public-dataset-exploration

To find raw public datasets for forecasting, you can use this Skill to search Kaggle, Hugging Face, and GitHub for domain-relevant data. It shortlists candidates that are not already instruction-tuned, ensuring the raw data is suitable for conversion into training seeds.

The best way to convert raw data files into training seeds is by using the SDK's conversion utilities. This Skill helps you download small subsets from discovered datasets and assemble them into an input dataset for your downstream labeling and training pipelines.

Yes, you can use GitHub repositories for document-based training data discovery. This Skill explores GitHub alongside Kaggle and Hugging Face to find raw or semi-structured datasets that can be converted into samples for document-based training.

Yes, the dataset discovery process checks for instruction-tuned data. It actively filters out instruction/Q&A formatted or synthetic datasets, prioritizing raw sources that can produce forecasting questions or document-style Q&A for your training readiness.

You can select public datasets suitable for seed generation by having this Skill evaluate candidates based on relevance and reasonable licensing. It maps shortlisted datasets into a conversion-to-seeds workflow, ensuring they are raw and not already training-ready.