Open-Platypus
A carefully curated dataset of 25K STEM and logic questions assembled from 11 open datasets with strict deduplication. Fine-tuning Llama 2 on this dataset for just 5 hours achieves GPT-4-level STEM performance, demonstrating quality trumps quantity.
from datasets import load_dataset
ds = load_dataset("garage-bAInd/Open-Platypus")Preview a sample row
{
"instruction": "Prove that the sum of the first n natural numbers is n(n+1)/2 using mathematical induction.",
"output": "Base case: n=1, LHS=1, RHS=1(2)/2=1. ✓\nInductive step: Assume true for k. For k+1: sum = k(k+1)/2 + (k+1) = (k+1)(k/2+1) = (k+1)(k+2)/2. ✓",
"input": ""
}Fine-tune with this dataset
Estimated VRAM to fine-tune with QLoRA (4-bit base model + LoRA adapters), using conservative defaults:
New to fine-tuning? Follow the step-by-step walkthrough: Fine-Tune Your First LLM in 1 Hour
Frequently asked questions
Can I use Open-Platypus commercially?
Not in a product — Open-Platypus is released under CC BY NC 4.0, which restricts use to research and other non-commercial purposes. For commercial fine-tuning, pick a permissively licensed dataset from the same category instead.
How much data does Open-Platypus contain, and do I need all of it?
Open-Platypus contains 25K Questions. You rarely need all of it: for style and format fine-tuning, a few hundred to a few thousand examples are enough — load a slice (e.g. split="train[:1000]") and scale up only if quality plateaus.
What is Open-Platypus best used for?
A quick STEM and logic boost on a small budget (non-commercial). It belongs to the Instruction / SFT section of our dataset hub, where you'll find alternatives and complementary sets.