LLMDataHub is a repository collecting datasets for LLM training, focusing on chatbot and alignment datasets with cataloged examples and dataset links. It lists datasets by release period and provides structured categories.
Collecting history — the radar snapshots this repo daily. The trend line appears after 3 days of data (1 so far).
What it is
A repository that curates datasets for training LLMs, specifically targeting chatbot training. It provides links, size, language, usage notes, and short descriptions for datasets across alignment, domain-specific, pretraining, and multimodal categories.
How it works
The README organizes datasets by release timeframe (e.g., November 2023, September 2023, August 2023, etc.) and by type (SFT, RLHF, PT, CoT, etc.). It includes tabled entries with Dataset name, Used by, Type, Language, Size, Description. It also includes references to external sources (e.g., HuggingFace datasets) and example dataset descriptions.
Getting started
No explicit installation or setup instructions are shown in the provided README excerpt. If you want to explore datasets, you would navigate the listed dataset links within each section. The repository description points to a curated collection of datasets for chatbot training.
Recent releases
The latest releases section shows "latest 0" with no entries. Therefore, there are no listed recent releases in the provided data.
Traction
Stars: 3412. Forks: 234. Open issues: 5.
Behind the repo
No linked startup or company is provided in the excerpt. The project appears to be an open-source collection of datasets for LLM training contributed by individuals or researchers.
Caveats
License: MIT. Created: 2023-04-10. Last push: 2023-11-28. Language field not specified (listed as "-"). The README includes multiple dataset entries with descriptions and links to external sources; no explicit usage instructions beyond the dataset listings are present in the excerpt.