README.md

metadata

language:
  - en
tags:
  - datasets
  - metadata
  - huggingface
  - training-data
  - model-cards
pretty_name: HF Datasets Master Index
size_categories:
  - 10k<n<100k

HF Datasets Master Index

A comprehensive rolled-together dataset of 4,810 unique datasets found across 227 HuggingFace profiles (90 people + 137 orgs).

What's Inside

This dataset consolidates all publicly available dataset references from:

1. Profile dataset pages - datasets published by each profile
2. Profile collections - datasets bookmarked in collections
3. Model cards - datasets referenced in model training descriptions
4. Model tree traversal - datasets inherited from base models through finetunes/merges/quants
5. Your personal collections - mc7ever's 52 HF collections

Columns

Column	Description
dataset	HuggingFace dataset ID (org/name)
profiles_using_it	Pipe-delimited list of profiles that use/reference this dataset
num_profiles	Number of profiles using it
collections_it_appears_in	Collections where this dataset appears
your_collection	Your personal collection name (if applicable)
direct_model_references	Number of models directly trained on this dataset
ancestral_model_references	Number of models that inherit this dataset through base model tree
total_model_tree_refs	Total model references (direct + ancestral)
direct_model_ids	Model IDs directly trained on this dataset
ancestral_model_ids	Model IDs that inherit this dataset
model_pipeline_types	Pipeline types of models using this dataset
total_downloads	Total downloads across all profiles
total_likes	Total likes across all profiles
found_via	How this dataset was discovered (profile_page, model_card, your_collection)

Stats

• 4,810 unique datasets
• 227 profiles scraped (90 people + 137 orgs)
• 5,041 models indexed (text + multimodal with text)
• 403 datasets appear in 2+ profiles
• 190 datasets in your personal collections

Usage

import pandas as pd

df = pd.read_parquet("hf_datasets_master.parquet")

# Most-used datasets
df.nlargest(10, 'num_profiles')[['dataset', 'num_profiles', 'total_model_tree_refs']]

# Datasets in your collections
df[df['your_collection'] != '']

# Datasets with most model references
df.nlargest(10, 'total_model_tree_refs')[['dataset', 'total_model_tree_refs']]