{"id":13754107,"url":"https://github.com/CrazyBoyM/phi3-Chinese","last_synced_at":"2025-05-09T22:30:45.204Z","repository":{"id":235537731,"uuid":"790882093","full_name":"CrazyBoyM/phi3-Chinese","owner":"CrazyBoyM","description":"Phi3 中文仓库 ","archived":false,"fork":false,"pushed_at":"2024-04-25T09:41:44.000Z","size":39,"stargazers_count":315,"open_issues_count":5,"forks_count":19,"subscribers_count":8,"default_branch":"main","last_synced_at":"2024-10-11T18:07:52.629Z","etag":null,"topics":["llm","llm-chinese","phi","phi3","phi3-chinese"],"latest_commit_sha":null,"homepage":"","language":"Python","has_issues":true,"has_wiki":null,"has_pages":null,"mirror_url":null,"source_name":null,"license":"mit","status":null,"scm":"git","pull_requests_enabled":true,"icon_url":"https://github.com/CrazyBoyM.png","metadata":{"files":{"readme":"README.md","changelog":null,"contributing":null,"funding":null,"license":"LICENSE","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":"2024-04-23T17:44:45.000Z","updated_at":"2024-09-20T18:56:21.000Z","dependencies_parsed_at":null,"dependency_job_id":"2666ed51-7645-4ffb-906e-6d8af86a3ea4","html_url":"https://github.com/CrazyBoyM/phi3-Chinese","commit_stats":null,"previous_names":["crazyboym/phi3-chinese"],"tags_count":0,"template":false,"template_full_name":null,"repository_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/CrazyBoyM%2Fphi3-Chinese","tags_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/CrazyBoyM%2Fphi3-Chinese/tags","releases_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/CrazyBoyM%2Fphi3-Chinese/releases","manifests_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/CrazyBoyM%2Fphi3-Chinese/manifests","owner_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners/CrazyBoyM","download_url":"https://codeload.github.com/CrazyBoyM/phi3-Chinese/tar.gz/refs/heads/main","host":{"name":"GitHub","url":"https://github.com","kind":"github","repositories_count":224884613,"owners_count":17386121,"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":["llm","llm-chinese","phi","phi3","phi3-chinese"],"created_at":"2024-08-03T09:01:40.536Z","updated_at":"2024-11-16T06:31:33.925Z","avatar_url":"https://github.com/CrazyBoyM.png","language":"Python","funding_links":[],"categories":["A01_文本生成_文本对话"],"sub_categories":["大语言对话模型及数据"],"readme":"# phi3-Chinese\nphi3以小搏大（从微软放出的跑分数据看），用不到1/2的小体积（3.8b）超越llama3 8b版性能表现，增大了在手机上部署的可行性。  \n该仓库致力于收录分散在开源社区的各种phi3的训练变体版本，让更多网友发现那些不为人知的特色有趣权重。  \n同时也会顺便整理phi相关训练、推理、部署的简单教程。  \n\n## Chat模型下载\n### Phi-3-chinese\n- Phi-3-mini-128k-instruct-Chinese\n  - 增量SFT版本：\n    - modelscope: https://modelscope.cn/models/baicai003/Phi-3-mini-128k-instruct-Chinese/summary\n  - 直接DPO版本：https://modelscope.cn/models/zhuangxialie/Phi-3-Chinese-ORPO/summary\n  - 扩充词表版本：计划中\n\n### Hugging Face（英文原版）\n- Phi-3-mini-128k-instruct：https://huggingface.co/microsoft/Phi-3-mini-128k-instruct\n- Phi-3-mini-4k-instruct：https://huggingface.co/microsoft/Phi-3-mini-4k-instruct\n\n### ModelScope（英文原版）\n- Phi-3-mini-128k-instruct：https://modelscope.cn/models/LLM-Research/Phi-3-mini-128k-instruct/summary\n- Phi-3-mini-4k-instruct：https://modelscope.cn/models/LLM-Research/Phi-3-mini-4k-instruct/summary\n\n## 网页部署\n```\nstreamlit run deploy/streamlit_for_instruct.py ./Phi-3-mini-128k-instruct-Chinese\n```\n\u003cimg width=\"1422\" alt=\"image\" src=\"https://github.com/CrazyBoyM/phi3-Chinese/assets/35400185/f77754e7-016b-4a66-9d8c-3e493faa11cb\"\u003e\n\n\n## 当前问题\n- 效果与跑分不符：理想是丰满的，但我实际深度体验英文原版、以及训练中文版体验后，发现phi3-mini并没有它说的那么好用，也许它有很大的刷分嫌疑？也许对它进行叠加block操作后很有潜力？\n- 32K词表过小：它的词表太小了，而且没什么中文token，经常约用3～5个token表示一个汉字，导致虽然它的体积小、加载快、运行快，但实际吐字速度比llama3 8b版还慢。也许应该对它进行词表扩充和增量预训练？   \n总体来说，我目前对它跑分超越llama3 8b的phi3-mini 3.8b版本是比较失望的，  \n当然也许这个版本适合更轻量级的下游垂直任务，我们不应该以gpt3.5的水平对它抱以期待？或许做个moe版本会更好？\n","project_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2FCrazyBoyM%2Fphi3-Chinese","html_url":"https://awesome.ecosyste.ms/projects/github.com%2FCrazyBoyM%2Fphi3-Chinese","lists_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2FCrazyBoyM%2Fphi3-Chinese/lists"}