[BUG] 最轻量部署依然安装torch、nvidia-XXX系列的依赖

[Mydreamandreality]

问题描述 / Problem Description 最轻量用的都是线上的api，确依然安装了torch等依赖

复现问题的步骤 / Steps to Reproduce pip install -r requirements_lite.txt

预期的结果 / Expected Result 安装web相关的依赖

实际结果 / Actual Result 安装torch，naidia-XXX一大堆系列的依赖

环境信息 / Environment Information

• langchain-ChatGLM 版本/commit 号：8721aad18d9f41d0c0687f12cd252b37e82c3b73
• 是否使用 Docker 部署（是/否）：否
• 使用的模型（ChatGLM2-6B / Qwen-7B 等）：ChatGLM-6B / Model used (ChatGLM2-6B / Qwen-7B, etc.): ChatGLM2-6B
• 使用的 Embedding 模型（moka-ai/m3e-base 等）：moka-ai/m3e-base / Embedding model used (moka-ai/m3e-base, etc.): moka-ai/m3e-base
• 使用的向量库类型 (faiss / milvus / pg_vector 等)： faiss / Vector library used (faiss, milvus, pg_vector, etc.): faiss
• 操作系统及版本 ubuntu22
• Python 版本 / Python version: 11
• 其他相关环境信息 / Other relevant environment information:

附加信息 / Additional Information 添加与问题相关的任何其他信息 / Add any other information related to the issue.

[dosubot[bot]]

It appears the issue with heavy dependencies like torch and NVIDIA-related ones being installed, despite aiming for a lightweight deployment, is due to indirect dependencies within the packages listed in your requirements_lite.txt. To address this, you'll need to scrutinize these packages for any heavy dependencies they might be pulling in.

Here's a strategy to refine your requirements_lite.txt:

• Review Packages for Heavy Dependencies: Check each package, especially langchain and langchain-experimental, for indirect heavy dependencies. Look for lighter alternatives that fulfill your requirements.
• Limit Optional Dependencies: For packages like unstructured[all-docs], consider limiting optional dependencies to avoid unnecessary heavy installations.
• Regularly Update Dependencies: Dependencies evolve, so newer versions might be lighter or heavier. Regularly review and update your dependencies to maintain a lightweight setup.

For example, ensure packages like faiss-cpu and accelerate do not indirectly include heavy dependencies. If they do, seek out lighter alternatives.

This approach should help you achieve a more lightweight deployment by avoiding unnecessary heavy dependencies.

Sources

• requirements_lite.txt
• requirements.txt

#### About Dosu

This response is meant to be useful and save you time. It is **not meant to be a precise solution**, but rather a starting point for your own research. Help me be more useful! Please leave a 👍 if this is helpful and 👎 if it is irrelevant. If you want to continue the conversation, start your reply with @dosu-bot.

[zRzRzRzRzRzRzR]

torch还是要的，但是应该是cpu的

[EscaticZheng]

是里面的unstructured默认需要torch2.2.0，导致默认安装torch，而linux下直接安装torch默认会安装gpu版的，所以应该先pip3 install torch==2.2.0 --index-url https://download.pytorch.org/whl/cpu。

[EscaticZheng]

顺带一提，在线api里的qianfan里的cachetools的依赖没有加上。我真不知道这是怎么写出来的

[EscaticZheng]

还有strsimpy这个依赖在requirements_lite.txt里也没有加上

[EscaticZheng]

还有markdownify， sentence_transformers, streamlit， 这些都没有，要不咱把requirements_lite.txt删了吧亲 @zRzRzRzRzRzRzR