Skip to content

Files

Latest commit

fddd626 · Dec 23, 2021

History

History
34 lines (21 loc) · 1.32 KB

README.md

File metadata and controls

34 lines (21 loc) · 1.32 KB

congrats_generator

Generator of the congratulations via web interface with web-scraper.

The generator originally works with pretrained model sberbank-ai/rugpt3small_based_on_gpt2 downloaded from huggingface.co.

The model fine tuning process is described in google colab notebook congrats_generator/my_gpt3.ipynb.

Web crawler is located in congrats_generator/crawler directory.

Stages to build and start up the generator:

  • Clone this repository into your local machine.

  • Make directory for the model and specify model and tokenizer path in congrats_generator/crawler/config.json. Put the model according to the path.

  • Build the app container. Start app process is realized with docker and docker compose services which allow the app to be more scalable further. You can use docker-compose build command to build your app, and docker-compose up to start the app.