Changed column names & added example to dbpedia

[rohanoai]

Can now call openai tools fine_tunes.prepare_data -f dbpedia_samples.jsonl without bugs.

[shyamal-anadkat]

@rohanoai do you still want to proceed with this change?

[ted-at-openai]

@rohanoai do you still want to proceed with this change?

My hesitation in merging this change is that it turns descriptive column names (text, category) into abstract fine-tuning-specific column names (prompt, completion). Given that our current example doesn't use fine-tuning, it feels like it might confuse folks if they import a data table that names categories as completions. They may wonder, "huh, is this machine generated data? Why is it called completions?" without realizing that it's written as a fine-tuning dataset.

My preference would be to keep the adjustment to the data class representation (so that it's less likely to cause problems with fine-tuning should someone wish to) but not to change the column names to prompt & completion.

Thoughts?