https://github.com/alekseyscorpi/vacancies_server

This is a server for vacancies generation using LLM (Saiga3)
https://github.com/alekseyscorpi/vacancies_server

code cuda cuda-toolkit docker dockerfile flask llama3 llamacpp llm ngrok pydantic saiga

Last synced: 7 days ago
JSON representation

This is a server for vacancies generation using LLM (Saiga3)

• Host: GitHub
• URL: https://github.com/alekseyscorpi/vacancies_server
• Owner: AlekseyScorpi
• Created: 2024-05-03T10:32:38.000Z (9 months ago)
• Default Branch: main
• Last Pushed: 2024-05-31T10:02:39.000Z (8 months ago)
• Last Synced: 2025-01-20T16:08:48.466Z (10 days ago)
• Topics: code, cuda, cuda-toolkit, docker, dockerfile, flask, llama3, llamacpp, llm, ngrok, pydantic, saiga
• Language: Python
• Homepage: https://vacancies-site.vercel.app/
• Size: 29.3 KB
• Stars: 1
• Watchers: 1
• Forks: 0
• Open Issues: 0
• Metadata Files:
  • Readme: README.md

Awesome Lists containing this project

README

        
# This is the server for the vacancies generation system

## About this server

This server works in pair with front-end part https://github.com/AlekseyScorpi/vacancies_site

It allows you to generate job texts at the request of users based on the form they submitted.

To generate text, the server uses a large language model. Model is finetuned Saiga3 - https://huggingface.co/AlekseyScorpi/saiga_llama3_vacancies_lora

## Installation

* You need to download any saiga_llama3_vacanices GGUF model from https://huggingface.co/AlekseyScorpi/saiga_llama3_vacancies_GGUF, create new folder and put the model in it

* Next you have to set model_config.json for your purpose or you can use my default config (default config assumes that you have created the generation_model folder and placed the model-Q5_K_M.gguf model in it)

* You need set several environments variables in .env file (create it):

  * NGROK_AUTH_TOKEN

  * NGROK_SERVER_DOMAIN

  * FLASK_PORT

  * CLIENT_DOMAIN

### Docker way

* You should be sure that you have the nvidia container toolkit installed to work with cuda inside containers

* Just build docker container with ```docker build -t {YOUR_IMAGE_NAME} . ```

* Run it 😉 , example run script ```docker run --gpus all -p 80:80 vacancies-server-saiga3```

### Default python way

* First, create new virtual environment and activate it (for this project python 3.11 is using)

* Second, set new environment variable ```ENV CMAKE_ARGS="-DLLAMA_CUDA=on"``` to build llama_cpp_python with CUDA (be sure, that you have CUDA Toolkit on your device)

* Then ```run pip install -r requirements.txt```

* Now you can run your server with ```python start.py```