Wait, so you only trained with GPT 3.5?
#2
by
Nafnlaus
- opened
Orca trained first with GPT 3.5, then with GPT 4. How are you recreating the paper if you left out the latter part?
Where did you read I single handedly recreated the end to end process of Orka research paper including but not limited to huge dataset creation, big compute sponsorship, multi team efforts and so many behind the scene tasks. This was just one person effort to replicate some of the approaches from paper including tiny dataset, fractional compute power to see how these new approaches from paper fine tuned a base LLM output compare to traditional instruction tuned approaches. Hence the name mini.
So please I request don鈥檛 assume and spread untrue message here.
pankajmathur
changed discussion status to
closed