Here we convert several existing recipe ingredient and instructions datasets into dialogue. Each notebook processes a different dataset and creates a final dataset to be uploaded to huggingface.
## tasty_recipes.ipynb
Takes this Kaggle dataset 'Recipes from Tasty' https://www.kaggle.com/datasets/zeeenb/recipes-from-tasty?select=ingredient_and_instructions.json, filters for the top 1,000 highest rated recipes, and turns them into basic dialogue using a preset list of user prompt templates.
This dataset is nicely structured, and the ingredients section includes the quantities and units separated out. Some, but not all already include a primary_unit (US) and metric_unit. We could find all recipes with both units and generate dialogue for the prompt 'convert the ingredients into metric', 'what are the ingredients in UK measurements'? etc..