Dataset · vicgalle
52k Alpaca instructions with GPT-4 answers
vicgalle/alpaca-gpt4
52k English instruction-response pairs from GPT-4 based on Alpaca prompts for fine-tuning LLMs.
What's inside
The dataset contains 52k English instruction-following examples: prompts are taken from Alpaca, and responses are generated by GPT-4. This is a wrapper for compatibility with the Hugging Face datasets library; the original was published in the GPT-4-LLM repository. The format is text instruction/response pairs.
What you can do with it
- SFT fine-tuning of English instruct-LLMs
- Evaluating a model for instruction-following
- Comparison with Alpaca-like SFT sets
- Tasks
- Text generation, Question answering
- Languages
- English
- Size
- 10K–100K rows
- Downloads this month
- 7.7K
- Likes
- 338
Models trained on it
- mradermacher/gpt2-alpaca-gpt4-GGUF 3.2K
- mradermacher/Llama-3-8B-SFT-RuoZhiBa-i1-GGUF 2K
- mradermacher/Admiral-Llama-3-8B-GGUF 1.2K
- mradermacher/NinjaMouse-3B-40L-danube-i1-GGUF 1.1K
- vicgalle/gpt2-alpaca-gpt4Text generation 755
- mradermacher/Llama-3-8B-SFT-RuoZhiBa-GGUF 594
- cgus/NinjaMouse2-2.5B-v0.1-iMat-GGUF 568
- mradermacher/HamzahLMV2-1B-i1-GGUF 518
How to load it
from datasets import load_dataset
ds = load_dataset("vicgalle/alpaca-gpt4")