GPTeacher General-Instruct dataset is GPT-4 Generated self-instruct dataset.

There are multiple versions, with more or less similarity reductions.

The dedupe only dataset contains 18194 entries, with less the more similarity is reduced.

Format is identical to alpaca's, with a varyiable mix of Instruction/Input/Response, and Instruction/NullInput/Response fields.

Learn more on github here:
https://github.com/teknium1/GPTeacher