Team Name - Wingardium Laviousa
Folder structure:
- backend All Django backend code is present in this folder.
- frontend, React frontend code is present in this folder.
- notebooks, Jupyter notebooks for the LLM pipeline along with fine tuning dataset is present in this folder.
Google Login:
The application uses Google OAuth for signing in users.
To use this feature, create a project on google cloud api console and obtain CLIENT ID for web application.
The client id is to be inserted in google_auth/views.py and index.js files.
Postgres SQL:
The application is stored in postgreSQL for persistance. Any other database can be used. The config in settings.py needs to be updated accordingly.
Redis stack server:
The embedding of scrapped fashion trends are stored in redis vector store. To use this functionality, redis stack server needs to be installed.
python -m venv venv # use python version 3
source venv/bin/activate
pip install -r requirements
python manage.py migrate
python manage.py createsuperuser
Install and run redis stack server. This can be done using apt for linux systems.
Example Prompt for the web application:
I am a 22 year old male looking for a casual outfit.
Notebooks:
Formatted_Product-Metadata_trial.csv: Custom dataset created for finetuning Llama 2 model for better recommendations. The finetuned model is available atDakshi22/Flipkart_GridSQL+RAG.ipynb: The LLM pipeline used for generating recommendations using Retrieval Augmented Generation.tryfinetune.ipynb: Notebook used to finetune llama 2 model with custom dataset in file 1.vogue__bs_scrape.ipynb: Notebook used to scrap data from vogue india's web site.
