Datasets:
Request access to Chat/instruction data (filtered)
Chat/instruction data (filtered) is released by The Fin AI for research. Access is granted automatically after you complete this short form.
By accessing this dataset you agree to its license and to cite the The Fin AI paper in any resulting publication. Use of the content is also subject to the terms of the original source.
Log in or Sign Up to review the conditions and access this dataset content.
Chat/instruction data (filtered)
Pretraining / reference corpus released by The Fin AI. Source: Instruction-tuning mixtures (e.g. Alpaca, SIQA, OASST1, GSM8K, Super-NaturalInstructions).
Source
A mixture of public instruction datasets; licensed CC BY-NC 4.0 because it includes Alpaca (non-commercial).
Structure
- Rows: 3,106,579
- Columns:
source,date,text,token_count,category
Quick Start
from datasets import load_dataset
ds = load_dataset("TheFinAI/chat_filtered", split="train", streaming=True)
print(next(iter(ds)))
License
CC BY NC 4.0, following the original source. Please attribute the original source.
- Downloads last month
- 32