Are you sure you want to delete this task? Once this task is deleted, it cannot be recovered.
houysh d5bd0e5622 | 2 years ago | |
---|---|---|
docs/material | 2 years ago | |
jointai-agent | 2 years ago | |
jointai-client | 2 years ago | |
jointai-client-npu | 2 years ago | |
text_classification_paddle | 2 years ago | |
README.md | 2 years ago | |
introduction.md | 2 years ago |
情感分析
构建方式:把原始ChnSentiCorp数据集的训练数据随机拆分成part_a、part_b和test三部分
数据集 | 样本数 | 格式 | 位置 |
---|---|---|---|
训练集big | 6000 | 标签\t文本内容 | ChnSentiCorp_a/train.tsv |
训练集small | 2524 | 标签\t文本内容 | ChnSentiCorp_b/train.tsv |
验证集 | 1200 | 样本id\t标签\t文本内容 | dev.tsv |
测试集 | 1076 | 标签\t文本内容 | test.tsv |
模型训练 | 数据集 | 数据规模 |
---|---|---|
单机训练 | 全量训练集 | 8524 |
单节点训练 | 全量训练集 | 8524 |
client_A | 训练集big | 6000 |
client_B | 训练集small | 2524 |
A_B融合 | 对A和B进行模型融合 | 8524 |
模型 | epoc1 | epoc2 | epoc3 | epoc4 | epoc5 | epoc6 | epoc7 | epoc8 | epoc9 | epoc10 |
---|---|---|---|---|---|---|---|---|---|---|
单机 | 0.8808 | 0.8900 | 0.8850 | 0.8742 | 0.8683 | 0.8767 | 0.8775 | 0.8933 | 0.8800 | 0.8833 |
单节点 | 0.8808 | 0.8875 | 0.8825 | 0.8842 | 0.8758 | 0.8808 | 0.8775 | 0.8792 | 0.8750 | 0.8658 |
client_A | 0.8508 | 0.8575 | 0.8675 | 0.8608 | 0.8683 | 0.8533 | 0.8608 | 0.8642 | 0.8592 | 0.8400 |
client_B | 0.6725 | 0.8242 | 0.8092 | 0.8300 | 0.8192 | 0.7867 | 0.8050 | 0.8025 | 0.8108 | 0.8142 |
A_B融合 | 0.7992 | 0.8617 | 0.8833 | 0.8750 | 0.8783 | 0.8733 | 0.8700 | 0.8683 | 0.8742 | 0.8667 |
模型训练 | 迭代轮数 | 总耗时(s) |
---|---|---|
单机训练 | 10 | 228 |
云际学习(2个client) | 10 | 1570 |
云际学习模块 | 数据加载 | 初始化 | 参数加载 | 模型训练 | 模型评估 | 结果上传 | 参数上传 | 总计 |
---|---|---|---|---|---|---|---|---|
耗时(s) | 15 | 27 | 393 | 351 | 10 | 361 | 413 | 1570 |
Dear OpenI User
Thank you for your continuous support to the Openl Qizhi Community AI Collaboration Platform. In order to protect your usage rights and ensure network security, we updated the Openl Qizhi Community AI Collaboration Platform Usage Agreement in January 2024. The updated agreement specifies that users are prohibited from using intranet penetration tools. After you click "Agree and continue", you can continue to use our services. Thank you for your cooperation and understanding.
For more agreement content, please refer to the《Openl Qizhi Community AI Collaboration Platform Usage Agreement》