625 loss:0.7848094701766968 time:0.02 sec 2021-08-02 23:46:00,112 - [10189] [line:200] - INFO: pass 7, batch 50, accuracy:0.7890625 loss:0.7848094701766968 time:0.02 sec 2021-08-02 23:46:00,344-INFO: pass 7, batch 60, accuracy:0.78125 loss:0.7815483212471008 time:0.02 sec 2021-08-02 23:46:00,344 - [10189] [line:200] - INFO: pass 7, batch 60, accuracy:0.78125 loss:0.7815483212471008 time:0.02 sec 2021-08-02 23:46:00,570-INFO: pass 7, batch 70, accuracy:0.75 loss:0.8120790123939514 time:0.02 sec 2021-08-02 23:46:00,570 - [10189] [line:200] - INFO: pass 7, batch 70, accuracy:0.75 loss:0.8120790123939514 time:0.02 sec 2021-08-02 23:46:00,681-INFO: pass 7 train result, current pass mean accuracy:0.7443219195855292 loss:0.8220906955631155 2021-08-02 23:46:00,681 - [10189] [line:207] - INFO: pass 7 train result, current pass mean accuracy:0.7443219195855292 loss:0.8220906955631155 2021-08-02 23:46:00,681-INFO: current pass: 8, start read text 2021-08-02 23:46:00,681 - [10189] [line:179] - INFO: current pass: 8, start read text 2021-08-02 23:46:00,704-INFO: pass 8, batch 0, accuracy:0.8359375 loss:0.7352373600006104 time:0.02 sec 2021-08-02 23:46:00,704 - [10189] [line:200] - INFO: pass 8, batch 0, accuracy:0.8359375 loss:0.7352373600006104 time:0.02 sec 2021-08-02 23:46:00,925-INFO: pass 8, batch 10, accuracy:0.6953125 loss:0.8599326610565186 time:0.02 sec 2021-08-02 23:46:00,925 - [10189] [line:200] - INFO: pass 8, batch 10, accuracy:0.6953125 loss:0.8599326610565186 time:0.02 sec 2021-08-02 23:46:01,149-INFO: pass 8, batch 20, accuracy:0.734375 loss:0.8253572583198547 time:0.02 sec 2021-08-02 23:46:01,149 - [10189] [line:200] - INFO: pass 8, batch 20, accuracy:0.734375 loss:0.8253572583198547 time:0.02 sec 2021-08-02 23:46:01,381-INFO: pass 8, batch 30, accuracy:0.7109375 loss:0.8464549779891968 time:0.02 sec 2021-08-02 23:46:01,381 - [10189] [line:200] - INFO: pass 8, batch 30, accuracy:0.7109375 loss:0.8464549779891968 time:0.02 sec 2021-08-02 23:46:01,604-INFO: pass 8, batch 40, accuracy:0.7734375 loss:0.7885726690292358 time:0.02 sec 2021-08-02 23:46:01,604 - [10189] [line:200] - INFO: pass 8, batch 40, accuracy:0.7734375 loss:0.7885726690292358 time:0.02 sec 2021-08-02 23:46:01,828-INFO: pass 8, batch 50, accuracy:0.71875 loss:0.8387879133224487 time:0.02 sec 2021-08-02 23:46:01,828 - [10189] [line:200] - INFO: pass 8, batch 50, accuracy:0.71875 loss:0.8387879133224487 time:0.02 sec 2021-08-02 23:46:02,107-INFO: pass 8, batch 60, accuracy:0.7265625 loss:0.8361030220985413 time:0.02 sec 2021-08-02 23:46:02,107 - [10189] [line:200] - INFO: pass 8, batch 60, accuracy:0.7265625 loss:0.8361030220985413 time:0.02 sec 2021-08-02 23:46:02,329-INFO: pass 8, batch 70, accuracy:0.7421875 loss:0.8175181746482849 time:0.02 sec 2021-08-02 23:46:02,329 - [10189] [line:200] - INFO: pass 8, batch 70, accuracy:0.7421875 loss:0.8175181746482849 time:0.02 sec 2021-08-02 23:46:02,436-INFO: pass 8 train result, current pass mean accuracy:0.7448676738299822 loss:0.81523066523828 2021-08-02 23:46:02,436 - [10189] [line:207] - INFO: pass 8 train result, current pass mean accuracy:0.7448676738299822 loss:0.81523066523828 2021-08-02 23:46:02,437-INFO: current pass: 9, start read text 2021-08-02 23:46:02,437 - [10189] [line:179] - INFO: current pass: 9, start read text 2021-08-02 23:46:02,459-INFO: pass 9, batch 0, accuracy:0.7421875 loss:0.813361406326294 time:0.02 sec 2021-08-02 23:46:02,459 - [10189] [line:200] - INFO: pass 9, batch 0, accuracy:0.7421875 loss:0.813361406326294 time:0.02 sec 2021-08-02 23:46:02,679-INFO: pass 9, batch 10, accuracy:0.703125 loss:0.8470901846885681 time:0.02 sec 2021-08-02 23:46:02,679 - [10189] [line:200] - INFO: pass 9, batch 10, accuracy:0.703125 loss:0.8470901846885681 time:0.02 sec 2021-08-02 23:46:02,904-INFO: pass 9, batch 20, accuracy:0.7421875 loss:0.8168234825134277 time:0.02 sec 2021-08-02 23:46:02,904 - [10189] [line:200] - INFO: pass 9, batch 20, accuracy:0.7421875 loss:0.8168234825134277 time:0.02 sec 2021-08-02 23:46:03,128-INFO: pass 9, batch 30, accuracy:0.8203125 loss:0.7440495491027832 time:0.02 sec 2021-08-02 23:46:03,128 - [10189] [line:200] - INFO: pass 9, batch 30, accuracy:0.8203125 loss:0.7440495491027832 time:0.02 sec 2021-08-02 23:46:03,360-INFO: pass 9, batch 40, accuracy:0.765625 loss:0.7905645966529846 time:0.02 sec 2021-08-02 23:46:03,360 - [10189] [line:200] - INFO: pass 9, batch 40, accuracy:0.765625 loss:0.7905645966529846 time:0.02 sec 2021-08-02 23:46:03,583-INFO: pass 9, batch 50, accuracy:0.6953125 loss:0.8578976988792419 time:0.02 sec 2021-08-02 23:46:03,583 - [10189] [line:200] - INFO: pass 9, batch 50, accuracy:0.6953125 loss:0.8578976988792419 time:0.02 sec 2021-08-02 23:46:03,813-INFO: pass 9, batch 60, accuracy:0.75 loss:0.8052486181259155 time:0.02 sec 2021-08-02 23:46:03,813 - [10189] [line:200] - INFO: pass 9, batch 60, accuracy:0.75 loss:0.8052486181259155 time:0.02 sec 2021-08-02 23:46:04,033-INFO: pass 9, batch 70, accuracy:0.6953125 loss:0.8516145944595337 time:0.02 sec 2021-08-02 23:46:04,033 - [10189] [line:200] - INFO: pass 9, batch 70, accuracy:0.6953125 loss:0.8516145944595337 time:0.02 sec 2021-08-02 23:46:04,141-INFO: pass 9 train result, current pass mean accuracy:0.7450041119989596 loss:0.8109210525688372 2021-08-02 23:46:04,141 - [10189] [line:207] - INFO: pass 9 train result, current pass mean accuracy:0.7450041119989596 loss:0.8109210525688372 2021-08-02 23:46:04,141-INFO: temp save pass 9 train result, current best pass mean accuracy:0.7450041119989596 2021-08-02 23:46:04,141 - [10189] [line:209] - INFO: temp save pass 9 train result, current best pass mean accuracy:0.7450041119989596 2021-08-02 23:46:04,774-INFO: current pass: 10, start read text 2021-08-02 23:46:04,774 - [10189] [line:179] - INFO: current pass: 10, start read text 2021-08-02 23:46:04,806-INFO: pass 10, batch 0, accuracy:0.78125 loss:0.7753205895423889 time:0.03 sec 2021-08-02 23:46:04,806 - [10189] [line:200] - INFO: pass 10, batch 0, accuracy:0.78125 loss:0.7753205895423889 time:0.03 sec 2021-08-02 23:46:05,039-INFO: pass 10, batch 10, accuracy:0.7265625 loss:0.8241345882415771 time:0.02 sec 2021-08-02 23:46:05,039 - [10189] [line:200] - INFO: pass 10, batch 10, accuracy:0.7265625 loss:0.8241345882415771 time:0.02 sec 2021-08-02 23:46:05,263-INFO: pass 10, batch 20, accuracy:0.765625 loss:0.7888714671134949 time:0.02 sec 2021-08-02 23:46:05,263 - [10189] [line:200] - INFO: pass 10, batch 20, accuracy:0.765625 loss:0.7888714671134949 time:0.02 sec 2021-08-02 23:46:05,482-INFO: pass 10, batch 30, accuracy:0.703125 loss:0.8527122735977173 time:0.02 sec 2021-08-02 23:46:05,482 - [10189] [line:200] - INFO: pass 10, batch 30, accuracy:0.703125 loss:0.8527122735977173 time:0.02 sec 2021-08-02 23:46:05,705-INFO: pass 10, batch 40, accuracy:0.65625 loss:0.8972397446632385 time:0.02 sec 2021-08-02 23:46:05,705 - [10189] [line:200] - INFO: pass 10, batch 40, accuracy:0.65625 loss:0.8972397446632385 time:0.02 sec 2021-08-02 23:46:05,925-INFO: pass 10, batch 50, accuracy:0.7890625 loss:0.7625322341918945 time:0.02 sec 2021-08-02 23:46:05,925 - [10189] [line:200] - INFO: pass 10, batch 50, accuracy:0.7890625 loss:0.7625322341918945 time:0.02 sec 2021-08-02 23:46:06,150-INFO: pass 10, batch 60, accuracy:0.71875 loss:0.8338291645050049 time:0.02 sec 2021-08-02 23:46:06,150 - [10189] [line:200] - INFO: pass 10, batch 60, accuracy:0.71875 loss:0.8338291645050049 time:0.02 sec 2021-08-02 23:46:06,374-INFO: pass 10, batch 70, accuracy:0.7109375 loss:0.8392796516418457 time:0.02 sec 2021-08-02 23:46:06,374 - [10189] [line:200] - INFO: pass 10, batch 70, accuracy:0.7109375 loss:0.8392796516418457 time:0.02 sec 2021-08-02 23:46:06,486-INFO: pass 10 train result, current pass mean accuracy:0.7444583585387782 loss:0.8088854480730859 2021-08-02 23:46:06,486 - [10189] [line:207] - INFO: pass 10 train result, current pass mean accuracy:0.7444583585387782 loss:0.8088854480730859 2021-08-02 23:46:06,486-INFO: current pass: 11, start read text 2021-08-02 23:46:06,486 - [10189] [line:179] - INFO: current pass: 11, start read text 2021-08-02 23:46:06,508-INFO: pass 11, batch 0, accuracy:0.75 loss:0.8019920587539673 time:0.02 sec 2021-08-02 23:46:06,508 - [10189] [line:200] - INFO: pass 11, batch 0, accuracy:0.75 loss:0.8019920587539673 time:0.02 sec 2021-08-02 23:46:06,732-INFO: pass 11, batch 10, accuracy:0.765625 loss:0.7893277406692505 time:0.02 sec 2021-08-02 23:46:06,732 - [10189] [line:200] - INFO: pass 11, batch 10, accuracy:0.765625 loss:0.7893277406692505 time:0.02 sec 2021-08-02 23:46:06,962-INFO: pass 11, batch 20, accuracy:0.7421875 loss:0.8125526905059814 time:0.02 sec 2021-08-02 23:46:06,962 - [10189] [line:200] - INFO: pass 11, batch 20, accuracy:0.7421875 loss:0.8125526905059814 time:0.02 sec 2021-08-02 23:46:07,188-INFO: pass 11, batch 30, accuracy:0.78125 loss:0.7728949189186096 time:0.03 sec 2021-08-02 23:46:07,188 - [10189] [line:200] - INFO: pass 11, batch 30, accuracy:0.78125 loss:0.7728949189186096 time:0.03 sec 2021-08-02 23:46:07,412-INFO: pass 11, batch 40, accuracy:0.765625 loss:0.7839987277984619 time:0.02 sec 2021-08-02 23:46:07,412 - [10189] [line:200] - INFO: pass 11, batch 40, accuracy:0.765625 loss:0.7839987277984619 time:0.02 sec 2021-08-02 23:46:07,637-INFO: pass 11, batch 50, accuracy:0.6875 loss:0.8568463921546936 time:0.02 sec 2021-08-02 23:46:07,637 - [10189] [line:200] - INFO: pass 11, batch 50, accuracy:0.6875 loss:0.8568463921546936 time:0.02 sec 2021-08-02 23:46:07,860-INFO: pass 11, batch 60, accuracy:0.6953125 loss:0.8541688323020935 time:0.02 sec 2021-08-02 23:46:07,860 - [10189] [line:200] - INFO: pass 11, batch 60, accuracy:0.6953125 loss:0.8541688323020935 time:0.02 sec 2021-08-02 23:46:08,085-INFO: pass 11, batch 70, accuracy:0.765625 loss:0.7836540937423706 time:0.02 sec 2021-08-02 23:46:08,085 - [10189] [line:200] - INFO: pass 11, batch 70, accuracy:0.765625 loss:0.7836540937423706 time:0.02 sec 2021-08-02 23:46:08,198-INFO: pass 11 train result, current pass mean accuracy:0.7441854814165517 loss:0.8075789707271677 2021-08-02 23:46:08,198 - [10189] [line:207] - INFO: pass 11 train result, current pass mean accuracy:0.7441854814165517 loss:0.8075789707271677 2021-08-02 23:46:08,198-INFO: training till last epcho, end training 2021-08-02 23:46:08,198 - [10189] [line:214] - INFO: training till last epcho, end training

3.3.5.2 固化训练参数

In [16]

!python freeze.py

2021-08-02 23:46:10,793-INFO: build network lstm_net to freeze 2021-08-02 23:46:10,793 - [10293] [line:46] - INFO: build network lstm_net to freeze 2021-08-02 23:46:11,884-INFO: freeze lstm_net end 2021-08-02 23:46:11,884 - [10293] [line:66] - INFO: freeze lstm_net end

3.3.5.3 验证训练结果

In [17]

!python eval.py

W0802 23:46:49.790235 10388 device_context.cc:252] Please NOTE: device: 0, CUDA Capability: 70, Driver API Version: 10.1, Runtime API Version: 9.0 W0802 23:46:49.795176 10388 device_context.cc:260] device: 0, cuDNN Version: 7.6. total eval count:1036 cost time:7.09 sec predict accuracy:0.750965250965251

3.3.5.4 使用模型进行预测

In [18]

!python infer.py

W0802 23:47:21.064400 10446 device_context.cc:252] Please NOTE: device: 0, CUDA Capability: 70, Driver API Version: 10.1, Runtime API Version: 9.0 W0802 23:47:21.069312 10446 device_context.cc:260] device: 0, cuDNN Version: 7.6. infer text:我 就是 喜欢 你的 知性 大方 温柔 infer result:1

3.3.6 Bi-LSTM模型的训练与预测

3.3.6.1 训练模型

更改train_parameters中"model_type"为"bilstm_net"

In [20]

!python train.py

2021-08-02 23:48:35,457-INFO: start train text classification, train params: {'class_dim': 3, 'sample_count': 9656, 'label_dict': {'消极': 0, '中性': 1, '积极': 2}, 'data_dir': 'data/data9740', 'vocabulary': 'vocab.txt', 'train_file_list': 'train.txt', 'label_file': 'label_list.txt', 'save_model_dir': './train_model', 'save_freeze_dir': './infer_model', 'pretrained_model_dir': './pretrained_model', 'continue_train': False, 'pretrained': False, 'model_type': 'bilstm_net', 'num_epochs': 12, 'train_batch_size': 128, 'use_gpu': True, 'sample_frequency': 10, 'rsm_strategy': {'learning_rate': 0.001, 'lr_epochs': [6, 8, 10], 'lr_decay': [1, 0.5, 0.25, 0.1]}, 'momentum_strategy': {'learning_rate': 0.001, 'lr_epochs': [6, 8, 10], 'lr_decay': [1, 0.5, 0.25, 0.1]}, 'sgd_strategy': {'learning_rate': 0.001, 'lr_epochs': [6, 8, 10], 'lr_decay': [1, 0.5, 0.25, 0.1]}, 'adam_strategy': {'learning_rate': 0.001}} 2021-08-02 23:48:35,457 - [10652] [line:133] - INFO: start train text classification, train params: {'class_dim': 3, 'sample_count': 9656, 'label_dict': {'消极': 0, '中性': 1, '积极': 2}, 'data_dir': 'data/data9740', 'vocabulary': 'vocab.txt', 'train_file_list': 'train.txt', 'label_file': 'label_list.txt', 'save_model_dir': './train_model', 'save_freeze_dir': './infer_model', 'pretrained_model_dir': './pretrained_model', 'continue_train': False, 'pretrained': False, 'model_type': 'bilstm_net', 'num_epochs': 12, 'train_batch_size': 128, 'use_gpu': True, 'sample_frequency': 10, 'rsm_strategy': {'learning_rate': 0.001, 'lr_epochs': [6, 8, 10], 'lr_decay': [1, 0.5, 0.25, 0.1]}, 'momentum_strategy': {'learning_rate': 0.001, 'lr_epochs': [6, 8, 10], 'lr_decay': [1, 0.5, 0.25, 0.1]}, 'sgd_strategy': {'learning_rate': 0.001, 'lr_epochs': [6, 8, 10], 'lr_decay': [1, 0.5, 0.25, 0.1]}, 'adam_strategy': {'learning_rate': 0.001}} 2021-08-02 23:48:35,457-INFO: create place, use gpu: True 2021-08-02 23:48:35,457 - [10652] [line:134] - INFO: create place, use gpu: True 2021-08-02 23:48:35,890-INFO: define input data tensor 2021-08-02 23:48:35,890 - [10652] [line:139] - INFO: define input data tensor 2021-08-02 23:48:35,891-INFO: build custom reader 2021-08-02 23:48:35,891 - [10652] [line:144] - INFO: build custom reader 2021-08-02 23:48:35,913-INFO: build network bilstm_net 2021-08-02 23:48:35,913 - [10652] [line:147] - INFO: build network bilstm_net 2021-08-02 23:48:35,931-INFO: build optimizer 2021-08-02 23:48:35,931 - [10652] [line:164] - INFO: build optimizer W0802 23:48:35.951720 10652 device_context.cc:252] Please NOTE: device: 0, CUDA Capability: 70, Driver API Version: 10.1, Runtime API Version: 9.0 W0802 23:48:35.956382 10652 device_context.cc:260] device: 0, cuDNN Version: 7.6. 2021-08-02 23:48:49,744-INFO: current pass: 0, start read text 2021-08-02 23:48:49,744 - [10652] [line:179] - INFO: current pass: 0, start read text 2021-08-02 23:48:49,800-INFO: pass 0, batch 0, accuracy:0.3359375 loss:1.09859037399292 time:0.05 sec 2021-08-02 23:48:49,800 - [10652] [line:200] - INFO: pass 0, batch 0, accuracy:0.3359375 loss:1.09859037399292 time:0.05 sec 2021-08-02 23:48:50,091-INFO: pass 0, batch 10, accuracy:0.8125 loss:1.088160753250122 time:0.02 sec 2021-08-02 23:48:50,091 - [10652] [line:200] - INFO: pass 0, batch 10, accuracy:0.8125 loss:1.088160753250122 time:0.02 sec 2021-08-02 23:48:50,412-INFO: pass 0, batch 20, accuracy:0.71875 loss:1.0807138681411743 time:0.03 sec 2021-08-02 23:48:50,412 - [10652] [line:200] - INFO: pass 0, batch 20, accuracy:0.71875 loss:1.0807138681411743 time:0.03 sec 2021-08-02 23:48:50,726-INFO: pass 0, batch 30, accuracy:0.703125 loss:1.0731867551803589 time:0.04 sec 2021-08-02 23:48:50,726 - [10652] [line:200] - INFO: pass 0, batch 30, accuracy:0.703125 loss:1.0731867551803589 time:0.04 sec 2021-08-02 23:48:51,027-INFO: pass 0, batch 40, accuracy:0.703125 loss:1.0651404857635498 time:0.03 sec 2021-08-02 23:48:51,027 - [10652] [line:200] - INFO: pass 0, batch 40, accuracy:0.703125 loss:1.0651404857635498 time:0.03 sec 2021-08-02 23:48:51,322-INFO: pass 0, batch 50, accuracy:0.7421875 loss:1.0528974533081055 time:0.03 sec 2021-08-02 23:48:51,322 - [10652] [line:200] - INFO: pass 0, batch 50, accuracy:0.7421875 loss:1.0528974533081055 time:0.03 sec 2021-08-02 23:48:51,641-INFO: pass 0, batch 60, accuracy:0.7109375 loss:1.0488063097000122 time:0.03 sec 2021-08-02 23:48:51,641 - [10652] [line:200] - INFO: pass 0, batch 60, accuracy:0.7109375 loss:1.0488063097000122 time:0.03 sec 2021-08-02 23:48:51,933-INFO: pass 0, batch 70, accuracy:0.75 loss:1.0359910726547241 time:0.02 sec 2021-08-02 23:48:51,933 - [10652] [line:200] - INFO: pass 0, batch 70, accuracy:0.75 loss:1.0359910726547241 time:0.02 sec 2021-08-02 23:48:52,083-INFO: pass 0 train result, current pass mean accuracy:0.7393166869878769 loss:1.0643240103596134 2021-08-02 23:48:52,083 - [10652] [line:207] - INFO: pass 0 train result, current pass mean accuracy:0.7393166869878769 loss:1.0643240103596134 2021-08-02 23:48:52,083-INFO: temp save pass 0 train result, current best pass mean accuracy:0.7393166869878769 2021-08-02 23:48:52,083 - [10652] [line:209] - INFO: temp save pass 0 train result, current best pass mean accuracy:0.7393166869878769 2021-08-02 23:48:52,684-INFO: current pass: 1, start read text 2021-08-02 23:48:52,684 - [10652] [line:179] - INFO: current pass: 1, start read text 2021-08-02 23:48:52,714-INFO: pass 1, batch 0, accuracy:0.7734375 loss:1.0274670124053955 time:0.03 sec 2021-08-02 23:48:52,714 - [10652] [line:200] - INFO: pass 1, batch 0, accuracy:0.7734375 loss:1.0274670124053955 time:0.03 sec 2021-08-02 23:48:53,032-INFO: pass 1, batch 10, accuracy:0.75 loss:1.0238920450210571 time:0.03 sec 2021-08-02 23:48:53,032 - [10652] [line:200] - INFO: pass 1, batch 10, accuracy:0.75 loss:1.0238920450210571 time:0.03 sec 2021-08-02 23:48:53,345-INFO: pass 1, batch 20, accuracy:0.8125 loss:1.0014402866363525 time:0.03 sec 2021-08-02 23:48:53,345 - [10652] [line:200] - INFO: pass 1, batch 20, accuracy:0.8125 loss:1.0014402866363525 time:0.03 sec 2021-08-02 23:48:53,637-INFO: pass 1, batch 30, accuracy:0.75 loss:1.00754976272583 time:0.03 sec 2021-08-02 23:48:53,637 - [10652] [line:200] - INFO: pass 1, batch 30, accuracy:0.75 loss:1.00754976272583 time:0.03 sec 2021-08-02 23:48:53,943-INFO: pass 1, batch 40, accuracy:0.765625 loss:0.9965028166770935 time:0.03 sec 2021-08-02 23:48:53,943 - [10652] [line:200] - INFO: pass 1, batch 40, accuracy:0.765625 loss:0.9965028166770935 time:0.03 sec 2021-08-02 23:48:54,248-INFO: pass 1, batch 50, accuracy:0.7734375 loss:0.9846490621566772 time:0.03 sec 2021-08-02 23:48:54,248 - [10652] [line:200] - INFO: pass 1, batch 50, accuracy:0.7734375 loss:0.9846490621566772 time:0.03 sec 2021-08-02 23:48:54,565-INFO: pass 1, batch 60, accuracy:0.75 loss:0.9861129522323608 time:0.03 sec 2021-08-02 23:48:54,565 - [10652] [line:200] - INFO: pass 1, batch 60, accuracy:0.75 loss:0.9861129522323608 time:0.03 sec 2021-08-02 23:48:54,863-INFO: pass 1, batch 70, accuracy:0.6953125 loss:0.9921696186065674 time:0.03 sec 2021-08-02 23:48:54,863 - [10652] [line:200] - INFO: pass 1, batch 70, accuracy:0.6953125 loss:0.9921696186065674 time:0.03 sec 2021-08-02 23:48:55,010-INFO: pass 1 train result, current pass mean accuracy:0.7440490432475743 loss:1.0033885460150869 2021-08-02 23:48:55,010 - [10652] [line:207] - INFO: pass 1 train result, current pass mean accuracy:0.7440490432475743 loss:1.0033885460150869 2021-08-02 23:48:55,010-INFO: temp save pass 1 train result, current best pass mean accuracy:0.7440490432475743 2021-08-02 23:48:55,010 - [10652] [line:209] - INFO: temp save pass 1 train result, current best pass mean accuracy:0.7440490432475743 2021-08-02 23:48:55,645-INFO: current pass: 2, start read text 2021-08-02 23:48:55,645 - [10652] [line:179] - INFO: current pass: 2, start read text 2021-08-02 23:48:55,686-INFO: pass 2, batch 0, accuracy:0.796875 loss:0.9592280387878418 time:0.04 sec 2021-08-02 23:48:55,686 - [10652] [line:200] - INFO: pass 2, batch 0, accuracy:0.796875 loss:0.9592280387878418 time:0.04 sec 2021-08-02 23:48:56,010-INFO: pass 2, batch 10, accuracy:0.734375 loss:0.9714004993438721 time:0.03 sec 2021-08-02 23:48:56,010 - [10652] [line:200] - INFO: pass 2, batch 10, accuracy:0.734375 loss:0.9714004993438721 time:0.03 sec 2021-08-02 23:48:56,334-INFO: pass 2, batch 20, accuracy:0.7578125 loss:0.9608726501464844 time:0.03 sec 2021-08-02 23:48:56,334 - [10652] [line:200] - INFO: pass 2, batch 20, accuracy:0.7578125 loss:0.9608726501464844 time:0.03 sec 2021-08-02 23:48:56,628-INFO: pass 2, batch 30, accuracy:0.75 loss:0.9566822052001953 time:0.03 sec 2021-08-02 23:48:56,628 - [10652] [line:200] - INFO: pass 2, batch 30, accuracy:0.75 loss:0.9566822052001953 time:0.03 sec 2021-08-02 23:48:56,936-INFO: pass 2, batch 40, accuracy:0.71875 loss:0.9620290398597717 time:0.02 sec 2021-08-02 23:48:56,936 - [10652] [line:200] - INFO: pass 2, batch 40, accuracy:0.71875 loss:0.9620290398597717 time:0.02 sec 2021-08-02 23:48:57,236-INFO: pass 2, batch 50, accuracy:0.6484375 loss:0.9851903319358826 time:0.03 sec 2021-08-02 23:48:57,236 - [10652] [line:200] - INFO: pass 2, batch 50, accuracy:0.6484375 loss:0.9851903319358826 time:0.03 sec 2021-08-02 23:48:57,545-INFO: pass 2, batch 60, accuracy:0.734375 loss:0.9441577196121216 time:0.03 sec 2021-08-02 23:48:57,545 - [10652] [line:200] - INFO: pass 2, batch 60, accuracy:0.734375 loss:0.9441577196121216 time:0.03 sec 2021-08-02 23:48:57,856-INFO: pass 2, batch 70, accuracy:0.796875 loss:0.9136397838592529 time:0.03 sec 2021-08-02 23:48:57,856 - [10652] [line:200] - INFO: pass 2, batch 70, accuracy:0.796875 loss:0.9136397838592529 time:0.03 sec 2021-08-02 23:48:58,011-INFO: pass 2 train result, current pass mean accuracy:0.7444583585387782 loss:0.9531475031062177 2021-08-02 23:48:58,011 - [10652] [line:207] - INFO: pass 2 train result, current pass mean accuracy:0.7444583585387782 loss:0.9531475031062177 2021-08-02 23:48:58,012-INFO: temp save pass 2 train result, current best pass mean accuracy:0.7444583585387782 2021-08-02 23:48:58,012 - [10652] [line:209] - INFO: temp save pass 2 train result, current best pass mean accuracy:0.7444583585387782 2021-08-02 23:48:58,648-INFO: current pass: 3, start read text 2021-08-02 23:48:58,648 - [10652] [line:179] - INFO: current pass: 3, start read text 2021-08-02 23:48:58,687-INFO: pass 3, batch 0, accuracy:0.7578125 loss:0.924994707107544 time:0.04 sec 2021-08-02 23:48:58,687 - [10652] [line:200] - INFO: pass 3, batch 0, accuracy:0.7578125 loss:0.924994707107544 time:0.04 sec 2021-08-02 23:48:58,971-INFO: pass 3, batch 10, accuracy:0.796875 loss:0.8998569846153259 time:0.02 sec 2021-08-02 23:48:58,971 - [10652] [line:200] - INFO: pass 3, batch 10, accuracy:0.796875 loss:0.8998569846153259 time:0.02 sec 2021-08-02 23:48:59,295-INFO: pass 3, batch 20, accuracy:0.7421875 loss:0.9239690899848938 time:0.03 sec 2021-08-02 23:48:59,295 - [10652] [line:200] - INFO: pass 3, batch 20, accuracy:0.7421875 loss:0.9239690899848938 time:0.03 sec 2021-08-02 23:48:59,629-INFO: pass 3, batch 30, accuracy:0.703125 loss:0.9341282844543457 time:0.03 sec 2021-08-02 23:48:59,629 - [10652] [line:200] - INFO: pass 3, batch 30, accuracy:0.703125 loss:0.9341282844543457 time:0.03 sec 2021-08-02 23:48:59,942-INFO: pass 3, batch 40, accuracy:0.75 loss:0.9094167947769165 time:0.03 sec 2021-08-02 23:48:59,942 - [10652] [line:200] - INFO: pass 3, batch 40, accuracy:0.75 loss:0.9094167947769165 time:0.03 sec 2021-08-02 23:49:00,239-INFO: pass 3, batch 50, accuracy:0.75 loss:0.9022434949874878 time:0.03 sec 2021-08-02 23:49:00,239 - [10652] [line:200] - INFO: pass 3, batch 50, accuracy:0.75 loss:0.9022434949874878 time:0.03 sec 2021-08-02 23:49:00,554-INFO: pass 3, batch 60, accuracy:0.78125 loss:0.8830137252807617 time:0.03 sec 2021-08-02 23:49:00,554 - [10652] [line:200] - INFO: pass 3, batch 60, accuracy:0.78125 loss:0.8830137252807617 time:0.03 sec 2021-08-02 23:49:00,853-INFO: pass 3, batch 70, accuracy:0.7578125 loss:0.8861168622970581 time:0.03 sec 2021-08-02 23:49:00,853 - [10652] [line:200] - INFO: pass 3, batch 70, accuracy:0.7578125 loss:0.8861168622970581 time:0.03 sec 2021-08-02 23:49:00,997-INFO: pass 3 train result, current pass mean accuracy:0.7447312348767331 loss:0.9118851274251938 2021-08-02 23:49:00,997 - [10652] [line:207] - INFO: pass 3 train result, current pass mean accuracy:0.7447312348767331 loss:0.9118851274251938 2021-08-02 23:49:00,997-INFO: temp save pass 3 train result, current best pass mean accuracy:0.7447312348767331 2021-08-02 23:49:00,997 - [10652] [line:209] - INFO: temp save pass 3 train result, current best pass mean accuracy:0.7447312348767331 2021-08-02 23:49:01,623-INFO: current pass: 4, start read text 2021-08-02 23:49:01,623 - [10652] [line:179] - INFO: current pass: 4, start read text 2021-08-02 23:49:01,660-INFO: pass 4, batch 0, accuracy:0.7421875 loss:0.8983062505722046 time:0.03 sec 2021-08-02 23:49:01,660 - [10652] [line:200] - INFO: pass 4, batch 0, accuracy:0.7421875 loss:0.8983062505722046 time:0.03 sec 2021-08-02 23:49:01,975-INFO: pass 4, batch 10, accuracy:0.6953125 loss:0.9240391254425049 time:0.02 sec 2021-08-02 23:49:01,975 - [10652] [line:200] - INFO: pass 4, batch 10, accuracy:0.6953125 loss:0.9240391254425049 time:0.02 sec 2021-08-02 23:49:02,274-INFO: pass 4, batch 20, accuracy:0.78125 loss:0.8611726760864258 time:0.03 sec 2021-08-02 23:49:02,274 - [10652] [line:200] - INFO: pass 4, batch 20, accuracy:0.78125 loss:0.8611726760864258 time:0.03 sec 2021-08-02 23:49:02,589-INFO: pass 4, batch 30, accuracy:0.78125 loss:0.8557838201522827 time:0.03 sec 2021-08-02 23:49:02,589 - [10652] [line:200] - INFO: pass 4, batch 30, accuracy:0.78125 loss:0.8557838201522827 time:0.03 sec 2021-08-02 23:49:02,912-INFO: pass 4, batch 40, accuracy:0.7578125 loss:0.8704908490180969 time:0.04 sec 2021-08-02 23:49:02,912 - [10652] [line:200] - INFO: pass 4, batch 40, accuracy:0.7578125 loss:0.8704908490180969 time:0.04 sec 2021-08-02 23:49:03,221-INFO: pass 4, batch 50, accuracy:0.765625 loss:0.8563944697380066 time:0.03 sec 2021-08-02 23:49:03,221 - [10652] [line:200] - INFO: pass 4, batch 50, accuracy:0.765625 loss:0.8563944697380066 time:0.03 sec 2021-08-02 23:49:03,521-INFO: pass 4, batch 60, accuracy:0.6953125 loss:0.9025968909263611 time:0.03 sec 2021-08-02 23:49:03,521 - [10652] [line:200] - INFO: pass 4, batch 60, accuracy:0.6953125 loss:0.9025968909263611 time:0.03 sec 2021-08-02 23:49:03,848-INFO: pass 4, batch 70, accuracy:0.7890625 loss:0.8335443735122681 time:0.03 sec 2021-08-02 23:49:03,848 - [10652] [line:200] - INFO: pass 4, batch 70, accuracy:0.7890625 loss:0.8335443735122681 time:0.03 sec 2021-08-02 23:49:04,000-INFO: pass 4 train result, current pass mean accuracy:0.7439126042943252 loss:0.878848555840944 2021-08-02 23:49:04,000 - [10652] [line:207] - INFO: pass 4 train result, current pass mean accuracy:0.7439126042943252 loss:0.878848555840944 2021-08-02 23:49:04,000-INFO: current pass: 5, start read text 2021-08-02 23:49:04,000 - [10652] [line:179] - INFO: current pass: 5, start read text 2021-08-02 23:49:04,031-INFO: pass 5, batch 0, accuracy:0.7421875 loss:0.8637392520904541 time:0.03 sec 2021-08-02 23:49:04,031 - [10652] [line:200] - INFO: pass 5, batch 0, accuracy:0.7421875 loss:0.8637392520904541 time:0.03 sec 2021-08-02 23:49:04,354-INFO: pass 5, batch 10, accuracy:0.765625 loss:0.8477797508239746 time:0.03 sec 2021-08-02 23:49:04,354 - [10652] [line:200] - INFO: pass 5, batch 10, accuracy:0.765625 loss:0.8477797508239746 time:0.03 sec 2021-08-02 23:49:04,637-INFO: pass 5, batch 20, accuracy:0.7109375 loss:0.8814026117324829 time:0.02 sec 2021-08-02 23:49:04,637 - [10652] [line:200] - INFO: pass 5, batch 20, accuracy:0.7109375 loss:0.8814026117324829 time:0.02 sec 2021-08-02 23:49:04,957-INFO: pass 5, batch 30, accuracy:0.7265625 loss:0.8628135919570923 time:0.03 sec 2021-08-02 23:49:04,957 - [10652] [line:200] - INFO: pass 5, batch 30, accuracy:0.7265625 loss:0.8628135919570923 time:0.03 sec 2021-08-02 23:49:05,274-INFO: pass 5, batch 40, accuracy:0.75 loss:0.8458261489868164 time:0.03 sec 2021-08-02 23:49:05,274 - [10652] [line:200] - INFO: pass 5, batch 40, accuracy:0.75 loss:0.8458261489868164 time:0.03 sec 2021-08-02 23:49:05,633-INFO: pass 5, batch 50, accuracy:0.7890625 loss:0.8113884925842285 time:0.08 sec 2021-08-02 23:49:05,633 - [10652] [line:200] - INFO: pass 5, batch 50, accuracy:0.7890625 loss:0.8113884925842285 time:0.08 sec 2021-08-02 23:49:05,932-INFO: pass 5, batch 60, accuracy:0.734375 loss:0.8523907661437988 time:0.03 sec 2021-08-02 23:49:05,932 - [10652] [line:200] - INFO: pass 5, batch 60, accuracy:0.734375 loss:0.8523907661437988 time:0.03 sec 2021-08-02 23:49:06,230-INFO: pass 5, batch 70, accuracy:0.7265625 loss:0.8549454808235168 time:0.04 sec 2021-08-02 23:49:06,230 - [10652] [line:200] - INFO: pass 5, batch 70, accuracy:0.7265625 loss:0.8549454808235168 time:0.04 sec 2021-08-02 23:49:06,389-INFO: pass 5 train result, current pass mean accuracy:0.7447312348767331 loss:0.8509354889392853 2021-08-02 23:49:06,389 - [10652] [line:207] - INFO: pass 5 train result, current pass mean accuracy:0.7447312348767331 loss:0.8509354889392853 2021-08-02 23:49:06,389-INFO: current pass: 6, start read text 2021-08-02 23:49:06,389 - [10652] [line:179] - INFO: current pass: 6, start read text 2021-08-02 23:49:06,422-INFO: pass 6, batch 0, accuracy:0.796875 loss:0.8081477880477905 time:0.03 sec 2021-08-02 23:49:06,422 - [10652] [line:200] - INFO: pass 6, batch 0, accuracy:0.796875 loss:0.8081477880477905 time:0.03 sec 2021-08-02 23:49:06,728-INFO: pass 6, batch 10, accuracy:0.6953125 loss:0.8746772408485413 time:0.03 sec 2021-08-02 23:49:06,728 - [10652] [line:200] - INFO: pass 6, batch 10, accuracy:0.6953125 loss:0.8746772408485413 time:0.03 sec 2021-08-02 23:49:07,044-INFO: pass 6, batch 20, accuracy:0.7890625 loss:0.8035359382629395 time:0.03 sec 2021-08-02 23:49:07,044 - [10652] [line:200] - INFO: pass 6, batch 20, accuracy:0.7890625 loss:0.8035359382629395 time:0.03 sec 2021-08-02 23:49:07,349-INFO: pass 6, batch 30, accuracy:0.78125 loss:0.7980951070785522 time:0.03 sec 2021-08-02 23:49:07,349 - [10652] [line:200] - INFO: pass 6, batch 30, accuracy:0.78125 loss:0.7980951070785522 time:0.03 sec 2021-08-02 23:49:07,657-INFO: pass 6, batch 40, accuracy:0.7890625 loss:0.8009939193725586 time:0.03 sec 2021-08-02 23:49:07,657 - [10652] [line:200] - INFO: pass 6, batch 40, accuracy:0.7890625 loss:0.8009939193725586 time:0.03 sec 2021-08-02 23:49:07,958-INFO: pass 6, batch 50, accuracy:0.78125 loss:0.8017900586128235 time:0.03 sec 2021-08-02 23:49:07,958 - [10652] [line:200] - INFO: pass 6, batch 50, accuracy:0.78125 loss:0.8017900586128235 time:0.03 sec 2021-08-02 23:49:08,259-INFO: pass 6, batch 60, accuracy:0.71875 loss:0.8546284437179565 time:0.03 sec 2021-08-02 23:49:08,259 - [10652] [line:200] - INFO: pass 6, batch 60, accuracy:0.71875 loss:0.8546284437179565 time:0.03 sec 2021-08-02 23:49:08,559-INFO: pass 6, batch 70, accuracy:0.71875 loss:0.8484985828399658 time:0.03 sec 2021-08-02 23:49:08,559 - [10652] [line:200] - INFO: pass 6, batch 70, accuracy:0.71875 loss:0.8484985828399658 time:0.03 sec 2021-08-02 23:49:08,689-INFO: pass 6 train result, current pass mean accuracy:0.7452769883369145 loss:0.833906288993986 2021-08-02 23:49:08,689 - [10652] [line:207] - INFO: pass 6 train result, current pass mean accuracy:0.7452769883369145 loss:0.833906288993986 2021-08-02 23:49:08,689-INFO: temp save pass 6 train result, current best pass mean accuracy:0.7452769883369145 2021-08-02 23:49:08,689 - [10652] [line:209] - INFO: temp save pass 6 train result, current best pass mean accuracy:0.7452769883369145 2021-08-02 23:49:09,330-INFO: current pass: 7, start read text 2021-08-02 23:49:09,330 - [10652] [line:179] - INFO: current pass: 7, start read text 2021-08-02 23:49:09,368-INFO: pass 7, batch 0, accuracy:0.7578125 loss:0.8156975507736206 time:0.04 sec 2021-08-02 23:49:09,368 - [10652] [line:200] - INFO: pass 7, batch 0, accuracy:0.7578125 loss:0.8156975507736206 time:0.04 sec 2021-08-02 23:49:09,669-INFO: pass 7, batch 10, accuracy:0.796875 loss:0.7870999574661255 time:0.02 sec 2021-08-02 23:49:09,669 - [10652] [line:200] - INFO: pass 7, batch 10, accuracy:0.796875 loss:0.7870999574661255 time:0.02 sec 2021-08-02 23:49:09,977-INFO: pass 7, batch 20, accuracy:0.7578125 loss:0.8177455067634583 time:0.03 sec 2021-08-02 23:49:09,977 - [10652] [line:200] - INFO: pass 7, batch 20, accuracy:0.7578125 loss:0.8177455067634583 time:0.03 sec 2021-08-02 23:49:10,276-INFO: pass 7, batch 30, accuracy:0.859375 loss:0.7317814826965332 time:0.03 sec 2021-08-02 23:49:10,276 - [10652] [line:200] - INFO: pass 7, batch 30, accuracy:0.859375 loss:0.7317814826965332 time:0.03 sec 2021-08-02 23:49:10,583-INFO: pass 7, batch 40, accuracy:0.75 loss:0.8145904541015625 time:0.03 sec 2021-08-02 23:49:10,583 - [10652] [line:200] - INFO: pass 7, batch 40, accuracy:0.75 loss:0.8145904541015625 time:0.03 sec 2021-08-02 23:49:10,894-INFO: pass 7, batch 50, accuracy:0.8125 loss:0.7646461725234985 time:0.03 sec 2021-08-02 23:49:10,894 - [10652] [line:200] - INFO: pass 7, batch 50, accuracy:0.8125 loss:0.7646461725234985 time:0.03 sec 2021-08-02 23:49:11,210-INFO: pass 7, batch 60, accuracy:0.796875 loss:0.7782437801361084 time:0.03 sec 2021-08-02 23:49:11,210 - [10652] [line:200] - INFO: pass 7, batch 60, accuracy:0.796875 loss:0.7782437801361084 time:0.03 sec 2021-08-02 23:49:11,538-INFO: pass 7, batch 70, accuracy:0.7578125 loss:0.8106646537780762 time:0.03 sec 2021-08-02 23:49:11,538 - [10652] [line:200] - INFO: pass 7, batch 70, accuracy:0.7578125 loss:0.8106646537780762 time:0.03 sec 2021-08-02 23:49:11,683-INFO: pass 7 train result, current pass mean accuracy:0.7441854814165517 loss:0.8246896737500241 2021-08-02 23:49:11,683 - [10652] [line:207] - INFO: pass 7 train result, current pass mean accuracy:0.7441854814165517 loss:0.8246896737500241 2021-08-02 23:49:11,684-INFO: current pass: 8, start read text 2021-08-02 23:49:11,684 - [10652] [line:179] - INFO: current pass: 8, start read text 2021-08-02 23:49:11,713-INFO: pass 8, batch 0, accuracy:0.7109375 loss:0.8437689542770386 time:0.03 sec 2021-08-02 23:49:11,713 - [10652] [line:200] - INFO: pass 8, batch 0, accuracy:0.7109375 loss:0.8437689542770386 time:0.03 sec 2021-08-02 23:49:12,027-INFO: pass 8, batch 10, accuracy:0.7421875 loss:0.8261438012123108 time:0.03 sec 2021-08-02 23:49:12,027 - [10652] [line:200] - INFO: pass 8, batch 10, accuracy:0.7421875 loss:0.8261438012123108 time:0.03 sec 2021-08-02 23:49:12,333-INFO: pass 8, batch 20, accuracy:0.7421875 loss:0.820777416229248 time:0.03 sec 2021-08-02 23:49:12,333 - [10652] [line:200] - INFO: pass 8, batch 20, accuracy:0.7421875 loss:0.820777416229248 time:0.03 sec 2021-08-02 23:49:12,660-INFO: pass 8, batch 30, accuracy:0.6875 loss:0.8689138293266296 time:0.03 sec 2021-08-02 23:49:12,660 - [10652] [line:200] - INFO: pass 8, batch 30, accuracy:0.6875 loss:0.8689138293266296 time:0.03 sec 2021-08-02 23:49:12,976-INFO: pass 8, batch 40, accuracy:0.734375 loss:0.8207021951675415 time:0.03 sec 2021-08-02 23:49:12,976 - [10652] [line:200] - INFO: pass 8, batch 40, accuracy:0.734375 loss:0.8207021951675415 time:0.03 sec 2021-08-02 23:49:13,263-INFO: pass 8, batch 50, accuracy:0.765625 loss:0.8096222877502441 time:0.02 sec 2021-08-02 23:49:13,263 - [10652] [line:200] - INFO: pass 8, batch 50, accuracy:0.765625 loss:0.8096222877502441 time:0.02 sec 2021-08-02 23:49:13,578-INFO: pass 8, batch 60, accuracy:0.734375 loss:0.8275260925292969 time:0.03 sec 2021-08-02 23:49:13,578 - [10652] [line:200] - INFO: pass 8, batch 60, accuracy:0.734375 loss:0.8275260925292969 time:0.03 sec 2021-08-02 23:49:13,870-INFO: pass 8, batch 70, accuracy:0.7421875 loss:0.814354658126831 time:0.02 sec 2021-08-02 23:49:13,870 - [10652] [line:200] - INFO: pass 8, batch 70, accuracy:0.7421875 loss:0.814354658126831 time:0.02 sec 2021-08-02 23:49:14,025-INFO: pass 8 train result, current pass mean accuracy:0.7445947967077556 loss:0.817827045917511 2021-08-02 23:49:14,025 - [10652] [line:207] - INFO: pass 8 train result, current pass mean accuracy:0.7445947967077556 loss:0.817827045917511 2021-08-02 23:49:14,025-INFO: current pass: 9, start read text 2021-08-02 23:49:14,025 - [10652] [line:179] - INFO: current pass: 9, start read text 2021-08-02 23:49:14,049-INFO: pass 9, batch 0, accuracy:0.7734375 loss:0.7969448566436768 time:0.02 sec 2021-08-02 23:49:14,049 - [10652] [line:200] - INFO: pass 9, batch 0, accuracy:0.7734375 loss:0.7969448566436768 time:0.02 sec 2021-08-02 23:49:14,341-INFO: pass 9, batch 10, accuracy:0.734375 loss:0.8250979781150818 time:0.02 sec 2021-08-02 23:49:14,341 - [10652] [line:200] - INFO: pass 9, batch 10, accuracy:0.734375 loss:0.8250979781150818 time:0.02 sec 2021-08-02 23:49:14,642-INFO: pass 9, batch 20, accuracy:0.7578125 loss:0.8018253445625305 time:0.03 sec 2021-08-02 23:49:14,642 - [10652] [line:200] - INFO: pass 9, batch 20, accuracy:0.7578125 loss:0.8018253445625305 time:0.03 sec 2021-08-02 23:49:14,952-INFO: pass 9, batch 30, accuracy:0.734375 loss:0.820142924785614 time:0.03 sec 2021-08-02 23:49:14,952 - [10652] [line:200] - INFO: pass 9, batch 30, accuracy:0.734375 loss:0.820142924785614 time:0.03 sec 2021-08-02 23:49:15,268-INFO: pass 9, batch 40, accuracy:0.75 loss:0.814818799495697 time:0.03 sec 2021-08-02 23:49:15,268 - [10652] [line:200] - INFO: pass 9, batch 40, accuracy:0.75 loss:0.814818799495697 time:0.03 sec 2021-08-02 23:49:15,572-INFO: pass 9, batch 50, accuracy:0.7109375 loss:0.8432407379150391 time:0.03 sec 2021-08-02 23:49:15,572 - [10652] [line:200] - INFO: pass 9, batch 50, accuracy:0.7109375 loss:0.8432407379150391 time:0.03 sec 2021-08-02 23:49:15,864-INFO: pass 9, batch 60, accuracy:0.6875 loss:0.861954927444458 time:0.03 sec 2021-08-02 23:49:15,864 - [10652] [line:200] - INFO: pass 9, batch 60, accuracy:0.6875 loss:0.861954927444458 time:0.03 sec 2021-08-02 23:49:16,177-INFO: pass 9, batch 70, accuracy:0.796875 loss:0.7713867425918579 time:0.02 sec 2021-08-02 23:49:16,177 - [10652] [line:200] - INFO: pass 9, batch 70, accuracy:0.796875 loss:0.7713867425918579 time:0.02 sec 2021-08-02 23:49:16,327-INFO: pass 9 train result, current pass mean accuracy:0.7448676738299822 loss:0.8131639180999053 2021-08-02 23:49:16,327 - [10652] [line:207] - INFO: pass 9 train result, current pass mean accuracy:0.7448676738299822 loss:0.8131639180999053 2021-08-02 23:49:16,328-INFO: current pass: 10, start read text 2021-08-02 23:49:16,328 - [10652] [line:179] - INFO: current pass: 10, start read text 2021-08-02 23:49:16,360-INFO: pass 10, batch 0, accuracy:0.828125 loss:0.7386550903320312 time:0.03 sec 2021-08-02 23:49:16,360 - [10652] [line:200] - INFO: pass 10, batch 0, accuracy:0.828125 loss:0.7386550903320312 time:0.03 sec 2021-08-02 23:49:16,653-INFO: pass 10, batch 10, accuracy:0.6796875 loss:0.8697382211685181 time:0.03 sec 2021-08-02 23:49:16,653 - [10652] [line:200] - INFO: pass 10, batch 10, accuracy:0.6796875 loss:0.8697382211685181 time:0.03 sec 2021-08-02 23:49:16,950-INFO: pass 10, batch 20, accuracy:0.75 loss:0.8044241666793823 time:0.03 sec 2021-08-02 23:49:16,950 - [10652] [line:200] - INFO: pass 10, batch 20, accuracy:0.75 loss:0.8044241666793823 time:0.03 sec 2021-08-02 23:49:17,269-INFO: pass 10, batch 30, accuracy:0.7109375 loss:0.8448526263237 time:0.03 sec 2021-08-02 23:49:17,269 - [10652] [line:200] - INFO: pass 10, batch 30, accuracy:0.7109375 loss:0.8448526263237 time:0.03 sec 2021-08-02 23:49:17,558-INFO: pass 10, batch 40, accuracy:0.703125 loss:0.8442284464836121 time:0.02 sec 2021-08-02 23:49:17,558 - [10652] [line:200] - INFO: pass 10, batch 40, accuracy:0.703125 loss:0.8442284464836121 time:0.02 sec 2021-08-02 23:49:17,845-INFO: pass 10, batch 50, accuracy:0.75 loss:0.8083001971244812 time:0.03 sec 2021-08-02 23:49:17,845 - [10652] [line:200] - INFO: pass 10, batch 50, accuracy:0.75 loss:0.8083001971244812 time:0.03 sec 2021-08-02 23:49:18,154-INFO: pass 10, batch 60, accuracy:0.7734375 loss:0.7828631401062012 time:0.03 sec 2021-08-02 23:49:18,154 - [10652] [line:200] - INFO: pass 10, batch 60, accuracy:0.7734375 loss:0.7828631401062012 time:0.03 sec 2021-08-02 23:49:18,479-INFO: pass 10, batch 70, accuracy:0.7578125 loss:0.7909150719642639 time:0.03 sec 2021-08-02 23:49:18,479 - [10652] [line:200] - INFO: pass 10, batch 70, accuracy:0.7578125 loss:0.7909150719642639 time:0.03 sec 2021-08-02 23:49:18,635-INFO: pass 10 train result, current pass mean accuracy:0.7444583585387782 loss:0.8109645419999173 2021-08-02 23:49:18,635 - [10652] [line:207] - INFO: pass 10 train result, current pass mean accuracy:0.7444583585387782 loss:0.8109645419999173 2021-08-02 23:49:18,635-INFO: current pass: 11, start read text 2021-08-02 23:49:18,635 - [10652] [line:179] - INFO: current pass: 11, start read text 2021-08-02 23:49:18,667-INFO: pass 11, batch 0, accuracy:0.734375 loss:0.8220481872558594 time:0.03 sec 2021-08-02 23:49:18,667 - [10652] [line:200] - INFO: pass 11, batch 0, accuracy:0.734375 loss:0.8220481872558594 time:0.03 sec 2021-08-02 23:49:18,971-INFO: pass 11, batch 10, accuracy:0.71875 loss:0.8317664861679077 time:0.03 sec 2021-08-02 23:49:18,971 - [10652] [line:200] - INFO: pass 11, batch 10, accuracy:0.71875 loss:0.8317664861679077 time:0.03 sec 2021-08-02 23:49:19,284-INFO: pass 11, batch 20, accuracy:0.78125 loss:0.7798622250556946 time:0.03 sec 2021-08-02 23:49:19,284 - [10652] [line:200] - INFO: pass 11, batch 20, accuracy:0.78125 loss:0.7798622250556946 time:0.03 sec 2021-08-02 23:49:19,606-INFO: pass 11, batch 30, accuracy:0.765625 loss:0.7938206791877747 time:0.03 sec 2021-08-02 23:49:19,606 - [10652] [line:200] - INFO: pass 11, batch 30, accuracy:0.765625 loss:0.7938206791877747 time:0.03 sec 2021-08-02 23:49:19,908-INFO: pass 11, batch 40, accuracy:0.7578125 loss:0.8055585622787476 time:0.03 sec 2021-08-02 23:49:19,908 - [10652] [line:200] - INFO: pass 11, batch 40, accuracy:0.7578125 loss:0.8055585622787476 time:0.03 sec 2021-08-02 23:49:20,224-INFO: pass 11, batch 50, accuracy:0.7421875 loss:0.8090927600860596 time:0.03 sec 2021-08-02 23:49:20,224 - [10652] [line:200] - INFO: pass 11, batch 50, accuracy:0.7421875 loss:0.8090927600860596 time:0.03 sec 2021-08-02 23:49:20,539-INFO: pass 11, batch 60, accuracy:0.78125 loss:0.7800883054733276 time:0.03 sec 2021-08-02 23:49:20,539 - [10652] [line:200] - INFO: pass 11, batch 60, accuracy:0.78125 loss:0.7800883054733276 time:0.03 sec 2021-08-02 23:49:20,823-INFO: pass 11, batch 70, accuracy:0.8203125 loss:0.7432873845100403 time:0.03 sec 2021-08-02 23:49:20,823 - [10652] [line:200] - INFO: pass 11, batch 70, accuracy:0.8203125 loss:0.7432873845100403 time:0.03 sec 2021-08-02 23:49:20,978-INFO: pass 11 train result, current pass mean accuracy:0.7445947967077556 loss:0.8092331666695444 2021-08-02 23:49:20,978 - [10652] [line:207] - INFO: pass 11 train result, current pass mean accuracy:0.7445947967077556 loss:0.8092331666695444 2021-08-02 23:49:20,978-INFO: training till last epcho, end training 2021-08-02 23:49:20,978 - [10652] [line:214] - INFO: training till last epcho, end training

3.3.6.2 固化训练参数

In [21]

!python freeze.py

2021-08-02 23:49:25,071-INFO: build network bilstm_net to freeze 2021-08-02 23:49:25,071 - [10749] [line:46] - INFO: build network bilstm_net to freeze 2021-08-02 23:49:26,166-INFO: freeze bilstm_net end 2021-08-02 23:49:26,166 - [10749] [line:66] - INFO: freeze bilstm_net end

3.3.6.3 验证训练结果

In [22]

!python eval.py

W0802 23:49:44.475302 10805 device_context.cc:252] Please NOTE: device: 0, CUDA Capability: 70, Driver API Version: 10.1, Runtime API Version: 9.0 W0802 23:49:44.480252 10805 device_context.cc:260] device: 0, cuDNN Version: 7.6. total eval count:1036 cost time:7.18 sec predict accuracy:0.750965250965251

3.3.6.4 使用模型进行预测

In [23]

!python infer.py

W0802 23:50:11.259743 10899 device_context.cc:252] Please NOTE: device: 0, CUDA Capability: 70, Driver API Version: 10.1, Runtime API Version: 9.0 W0802 23:50:11.264673 10899 device_context.cc:260] device: 0, cuDNN Version: 7.6. infer text:我 就是 喜欢 你的 知性 大方 温柔 infer result:1

3.3.7 TEXTCNN模型的训练与预测

3.3.7.1 训练模型

更改train_parameters中"model_type"为"textcnn_net"

In [24]

!python train.py

2021-08-02 23:50:51,469-INFO: start train text classification, train params: {'class_dim': 3, 'sample_count': 9656, 'label_dict': {'消极': 0, '中性': 1, '积极': 2}, 'data_dir': 'data/data9740', 'vocabulary': 'vocab.txt', 'train_file_list': 'train.txt', 'label_file': 'label_list.txt', 'save_model_dir': './train_model', 'save_freeze_dir': './infer_model', 'pretrained_model_dir': './pretrained_model', 'continue_train': False, 'pretrained': False, 'model_type': 'textcnn_net', 'num_epochs': 12, 'train_batch_size': 128, 'use_gpu': True, 'sample_frequency': 10, 'rsm_strategy': {'learning_rate': 0.001, 'lr_epochs': [6, 8, 10], 'lr_decay': [1, 0.5, 0.25, 0.1]}, 'momentum_strategy': {'learning_rate': 0.001, 'lr_epochs': [6, 8, 10], 'lr_decay': [1, 0.5, 0.25, 0.1]}, 'sgd_strategy': {'learning_rate': 0.001, 'lr_epochs': [6, 8, 10], 'lr_decay': [1, 0.5, 0.25, 0.1]}, 'adam_strategy': {'learning_rate': 0.001}} 2021-08-02 23:50:51,469 - [11011] [line:133] - INFO: start train text classification, train params: {'class_dim': 3, 'sample_count': 9656, 'label_dict': {'消极': 0, '中性': 1, '积极': 2}, 'data_dir': 'data/data9740', 'vocabulary': 'vocab.txt', 'train_file_list': 'train.txt', 'label_file': 'label_list.txt', 'save_model_dir': './train_model', 'save_freeze_dir': './infer_model', 'pretrained_model_dir': './pretrained_model', 'continue_train': False, 'pretrained': False, 'model_type': 'textcnn_net', 'num_epochs': 12, 'train_batch_size': 128, 'use_gpu': True, 'sample_frequency': 10, 'rsm_strategy': {'learning_rate': 0.001, 'lr_epochs': [6, 8, 10], 'lr_decay': [1, 0.5, 0.25, 0.1]}, 'momentum_strategy': {'learning_rate': 0.001, 'lr_epochs': [6, 8, 10], 'lr_decay': [1, 0.5, 0.25, 0.1]}, 'sgd_strategy': {'learning_rate': 0.001, 'lr_epochs': [6, 8, 10], 'lr_decay': [1, 0.5, 0.25, 0.1]}, 'adam_strategy': {'learning_rate': 0.001}} 2021-08-02 23:50:51,469-INFO: create place, use gpu: True 2021-08-02 23:50:51,469 - [11011] [line:134] - INFO: create place, use gpu: True 2021-08-02 23:50:51,901-INFO: define input data tensor 2021-08-02 23:50:51,901 - [11011] [line:139] - INFO: define input data tensor 2021-08-02 23:50:51,902-INFO: build custom reader 2021-08-02 23:50:51,902 - [11011] [line:144] - INFO: build custom reader 2021-08-02 23:50:51,925-INFO: build network textcnn_net 2021-08-02 23:50:51,925 - [11011] [line:147] - INFO: build network textcnn_net 2021-08-02 23:50:51,941-INFO: build optimizer 2021-08-02 23:50:51,941 - [11011] [line:164] - INFO: build optimizer W0802 23:50:51.962895 11011 device_context.cc:252] Please NOTE: device: 0, CUDA Capability: 70, Driver API Version: 10.1, Runtime API Version: 9.0 W0802 23:50:51.967571 11011 device_context.cc:260] device: 0, cuDNN Version: 7.6. 2021-08-02 23:51:05,757-INFO: current pass: 0, start read text 2021-08-02 23:51:05,757 - [11011] [line:179] - INFO: current pass: 0, start read text 2021-08-02 23:51:05,815-INFO: pass 0, batch 0, accuracy:0.328125 loss:1.0972421169281006 time:0.06 sec 2021-08-02 23:51:05,815 - [11011] [line:200] - INFO: pass 0, batch 0, accuracy:0.328125 loss:1.0972421169281006 time:0.06 sec 2021-08-02 23:51:06,160-INFO: pass 0, batch 10, accuracy:0.6875 loss:1.0842721462249756 time:0.04 sec 2021-08-02 23:51:06,160 - [11011] [line:200] - INFO: pass 0, batch 10, accuracy:0.6875 loss:1.0842721462249756 time:0.04 sec 2021-08-02 23:51:06,494-INFO: pass 0, batch 20, accuracy:0.703125 loss:1.0707015991210938 time:0.03 sec 2021-08-02 23:51:06,494 - [11011] [line:200] - INFO: pass 0, batch 20, accuracy:0.703125 loss:1.0707015991210938 time:0.03 sec 2021-08-02 23:51:06,835-INFO: pass 0, batch 30, accuracy:0.7734375 loss:1.0518598556518555 time:0.04 sec 2021-08-02 23:51:06,835 - [11011] [line:200] - INFO: pass 0, batch 30, accuracy:0.7734375 loss:1.0518598556518555 time:0.04 sec 2021-08-02 23:51:07,154-INFO: pass 0, batch 40, accuracy:0.8125 loss:1.0304169654846191 time:0.03 sec 2021-08-02 23:51:07,154 - [11011] [line:200] - INFO: pass 0, batch 40, accuracy:0.8125 loss:1.0304169654846191 time:0.03 sec 2021-08-02 23:51:07,484-INFO: pass 0, batch 50, accuracy:0.7421875 loss:1.0275747776031494 time:0.03 sec 2021-08-02 23:51:07,484 - [11011] [line:200] - INFO: pass 0, batch 50, accuracy:0.7421875 loss:1.0275747776031494 time:0.03 sec 2021-08-02 23:51:07,821-INFO: pass 0, batch 60, accuracy:0.7421875 loss:1.0170284509658813 time:0.04 sec 2021-08-02 23:51:07,821 - [11011] [line:200] - INFO: pass 0, batch 60, accuracy:0.7421875 loss:1.0170284509658813 time:0.04 sec 2021-08-02 23:51:08,155-INFO: pass 0, batch 70, accuracy:0.75 loss:1.0012245178222656 time:0.03 sec 2021-08-02 23:51:08,155 - [11011] [line:200] - INFO: pass 0, batch 70, accuracy:0.75 loss:1.0012245178222656 time:0.03 sec 2021-08-02 23:51:08,316-INFO: pass 0 train result, current pass mean accuracy:0.7321228322229887 loss:1.045332748638956 2021-08-02 23:51:08,316 - [11011] [line:207] - INFO: pass 0 train result, current pass mean accuracy:0.7321228322229887 loss:1.045332748638956 2021-08-02 23:51:08,316-INFO: temp save pass 0 train result, current best pass mean accuracy:0.7321228322229887 2021-08-02 23:51:08,316 - [11011] [line:209] - INFO: temp save pass 0 train result, current best pass mean accuracy:0.7321228322229887 2021-08-02 23:51:08,928-INFO: current pass: 1, start read text 2021-08-02 23:51:08,928 - [11011] [line:179] - INFO: current pass: 1, start read text 2021-08-02 23:51:08,965-INFO: pass 1, batch 0, accuracy:0.7890625 loss:0.987476110458374 time:0.03 sec 2021-08-02 23:51:08,965 - [11011] [line:200] - INFO: pass 1, batch 0, accuracy:0.7890625 loss:0.987476110458374 time:0.03 sec 2021-08-02 23:51:09,320-INFO: pass 1, batch 10, accuracy:0.78125 loss:0.9749718308448792 time:0.04 sec 2021-08-02 23:51:09,320 - [11011] [line:200] - INFO: pass 1, batch 10, accuracy:0.78125 loss:0.9749718308448792 time:0.04 sec 2021-08-02 23:51:09,665-INFO: pass 1, batch 20, accuracy:0.75 loss:0.972743034362793 time:0.03 sec 2021-08-02 23:51:09,665 - [11011] [line:200] - INFO: pass 1, batch 20, accuracy:0.75 loss:0.972743034362793 time:0.03 sec 2021-08-02 23:51:10,005-INFO: pass 1, batch 30, accuracy:0.7109375 loss:0.9782093167304993 time:0.04 sec 2021-08-02 23:51:10,005 - [11011] [line:200] - INFO: pass 1, batch 30, accuracy:0.7109375 loss:0.9782093167304993 time:0.04 sec 2021-08-02 23:51:10,339-INFO: pass 1, batch 40, accuracy:0.71875 loss:0.9664385318756104 time:0.03 sec 2021-08-02 23:51:10,339 - [11011] [line:200] - INFO: pass 1, batch 40, accuracy:0.71875 loss:0.9664385318756104 time:0.03 sec 2021-08-02 23:51:10,669-INFO: pass 1, batch 50, accuracy:0.7421875 loss:0.9468758702278137 time:0.03 sec 2021-08-02 23:51:10,669 - [11011] [line:200] - INFO: pass 1, batch 50, accuracy:0.7421875 loss:0.9468758702278137 time:0.03 sec 2021-08-02 23:51:11,009-INFO: pass 1, batch 60, accuracy:0.7421875 loss:0.9381577968597412 time:0.03 sec 2021-08-02 23:51:11,009 - [11011] [line:200] - INFO: pass 1, batch 60, accuracy:0.7421875 loss:0.9381577968597412 time:0.03 sec 2021-08-02 23:51:11,339-INFO: pass 1, batch 70, accuracy:0.71875 loss:0.9377846717834473 time:0.03 sec 2021-08-02 23:51:11,339 - [11011] [line:200] - INFO: pass 1, batch 70, accuracy:0.71875 loss:0.9377846717834473 time:0.03 sec 2021-08-02 23:51:11,496-INFO: pass 1 train result, current pass mean accuracy:0.7440490432475743 loss:0.9589235163048694 2021-08-02 23:51:11,496 - [11011] [line:207] - INFO: pass 1 train result, current pass mean accuracy:0.7440490432475743 loss:0.9589235163048694 2021-08-02 23:51:11,496-INFO: temp save pass 1 train result, current best pass mean accuracy:0.7440490432475743 2021-08-02 23:51:11,496 - [11011] [line:209] - INFO: temp save pass 1 train result, current best pass mean accuracy:0.7440490432475743 2021-08-02 23:51:12,123-INFO: current pass: 2, start read text 2021-08-02 23:51:12,123 - [11011] [line:179] - INFO: current pass: 2, start read text 2021-08-02 23:51:12,166-INFO: pass 2, batch 0, accuracy:0.7421875 loss:0.922042727470398 time:0.04 sec 2021-08-02 23:51:12,166 - [11011] [line:200] - INFO: pass 2, batch 0, accuracy:0.7421875 loss:0.922042727470398 time:0.04 sec 2021-08-02 23:51:15,998-INFO: pass 3, batch 20, accuracy:0.6953125 loss:0.8947010040283203 time:0.03 sec 2021-08-02 23:51:15,998 - [11011] [line:200] - INFO: pass 3, batch 20, accuracy:0.6953125 loss:0.8947010040283203 time:0.03 sec 2021-08-02 23:51:16,337-INFO: pass 3, batch 30, accuracy:0.765625 loss:0.8366822004318237 time:0.04 sec 2021-08-02 23:51:16,337 - [11011] [line:200] - INFO: pass 3, batch 30, accuracy:0.765625 loss:0.8366822004318237 time:0.04 sec 2021-08-02 23:51:16,671-INFO: pass 3, batch 40, accuracy:0.75 loss:0.8411478996276855 time:0.03 sec 2021-08-02 23:51:16,671 - [11011] [line:200] - INFO: pass 3, batch 40, accuracy:0.75 loss:0.8411478996276855 time:0.03 sec 2021-08-02 23:51:17,004-INFO: pass 3, batch 50, accuracy:0.7734375 loss:0.8299131393432617 time:0.03 sec 2021-08-02 23:51:17,004 - [11011] [line:200] - INFO: pass 3, batch 50, accuracy:0.7734375 loss:0.8299131393432617 time:0.03 sec 2021-08-02 23:51:17,328-INFO: pass 3, batch 60, accuracy:0.78125 loss:0.8135496377944946 time:0.03 sec 2021-08-02 23:51:17,328 - [11011] [line:200] - INFO: pass 3, batch 60, accuracy:0.78125 loss:0.8135496377944946 time:0.03 sec 2021-08-02 23:51:17,661-INFO: pass 3, batch 70, accuracy:0.7265625 loss:0.8516656756401062 time:0.03 sec 2021-08-02 23:51:17,661 - [11011] [line:200] - INFO: pass 3, batch 70, accuracy:0.7265625 loss:0.8516656756401062 time:0.03 sec 2021-08-02 23:51:17,822-INFO: pass 3 train result, current pass mean accuracy:0.7440490432475743 loss:0.8495429417020396 2021-08-02 23:51:17,822 - [11011] [line:207] - INFO: pass 3 train result, current pass mean accuracy:0.7440490432475743 loss:0.8495429417020396 2021-08-02 23:51:17,822-INFO: current pass: 4, start read text 2021-08-02 23:51:17,822 - [11011] [line:179] - INFO: current pass: 4, start read text 2021-08-02 23:51:17,854-INFO: pass 4, batch 0, accuracy:0.703125 loss:0.8586623668670654 time:0.03 sec 2021-08-02 23:51:17,854 - [11011] [line:200] - INFO: pass 4, batch 0, accuracy:0.703125 loss:0.8586623668670654 time:0.03 sec 2021-08-02 23:51:18,192-INFO: pass 4, batch 10, accuracy:0.75 loss:0.8180871605873108 time:0.04 sec 2021-08-02 23:51:18,192 - [11011] [line:200] - INFO: pass 4, batch 10, accuracy:0.75 loss:0.8180871605873108 time:0.04 sec 2021-08-02 23:51:18,525-INFO: pass 4, batch 20, accuracy:0.7578125 loss:0.8063067197799683 time:0.03 sec 2021-08-02 23:51:18,525 - [11011] [line:200] - INFO: pass 4, batch 20, accuracy:0.7578125 loss:0.8063067197799683 time:0.03 sec 2021-08-02 23:51:18,860-INFO: pass 4, batch 30, accuracy:0.7421875 loss:0.8158521056175232 time:0.03 sec 2021-08-02 23:51:18,860 - [11011] [line:200] - INFO: pass 4, batch 30, accuracy:0.7421875 loss:0.8158521056175232 time:0.03 sec 2021-08-02 23:51:19,197-INFO: pass 4, batch 40, accuracy:0.7265625 loss:0.8336988091468811 time:0.03 sec 2021-08-02 23:51:19,197 - [11011] [line:200] - INFO: pass 4, batch 40, accuracy:0.7265625 loss:0.8336988091468811 time:0.03 sec 2021-08-02 23:51:19,531-INFO: pass 4, batch 50, accuracy:0.703125 loss:0.8522412776947021 time:0.03 sec 2021-08-02 23:51:19,531 - [11011] [line:200] - INFO: pass 4, batch 50, accuracy:0.703125 loss:0.8522412776947021 time:0.03 sec 2021-08-02 23:51:19,866-INFO: pass 4, batch 60, accuracy:0.6640625 loss:0.8907402753829956 time:0.03 sec 2021-08-02 23:51:19,866 - [11011] [line:200] - INFO: pass 4, batch 60, accuracy:0.6640625 loss:0.8907402753829956 time:0.03 sec 2021-08-02 23:51:20,203-INFO: pass 4, batch 70, accuracy:0.8046875 loss:0.753200888633728 time:0.03 sec 2021-08-02 23:51:20,203 - [11011] [line:200] - INFO: pass 4, batch 70, accuracy:0.8046875 loss:0.753200888633728 time:0.03 sec 2021-08-02 23:51:20,360-INFO: pass 4 train result, current pass mean accuracy:0.7443219195855292 loss:0.8157412260770798 2021-08-02 23:51:20,360 - [11011] [line:207] - INFO: pass 4 train result, current pass mean accuracy:0.7443219195855292 loss:0.8157412260770798 2021-08-02 23:51:20,360-INFO: current pass: 5, start read text 2021-08-02 23:51:20,360 - [11011] [line:179] - INFO: current pass: 5, start read text 2021-08-02 23:51:20,396-INFO: pass 5, batch 0, accuracy:0.7421875 loss:0.7929457426071167 time:0.03 sec 2021-08-02 23:51:20,396 - [11011] [line:200] - INFO: pass 5, batch 0, accuracy:0.7421875 loss:0.7929457426071167 time:0.03 sec 2021-08-02 23:51:20,727-INFO: pass 5, batch 10, accuracy:0.7265625 loss:0.815055787563324 time:0.03 sec 2021-08-02 23:51:20,727 - [11011] [line:200] - INFO: pass 5, batch 10, accuracy:0.7265625 loss:0.815055787563324 time:0.03 sec 2021-08-02 23:51:21,066-INFO: pass 5, batch 20, accuracy:0.796875 loss:0.7460964322090149 time:0.04 sec 2021-08-02 23:51:21,066 - [11011] [line:200] - INFO: pass 5, batch 20, accuracy:0.796875 loss:0.7460964322090149 time:0.04 sec 2021-08-02 23:51:21,400-INFO: pass 5, batch 30, accuracy:0.7421875 loss:0.7930421829223633 time:0.03 sec 2021-08-02 23:51:21,400 - [11011] [line:200] - INFO: pass 5, batch 30, accuracy:0.7421875 loss:0.7930421829223633 time:0.03 sec 2021-08-02 23:51:21,733-INFO: pass 5, batch 40, accuracy:0.8203125 loss:0.7163165807723999 time:0.03 sec 2021-08-02 23:51:21,733 - [11011] [line:200] - INFO: pass 5, batch 40, accuracy:0.8203125 loss:0.7163165807723999 time:0.03 sec 2021-08-02 23:51:22,071-INFO: pass 5, batch 50, accuracy:0.7734375 loss:0.7556018829345703 time:0.03 sec 2021-08-02 23:51:22,071 - [11011] [line:200] - INFO: pass 5, batch 50, accuracy:0.7734375 loss:0.7556018829345703 time:0.03 sec 2021-08-02 23:51:22,406-INFO: pass 5, batch 60, accuracy:0.7578125 loss:0.7768285274505615 time:0.03 sec 2021-08-02 23:51:22,406 - [11011] [line:200] - INFO: pass 5, batch 60, accuracy:0.7578125 loss:0.7768285274505615 time:0.03 sec 2021-08-02 23:51:22,744-INFO: pass 5, batch 70, accuracy:0.734375 loss:0.789546549320221 time:0.03 sec 2021-08-02 23:51:22,744 - [11011] [line:200] - INFO: pass 5, batch 70, accuracy:0.734375 loss:0.789546549320221 time:0.03 sec 2021-08-02 23:51:22,958-INFO: pass 5 train result, current pass mean accuracy:0.7443219195855292 loss:0.7917964105543337 2021-08-02 23:51:22,958 - [11011] [line:207] - INFO: pass 5 train result, current pass mean accuracy:0.7443219195855292 loss:0.7917964105543337 2021-08-02 23:51:22,959-INFO: current pass: 6, start read text 2021-08-02 23:51:22,959 - [11011] [line:179] - INFO: current pass: 6, start read text 2021-08-02 23:51:22,991-INFO: pass 6, batch 0, accuracy:0.71875 loss:0.8212243914604187 time:0.03 sec 2021-08-02 23:51:22,991 - [11011] [line:200] - INFO: pass 6, batch 0, accuracy:0.71875 loss:0.8212243914604187 time:0.03 sec 2021-08-02 23:51:23,325-INFO: pass 6, batch 10, accuracy:0.7265625 loss:0.8002000451087952 time:0.03 sec 2021-08-02 23:51:23,325 - [11011] [line:200] - INFO: pass 6, batch 10, accuracy:0.7265625 loss:0.8002000451087952 time:0.03 sec 2021-08-02 23:51:23,669-INFO: pass 6, batch 20, accuracy:0.7265625 loss:0.8038257360458374 time:0.04 sec 2021-08-02 23:51:23,669 - [11011] [line:200] - INFO: pass 6, batch 20, accuracy:0.7265625 loss:0.8038257360458374 time:0.04 sec 2021-08-02 23:51:24,003-INFO: pass 6, batch 30, accuracy:0.7890625 loss:0.7291620969772339 time:0.03 sec 2021-08-02 23:51:24,003 - [11011] [line:200] - INFO: pass 6, batch 30, accuracy:0.7890625 loss:0.7291620969772339 time:0.03 sec 2021-08-02 23:51:24,336-INFO: pass 6, batch 40, accuracy:0.78125 loss:0.7435768246650696 time:0.03 sec 2021-08-02 23:51:24,336 - [11011] [line:200] - INFO: pass 6, batch 40, accuracy:0.78125 loss:0.7435768246650696 time:0.03 sec 2021-08-02 23:51:24,665-INFO: pass 6, batch 50, accuracy:0.71875 loss:0.8026926517486572 time:0.03 sec 2021-08-02 23:51:24,665 - [11011] [line:200] - INFO: pass 6, batch 50, accuracy:0.71875 loss:0.8026926517486572 time:0.03 sec 2021-08-02 23:51:24,993-INFO: pass 6, batch 60, accuracy:0.6875 loss:0.8358154296875 time:0.03 sec 2021-08-02 23:51:24,993 - [11011] [line:200] - INFO: pass 6, batch 60, accuracy:0.6875 loss:0.8358154296875 time:0.03 sec 2021-08-02 23:51:25,331-INFO: pass 6, batch 70, accuracy:0.78125 loss:0.7322486042976379 time:0.04 sec 2021-08-02 23:51:25,331 - [11011] [line:200] - INFO: pass 6, batch 70, accuracy:0.78125 loss:0.7322486042976379 time:0.04 sec 2021-08-02 23:51:25,488-INFO: pass 6 train result, current pass mean accuracy:0.7440490432475743 loss:0.7786422003256647 2021-08-02 23:51:25,488 - [11011] [line:207] - INFO: pass 6 train result, current pass mean accuracy:0.7440490432475743 loss:0.7786422003256647 2021-08-02 23:51:25,488-INFO: current pass: 7, start read text 2021-08-02 23:51:25,488 - [11011] [line:179] - INFO: current pass: 7, start read text 2021-08-02 23:51:25,524-INFO: pass 7, batch 0, accuracy:0.78125 loss:0.7367449998855591 time:0.03 sec 2021-08-02 23:51:25,524 - [11011] [line:200] - INFO: pass 7, batch 0, accuracy:0.78125 loss:0.7367449998855591 time:0.03 sec 2021-08-02 23:51:25,865-INFO: pass 7, batch 10, accuracy:0.7421875 loss:0.7774536609649658 time:0.03 sec 2021-08-02 23:51:25,865 - [11011] [line:200] - INFO: pass 7, batch 10, accuracy:0.7421875 loss:0.7774536609649658 time:0.03 sec 2021-08-02 23:51:26,198-INFO: pass 7, batch 20, accuracy:0.6953125 loss:0.8315885066986084 time:0.03 sec 2021-08-02 23:51:26,198 - [11011] [line:200] - INFO: pass 7, batch 20, accuracy:0.6953125 loss:0.8315885066986084 time:0.03 sec 2021-08-02 23:51:26,536-INFO: pass 7, batch 30, accuracy:0.765625 loss:0.7557453513145447 time:0.04 sec 2021-08-02 23:51:26,536 - [11011] [line:200] - INFO: pass 7, batch 30, accuracy:0.765625 loss:0.7557453513145447 time:0.04 sec 2021-08-02 23:51:26,861-INFO: pass 7, batch 40, accuracy:0.7734375 loss:0.7385739684104919 time:0.03 sec 2021-08-02 23:51:26,861 - [11011] [line:200] - INFO: pass 7, batch 40, accuracy:0.7734375 loss:0.7385739684104919 time:0.03 sec 2021-08-02 23:51:27,192-INFO: pass 7, batch 50, accuracy:0.7421875 loss:0.7729023694992065 time:0.04 sec 2021-08-02 23:51:27,192 - [11011] [line:200] - INFO: pass 7, batch 50, accuracy:0.7421875 loss:0.7729023694992065 time:0.04 sec 2021-08-02 23:51:27,527-INFO: pass 7, batch 60, accuracy:0.7421875 loss:0.7716947197914124 time:0.03 sec 2021-08-02 23:51:27,527 - [11011] [line:200] - INFO: pass 7, batch 60, accuracy:0.7421875 loss:0.7716947197914124 time:0.03 sec 2021-08-02 23:51:27,859-INFO: pass 7, batch 70, accuracy:0.8203125 loss:0.6803871989250183 time:0.03 sec 2021-08-02 23:51:27,859 - [11011] [line:200] - INFO: pass 7, batch 70, accuracy:0.8203125 loss:0.6803871989250183 time:0.03 sec 2021-08-02 23:51:28,021-INFO: pass 7 train result, current pass mean accuracy:0.7448676738299822 loss:0.7704506861536127 2021-08-02 23:51:28,021 - [11011] [line:207] - INFO: pass 7 train result, current pass mean accuracy:0.7448676738299822 loss:0.7704506861536127 2021-08-02 23:51:28,021-INFO: temp save pass 7 train result, current best pass mean accuracy:0.7448676738299822 2021-08-02 23:51:28,021 - [11011] [line:209] - INFO: temp save pass 7 train result, current best pass mean accuracy:0.7448676738299822 2021-08-02 23:51:28,641-INFO: current pass: 8, start read text 2021-08-02 23:51:28,641 - [11011] [line:179] - INFO: current pass: 8, start read text 2021-08-02 23:51:28,680-INFO: pass 8, batch 0, accuracy:0.734375 loss:0.7832275032997131 time:0.04 sec 2021-08-02 23:51:28,680 - [11011] [line:200] - INFO: pass 8, batch 0, accuracy:0.734375 loss:0.7832275032997131 time:0.04 sec 2021-08-02 23:51:29,023-INFO: pass 8, batch 10, accuracy:0.7421875 loss:0.7637192606925964 time:0.03 sec 2021-08-02 23:51:29,023 - [11011] [line:200] - INFO: pass 8, batch 10, accuracy:0.7421875 loss:0.7637192606925964 time:0.03 sec 2021-08-02 23:51:29,363-INFO: pass 8, batch 20, accuracy:0.7421875 loss:0.7635141611099243 time:0.04 sec 2021-08-02 23:51:29,363 - [11011] [line:200] - INFO: pass 8, batch 20, accuracy:0.7421875 loss:0.7635141611099243 time:0.04 sec 2021-08-02 23:51:29,697-INFO: pass 8, batch 30, accuracy:0.71875 loss:0.7986685037612915 time:0.04 sec 2021-08-02 23:51:29,697 - [11011] [line:200] - INFO: pass 8, batch 30, accuracy:0.71875 loss:0.7986685037612915 time:0.04 sec 2021-08-02 23:51:30,019-INFO: pass 8, batch 40, accuracy:0.7265625 loss:0.7832265496253967 time:0.03 sec 2021-08-02 23:51:30,019 - [11011] [line:200] - INFO: pass 8, batch 40, accuracy:0.7265625 loss:0.7832265496253967 time:0.03 sec 2021-08-02 23:51:30,347-INFO: pass 8, batch 50, accuracy:0.734375 loss:0.7788916826248169 time:0.03 sec 2021-08-02 23:51:30,347 - [11011] [line:200] - INFO: pass 8, batch 50, accuracy:0.734375 loss:0.7788916826248169 time:0.03 sec 2021-08-02 23:51:30,676-INFO: pass 8, batch 60, accuracy:0.703125 loss:0.8206707239151001 time:0.04 sec 2021-08-02 23:51:30,676 - [11011] [line:200] - INFO: pass 8, batch 60, accuracy:0.703125 loss:0.8206707239151001 time:0.04 sec 2021-08-02 23:51:31,004-INFO: pass 8, batch 70, accuracy:0.7890625 loss:0.7126451730728149 time:0.03 sec 2021-08-02 23:51:31,004 - [11011] [line:200] - INFO: pass 8, batch 70, accuracy:0.7890625 loss:0.7126451730728149 time:0.03 sec 2021-08-02 23:51:31,164-INFO: pass 8 train result, current pass mean accuracy:0.7443219195855292 loss:0.7663892044832832 2021-08-02 23:51:31,164 - [11011] [line:207] - INFO: pass 8 train result, current pass mean accuracy:0.7443219195855292 loss:0.7663892044832832 2021-08-02 23:51:31,164-INFO: current pass: 9, start read text 2021-08-02 23:51:31,164 - [11011] [line:179] - INFO: current pass: 9, start read text 2021-08-02 23:51:31,196-INFO: pass 9, batch 0, accuracy:0.7890625 loss:0.7040406465530396 time:0.03 sec 2021-08-02 23:51:31,196 - [11011] [line:200] - INFO: pass 9, batch 0, accuracy:0.7890625 loss:0.7040406465530396 time:0.03 sec 2021-08-02 23:51:31,536-INFO: pass 9, batch 10, accuracy:0.7109375 loss:0.8077605366706848 time:0.04 sec 2021-08-02 23:51:31,536 - [11011] [line:200] - INFO: pass 9, batch 10, accuracy:0.7109375 loss:0.8077605366706848 time:0.04 sec 2021-08-02 23:51:31,875-INFO: pass 9, batch 20, accuracy:0.703125 loss:0.810735285282135 time:0.04 sec 2021-08-02 23:51:31,875 - [11011] [line:200] - INFO: pass 9, batch 20, accuracy:0.703125 loss:0.810735285282135 time:0.04 sec 2021-08-02 23:51:32,209-INFO: pass 9, batch 30, accuracy:0.7421875 loss:0.7697397470474243 time:0.03 sec 2021-08-02 23:51:32,209 - [11011] [line:200] - INFO: pass 9, batch 30, accuracy:0.7421875 loss:0.7697397470474243 time:0.03 sec 2021-08-02 23:51:32,548-INFO: pass 9, batch 40, accuracy:0.796875 loss:0.7038829922676086 time:0.04 sec 2021-08-02 23:51:32,548 - [11011] [line:200] - INFO: pass 9, batch 40, accuracy:0.796875 loss:0.7038829922676086 time:0.04 sec 2021-08-02 23:51:32,876-INFO: pass 9, batch 50, accuracy:0.734375 loss:0.7786746025085449 time:0.03 sec 2021-08-02 23:51:32,876 - [11011] [line:200] - INFO: pass 9, batch 50, accuracy:0.734375 loss:0.7786746025085449 time:0.03 sec 2021-08-02 23:51:33,224-INFO: pass 9, batch 60, accuracy:0.703125 loss:0.808058500289917 time:0.03 sec 2021-08-02 23:51:33,224 - [11011] [line:200] - INFO: pass 9, batch 60, accuracy:0.703125 loss:0.808058500289917 time:0.03 sec 2021-08-02 23:51:33,566-INFO: pass 9, batch 70, accuracy:0.7265625 loss:0.7933586239814758 time:0.03 sec 2021-08-02 23:51:33,566 - [11011] [line:200] - INFO: pass 9, batch 70, accuracy:0.7265625 loss:0.7933586239814758 time:0.03 sec 2021-08-02 23:51:33,726-INFO: pass 9 train result, current pass mean accuracy:0.7445947967077556 loss:0.763118348623577 2021-08-02 23:51:33,726 - [11011] [line:207] - INFO: pass 9 train result, current pass mean accuracy:0.7445947967077556 loss:0.763118348623577 2021-08-02 23:51:33,726-INFO: current pass: 10, start read text 2021-08-02 23:51:33,726 - [11011] [line:179] - INFO: current pass: 10, start read text 2021-08-02 23:51:33,758-INFO: pass 10, batch 0, accuracy:0.7109375 loss:0.7947152853012085 time:0.03 sec 2021-08-02 23:51:33,758 - [11011] [line:200] - INFO: pass 10, batch 0, accuracy:0.7109375 loss:0.7947152853012085 time:0.03 sec 2021-08-02 23:51:34,098-INFO: pass 10, batch 10, accuracy:0.75 loss:0.7497305870056152 time:0.03 sec 2021-08-02 23:51:34,098 - [11011] [line:200] - INFO: pass 10, batch 10, accuracy:0.75 loss:0.7497305870056152 time:0.03 sec 2021-08-02 23:51:34,428-INFO: pass 10, batch 20, accuracy:0.734375 loss:0.7728780508041382 time:0.03 sec 2021-08-02 23:51:34,428 - [11011] [line:200] - INFO: pass 10, batch 20, accuracy:0.734375 loss:0.7728780508041382 time:0.03 sec 2021-08-02 23:51:34,774-INFO: pass 10, batch 30, accuracy:0.6953125 loss:0.8097286224365234 time:0.03 sec 2021-08-02 23:51:34,774 - [11011] [line:200] - INFO: pass 10, batch 30, accuracy:0.6953125 loss:0.8097286224365234 time:0.03 sec 2021-08-02 23:51:35,122-INFO: pass 10, batch 40, accuracy:0.7578125 loss:0.7424492239952087 time:0.03 sec 2021-08-02 23:51:35,122 - [11011] [line:200] - INFO: pass 10, batch 40, accuracy:0.7578125 loss:0.7424492239952087 time:0.03 sec 2021-08-02 23:51:35,473-INFO: pass 10, batch 50, accuracy:0.8203125 loss:0.6749992370605469 time:0.03 sec 2021-08-02 23:51:35,473 - [11011] [line:200] - INFO: pass 10, batch 50, accuracy:0.8203125 loss:0.6749992370605469 time:0.03 sec 2021-08-02 23:51:35,802-INFO: pass 10, batch 60, accuracy:0.6171875 loss:0.9103174209594727 time:0.03 sec 2021-08-02 23:51:35,802 - [11011] [line:200] - INFO: pass 10, batch 60, accuracy:0.6171875 loss:0.9103174209594727 time:0.03 sec 2021-08-02 23:51:36,143-INFO: pass 10, batch 70, accuracy:0.7265625 loss:0.7783061265945435 time:0.04 sec 2021-08-02 23:51:36,143 - [11011] [line:200] - INFO: pass 10, batch 70, accuracy:0.7265625 loss:0.7783061265945435 time:0.04 sec 2021-08-02 23:51:36,299-INFO: pass 10 train result, current pass mean accuracy:0.7444583585387782 loss:0.7616675935293499 2021-08-02 23:51:36,299 - [11011] [line:207] - INFO: pass 10 train result, current pass mean accuracy:0.7444583585387782 loss:0.7616675935293499 2021-08-02 23:51:36,300-INFO: current pass: 11, start read text 2021-08-02 23:51:36,300 - [11011] [line:179] - INFO: current pass: 11, start read text 2021-08-02 23:51:36,335-INFO: pass 11, batch 0, accuracy:0.7421875 loss:0.7771193981170654 time:0.03 sec 2021-08-02 23:51:36,335 - [11011] [line:200] - INFO: pass 11, batch 0, accuracy:0.7421875 loss:0.7771193981170654 time:0.03 sec 2021-08-02 23:51:36,669-INFO: pass 11, batch 10, accuracy:0.7734375 loss:0.7245420217514038 time:0.03 sec 2021-08-02 23:51:36,669 - [11011] [line:200] - INFO: pass 11, batch 10, accuracy:0.7734375 loss:0.7245420217514038 time:0.03 sec 2021-08-02 23:51:37,042-INFO: pass 11, batch 20, accuracy:0.75 loss:0.757626473903656 time:0.04 sec 2021-08-02 23:51:37,042 - [11011] [line:200] - INFO: pass 11, batch 20, accuracy:0.75 loss:0.757626473903656 time:0.04 sec 2021-08-02 23:51:37,615-INFO: pass 11, batch 30, accuracy:0.75 loss:0.7432696223258972 time:0.04 sec 2021-08-02 23:51:37,615 - [11011] [line:200] - INFO: pass 11, batch 30, accuracy:0.75 loss:0.7432696223258972 time:0.04 sec 2021-08-02 23:51:37,987-INFO: pass 11, batch 40, accuracy:0.78125 loss:0.71085524559021 time:0.04 sec 2021-08-02 23:51:37,987 - [11011] [line:200] - INFO: pass 11, batch 40, accuracy:0.78125 loss:0.71085524559021 time:0.04 sec 2021-08-02 23:51:38,353-INFO: pass 11, batch 50, accuracy:0.78125 loss:0.7113056778907776 time:0.04 sec 2021-08-02 23:51:38,353 - [11011] [line:200] - INFO: pass 11, batch 50, accuracy:0.78125 loss:0.7113056778907776 time:0.04 sec 2021-08-02 23:51:38,701-INFO: pass 11, batch 60, accuracy:0.8125 loss:0.6788443326950073 time:0.03 sec 2021-08-02 23:51:38,701 - [11011] [line:200] - INFO: pass 11, batch 60, accuracy:0.8125 loss:0.6788443326950073 time:0.03 sec 2021-08-02 23:51:39,053-INFO: pass 11, batch 70, accuracy:0.78125 loss:0.7198144197463989 time:0.04 sec 2021-08-02 23:51:39,053 - [11011] [line:200] - INFO: pass 11, batch 70, accuracy:0.78125 loss:0.7198144197463989 time:0.04 sec 2021-08-02 23:51:39,211-INFO: pass 11 train result, current pass mean accuracy:0.7448676738299822 loss:0.7601113350767839 2021-08-02 23:51:39,211 - [11011] [line:207] - INFO: pass 11 train result, current pass mean accuracy:0.7448676738299822 loss:0.7601113350767839 2021-08-02 23:51:39,211-INFO: training till last epcho, end training 2021-08-02 23:51:39,211 - [11011] [line:214] - INFO: training till last epcho, end training

3.3.7.2 固化训练参数

In [25]

!python freeze.py

2021-08-02 23:51:58,800-INFO: build network textcnn_net to freeze 2021-08-02 23:51:58,800 - [11108] [line:46] - INFO: build network textcnn_net to freeze 2021-08-02 23:51:59,885-INFO: freeze textcnn_net end 2021-08-02 23:51:59,885 - [11108] [line:66] - INFO: freeze textcnn_net end

3.3.7.3 验证训练结果

In [26]

!python eval.py

W0802 23:52:03.972072 11164 device_context.cc:252] Please NOTE: device: 0, CUDA Capability: 70, Driver API Version: 10.1, Runtime API Version: 9.0 W0802 23:52:03.977432 11164 device_context.cc:260] device: 0, cuDNN Version: 7.6. total eval count:1036 cost time:7.11 sec predict accuracy:0.750965250965251

3.3.7.4 使用模型进行预测

In [27]

!python infer.py

W0802 23:52:40.870901 11262 device_context.cc:252] Please NOTE: device: 0, CUDA Capability: 70, Driver API Version: 10.1, Runtime API Version: 9.0 W0802 23:52:40.875870 11262 device_context.cc:260] device: 0, cuDNN Version: 7.6. infer text:我 就是 喜欢 你的 知性 大方 温柔 infer result:1

4 项目部署

APP 开发于 iOS 平台，模型通过 CoreML 转化至移动端运行。 通过 Apple 提供的接口进行用户信息的采集和处理，并将对应的参数向量传递到 View 层进行渲染。

5 成果展示

5.1 APP 界面

5.2 机器人联动

相关知识

基于 NLP 情感识别的电子宠物社交应用
基于宠物情绪识别的陪护方法、装置、服务器及存储介质
基于情感设计建立用户与虚拟宠物友伴关系之研究
一种宠物情绪识别的方法和终端
一种宠物情绪识别的方法和终端专利
AI识宠｜宠物登记、宠物保险、宠物医疗、宠物比赛、品种鉴别……一拍就知道
宠物社交应用最新资讯
宠物咕噜：把人和狗的故事讲得无比生动的一款遛狗应用，下一步将推出狗狗“领地”功能切入宠物社交
基于手势识别与情感识别的虚拟宠物设计
QQ宠物下线，在线情感谁来陪伴？

网址: 基于 NLP 情感识别的电子宠物社交应用 https://m.mcbbbk.com/newsview73362.html

上一篇: 网农以每年100%速度递增 90

下一篇: 有机食品标志