Dataset · timdettmers
Top Open Assistant dialogues for fine-tuning
timdettmers/openassistant-guanaco
A collection of 9,846 highly rated Open Assistant dialogues for instruction tuning.
What's inside
A text dataset and Open Assistant subset containing only the highest-rated paths in the dialogue tree, with a total of 9,846 examples. Used to train Guanaco via QLoRA under the Apache 2.0 license.
What you can do with it
- Train an SFT model on dialogues
- Fine-tune LoRA for instruction following
- Evaluate assistant response quality
- Size
- 10K–100K rows
- Downloads this month
- 15.4K
- Likes
- 456
Models trained on it
- bartowski/Tralalabs_Pythia-6.9B-Instruct-v1-Merged-GGUFText generation 1.8K
- mradermacher/mpt-7b-8k-chat-GGUF 1.1K
- shiva123782/Shiva-Llama-Instruct-v1Text generation 844
- mradermacher/mpt-7b-8k-chat-i1-GGUF 834
- TinyLlama/TinyLlama-1.1B-Chat-v0.1Text generation 786
- mradermacher/mpt-30b-chat-GGUF 727
- mradermacher/mpt-30b-chat-i1-GGUF 416
- mradermacher/croguana-RC2-GGUF 280
How to load it
from datasets import load_dataset
ds = load_dataset("timdettmers/openassistant-guanaco")