AI & ML interests

a repository to catalogue jsonify datasets which optimize a model for structured outputs for use with function calling, reward modeling, and efficiency.

Recent Activity

craffel  authored a paper about 1 month ago

TokSuite: Measuring the Impact of Tokenizer Choice on Language Model Behavior

craffel  authored a paper 8 months ago

FineWeb2: One Pipeline to Scale Them All -- Adapting Pre-Training Data Processing to Every Language

craffel  authored a paper 8 months ago

The Common Pile v0.1: An 8TB Dataset of Public Domain and Openly Licensed Text

View all activity

Team members 14