{"id":13754243,"url":"https://github.com/shawroad/CoSENT_Pytorch","last_synced_at":"2025-05-09T22:31:40.914Z","repository":{"id":47920869,"uuid":"445365680","full_name":"shawroad/CoSENT_Pytorch","owner":"shawroad","description":"CoSENT、STS、SentenceBERT","archived":false,"fork":false,"pushed_at":"2025-02-07T07:35:12.000Z","size":31011,"stargazers_count":162,"open_issues_count":17,"forks_count":22,"subscribers_count":1,"default_branch":"main","last_synced_at":"2025-02-07T08:26:53.400Z","etag":null,"topics":[],"latest_commit_sha":null,"homepage":"","language":"Python","has_issues":true,"has_wiki":null,"has_pages":null,"mirror_url":null,"source_name":null,"license":null,"status":null,"scm":"git","pull_requests_enabled":true,"icon_url":"https://github.com/shawroad.png","metadata":{"files":{"readme":"README.md","changelog":null,"contributing":null,"funding":null,"license":null,"code_of_conduct":null,"threat_model":null,"audit":null,"citation":null,"codeowners":null,"security":null,"support":null,"governance":null,"roadmap":null,"authors":null,"dei":null,"publiccode":null,"codemeta":null}},"created_at":"2022-01-07T01:49:49.000Z","updated_at":"2025-02-07T07:35:16.000Z","dependencies_parsed_at":"2024-08-03T09:17:17.518Z","dependency_job_id":null,"html_url":"https://github.com/shawroad/CoSENT_Pytorch","commit_stats":null,"previous_names":[],"tags_count":0,"template":false,"template_full_name":null,"repository_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/shawroad%2FCoSENT_Pytorch","tags_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/shawroad%2FCoSENT_Pytorch/tags","releases_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/shawroad%2FCoSENT_Pytorch/releases","manifests_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/shawroad%2FCoSENT_Pytorch/manifests","owner_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners/shawroad","download_url":"https://codeload.github.com/shawroad/CoSENT_Pytorch/tar.gz/refs/heads/main","host":{"name":"GitHub","url":"https://github.com","kind":"github","repositories_count":253335716,"owners_count":21892716,"icon_url":"https://github.com/github.png","version":null,"created_at":"2022-05-30T11:31:42.601Z","updated_at":"2022-07-04T15:15:14.044Z","host_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub","repositories_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories","repository_names_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repository_names","owners_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners"}},"keywords":[],"created_at":"2024-08-03T09:01:51.747Z","updated_at":"2025-05-09T22:31:35.903Z","avatar_url":"https://github.com/shawroad.png","language":"Python","funding_links":[],"categories":["文本匹配 文本检索 文本相似度"],"sub_categories":["其他_文本生成、文本对话"],"readme":"# CoSENT_Pytorch\n\n比Sentence-BERT更有效的句向量方案\n\n## \n- 参考: https://github.com/bojone/CoSENT\n- 对应博客：https://kexue.fm/archives/8847\n- 孟子预训练模型: https://github.com/Langboat/Mengzi\n\n\n## 实验结果\n实验效果来了。 预训练模型用的是孟子(换成其他模型同样可以。如google-bert、roberta等), 学习率2e-5,batch_size=64,等价苏神代码中的batch_size=32. 只用了训练集训练，然后在测试集上做测试。 分别训练了5个epoch，使用斯皮尔曼系数评价\n\n指定不同数据集，只需在config.py文件中，修改下面两个参数:  \nparser.add_argument('--train_data', default='./data/PAWSX/PAWSX.train.data', type=str, help='训练数据集')  \nparser.add_argument('--test_data', default='./data/PAWSX/PAWSX.test.data', type=str, help='测试数据集')\n\n**另外说明:** 本实验的句子编码向量是取embedding和最后一层池化后的结果。  也可以试试其他方式，如CLS, 最后一层池化等。 最近做了一些实现，发现cls更好一些。\n\n\u003cb\u003e我的实验结果\u003c/b\u003e\n| | ATEC | BQ | LCQMC | PAWSX | STS-B | Avg |\n| :-: | :-: | :-: | :-: | :-: | :-: | :-: |\n| MengZi+CoSENT | **50.5270** | **72.2789** | **78.6981** | **60.1437** | **80.1544** | **68.3604** |\n| Sentence-MengZi | 40.7809 | 70.6998 | 77.2590 | 46.31491 | 49.9348 | 56.9978 |\n| Roberta+CoSENT | **50.5969** | **72.5191** | 79.3777 | **60.5475** | **80.4344** | **68.6951** | \n| Sentence-Roberta | 48.5157 | 67.8545 | **79.6023** | 60.1675 | 71.0148 | 65.4309 | \n\n\n\u003cb\u003e苏神的结果:\u003c/b\u003e\ntrain训练、test测试：\n| | ATEC | BQ | LCQMC | PAWSX | STS-B | Avg |\n| :-: | :-: | :-: | :-: | :-: | :-: | :-: |\n| BERT+CoSENT | **49.74** | **72.38** | 78.69 | **60.00** | **80.14** | **68.19** |\n| Sentence-BERT | 46.36 | 70.36 | **78.72** | 46.86 | 66.41 | 61.74 |\n| RoBERTa+CoSENT | **50.81** | **71.45** | **79.31** | **61.56** | **81.13** | **68.85** |\n| Sentence-RoBERTa | 48.29 | 69.99 | 79.22 | 44.10 | 72.42 | 62.80 |\n\n## 使用\n1. 运行CoSENT模型  \n\n```\nsh start.sh\n```\n\n2. 运行SentenceBert模型\n\n```\n首先，执行 python sentence_bert/data_helper.py  生成对应的数据\n再执行 CUDA_VISIBLE_DEVICES=0 python sentence_bert/run_sentence_bert_transformers_reg_loss.py\n```\n\n## **更多句子表示学习的模型见: [链接](https://github.com/shawroad/Semantic-Textual-Similarity-Pytorch)**\n\n\n## Star History\n\n[![Star History Chart](https://api.star-history.com/svg?repos=shawroad/CoSENT_Pytorch\u0026type=Date)](https://star-history.com/#shawroad/CoSENT_Pytorch\u0026Date)\n","project_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Fshawroad%2FCoSENT_Pytorch","html_url":"https://awesome.ecosyste.ms/projects/github.com%2Fshawroad%2FCoSENT_Pytorch","lists_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Fshawroad%2FCoSENT_Pytorch/lists"}