TO DO
- basic functioning flask application to load model, get text and convert to SQL
- Search database based on schema and tables
- Added manual verification process before executing the SQL query
- change GPT to llama for sql generation
- create a select all and unselect all functionality based on schema and tables
- added few validations and preloader
- change the db creds to come from UI instead of .env
- fine-tune code-llama-instruct model with custom SQL datasets
- prepared dataset "shiroyasha13/llama_text_to_sql_dataset" (will increase data size)
- prepared the fine-tuning code and tested the functionality in colab
- fine-tuning "TheBloke/CodeLlama-7B-Instruct-GPTQ" (6-bit quuantized) model with SageMaker
- test the updated model and based on results upload to huggingface and use the model
- Replace the standard code-llama-instruct-7B with fine-tuned model
- store the response history for current session ( testing feasibility )
- Adding documentation to code and this repo