khu-llm-toolkit 0.1.8

No project description provided

My LLM Utilities

About The Project

個人工具箱，便於生成式AI的開發探索。私人研究，與工作無關，來窺伺動靜的人請自行退去。

2024-02-19

修改專案及套件名稱

Built With

• 
• 

Getting Started

本節說明如何建置本專案, 以及如何將建置成果供用戶使用.

Prerequisites

開發環境需要先安裝好Python和poetry

• pyenv

  curl https://pyenv.run | bash
  

• Python

  pyenv install 3.11
  

• poetry

  curl -sSL https://install.python-poetry.org | python3 -
  

• Virtual environment

  poetry install
  poetry shell
  

Development

see How to upload your python package to PyPi

使用 Poetry

• Virtual environment

  poetry shell
  

• Test

  pytest 
  

• Build

  poetry build
  

• Upload

  poetry publish --username=__token__ --password=pypi token值
  

手動執行 setuptools

• Virtual environment

  source .venv/bin/activate
  

• Test

  pytest 
  

• Github Release

  建立並發布git tag

  git tag -a 0.1.1 -m "adjust config file content layout"
  git push origin 0.1.1
  

  然後到Github倉庫頁面建立一個新的release

  複製Assets中Source code(.tar.gz)的URL, 把它貼到setup.py裡的download_url中

• Build

  python3 setup.py sdist
  python3 setup.py clean --all
  

• Upload

  PyPi不再允許在上傳過程中用個人帳號密碼做為身份驗證方式, 參考Hackmd記錄。 在上傳過程中要求認證的時候, 以"token"做為username, 以該檔案中的token值做為密碼

  twine upload dist/*
  

Usage

完成建置並推送到PyPi server後, 就可以在其它專案中將它設為相依套件, 並在程式碼中使用.

• 相依套件

Poetry: 在 pyproject.toml 填入

khu_llm_toolkit = "^0.1.6"

Pip: 在 requirements.txt 填入

khu_llm_toolkit >= 0.1.6

• 匯入套件

from khu_llm_toolkit import ModelDefinition
from khu_llm_toolkit.commons import ProviderType

config_file_path = os.path.join(os.getcwd(), f"instance/model_definition.ini")
model_def = ModelDefinition(ProviderType.AZURE, config_file_path)
llm, embeddings = model_def.get_models(temperature=temperature)

• 設定檔範例

[DEFAULT]

[openai]
USE_AZURE = False
API_KEY = OPENAI_API_KEY
CHAT_COMPLETIONS_MODEL = gpt-4-0613
EMBEDDINGS_MODEL = text-embedding-ada-002

[azure]
USE_AZURE = True
API_KEY = AZURE_OPENAI_API_KEY
API_BASE = AZURE_OPENAI_API_ENDPOINT_URL
API_VERSION = 2023-05-15
COMPLETIONS_MODEL = gpt-35-turbo
EMBEDDINGS_MODEL = text-embedding-ada-002

Roadmap

• 支援 Gemini (0.1.7 2024-04-25)
• 一設定檔,多模型 (0.1.7 2024-04-25)
  • 排除Gemini Embeddings模型使用錯誤問題 (0.1.8 2024-06-15)
• 支援 Huggingface上的開源模型

See the open issues for a full list of proposed features (and known issues).

License

Distributed under the MIT License. See LICENSE.txt for more information.

(back to top)

Contact

Ken Hu - [e-mail]

Project Link: https://github.com/kenhutaiwan/MyLlmUtils