Codex Flashcard

Front

How much Python data was used for fine-tuning Codex after filtering?

Back

159 GB

Front

True or false: Researchers observed improvements in the fine-tuning process when starting from GPT.

Back

True

Front

What was the most notable tokenization technique mentioned in today's presentation?

Back

Special tokens for white space

Front

Codex-D specializes in generating...

Back

Docstrings

Front

True or false, Codex-S prefers a higher temperature value than vanilla Codex?

Back

True

Front

Fine-tuning Codex from a GPT base model provided...

Back

Faster convergence

Front

Which surprising capabilities did Codex exhibit?

Back

None of these silly answers, April Fools!