and the EOS is "<turn|>". "<|channel>thought\n" is also used for the thinking tr... | Hacker News

Hacker Newsnew | past | comments | ask | show | jobs | submit

		trashcan2137 58 days ago \| parent \| context \| favorite \| on: Google releases Gemma 4 open models `and the EOS is "<turn\|>". "<\|channel>thought\n" is also used for the thinking trace!` Can someone explain this to me? Why is this faux-XML important here?

pertymcpert 58 days ago | [–]

That’s how the model is trained to signal the end to its generation and to indicate its thinking.

sroussey 58 days ago | [–]

These are likely individual tokens. They are super common.

Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact