datasets
Training and evaluation data, with the modality, task and licence stated up front. Listed live from the Hugging Face Hub.
starcoderdata
StarCoder Training Dataset
Dataset description
This is the dataset used for training StarCoder and StarCoderBase. It contains 783GB of code in 86 programming languages, and includes 54GB GitHub Issues + 13GB Jupyter notebooks in scripts and text-code pairs,
and 32GB of GitHub commits, which is approximately 250 Billion tokens.
Dataset creation
The creation and filtering of The Stack is explained in the original dataset, we additionally decontaminate and… See the full description on the dataset page: https://huggingface.co/datasets/bigcode/starcoderdata.gpn-star-scores
GPN-Star genome-wide scores
Genome-wide, mutation-rate-calibrated GPN-Star constraint and variant scores for
eight score sets covering human, mouse, chicken, D. melanogaster,
C. elegans, and A. thaliana. Canonical scores are available as
chromosome-sharded Parquet. Hugging Face hosts 72 BigWigs; a multi-assembly
UCSC track hub references the 64 logo/LLR views.
Overview and quick links
Resource
Link
Files
Browse all dataset files
Default Dataset… See the full description on the dataset page: https://huggingface.co/datasets/songlab/gpn-star-scores.pa-warm-start-sft-heavy-25b-mix
geodesic-research/pa-warm-start-sft-heavy-25b-mix
Auto-generated by dataset-builder.
Each config below is a separate dataset produced from a versioned YAML build
config. Load with:
from datasets import load_dataset
ds = load_dataset("geodesic-research/pa-warm-start-sft-heavy-25b-mix", "<config_name>", revision="<commit-sha>")
Pin revision= to the specific commit SHA you want; without it, you get the
current HEAD of the dataset repo, which may change when the builder re-pushes.… See the full description on the dataset page: https://huggingface.co/datasets/geodesic-research/pa-warm-start-sft-heavy-25b-mix.the-join
The Join
A broad collection of relational databases spanning many domains (academic,
e-commerce, finance, sports, biomedical, government, text2sql, and more), ported to the
RelBench manifest format. The Join is built for pretraining relational/tabular foundation
models: each database is self-describing and tasks ship labels as-is for large-scale
pretraining rather than held-out benchmarking.
Each dataset lives in its own subdirectory in the self-describing manifest layout (plain… See the full description on the dataset page: https://huggingface.co/datasets/stanford-star/the-join.relbench-v1
RelBench v1 databases
The original RelBench v1 relational databases and tasks, each
database in its own subdirectory in the self-describing manifest layout (plain parquet +
manifest.yaml):
<dataset>/
manifest.yaml # tables, primary keys, foreign-key graph, val/test timestamps
schema.svg # ER diagram
db/*.parquet # relational tables (plain parquet)
tasks/<task>/manifest.yaml # task spec (+ duckdb SQL for `forecast`… See the full description on the dataset page: https://huggingface.co/datasets/stanford-star/relbench-v1.STARK_10k
STARK: Spatial-Temporal reAsoning benchmaRK
STARK is a comprehensive benchmark designed to systematically evaluate large language models (LLMs) and large reasoning models (LRMs) on spatial-temporal reasoning tasks, particularly for applications in cyber-physical systems (CPS) such as robotics, autonomous vehicles, and smart city infrastructure.
Dataset Summary
Hierarchical Benchmark: Tasks are structured across three levels of reasoning complexity:
State Estimation:… See the full description on the dataset page: https://huggingface.co/datasets/prquan/STARK_10k.redelex
CTU relational datasets (redelex)
Relational databases from the CTU Prague Relational Learning Repository (a.k.a. the CTU relational repository), ported to the RelBench manifest format via redelex. Each dataset lives in its own subdirectory in the self-describing manifest layout (plain parquet + manifest.yaml):
<dataset>/
manifest.yaml # tables, primary keys, foreign-key graph, val/test timestamps
schema.svg # zoomable ER diagram… See the full description on the dataset page: https://huggingface.co/datasets/stanford-star/redelex.starcoder2data-extras
StarCoder2 Extras
This is the dataset of extra sources (besides Stack v2 code data) used to train the StarCoder2 family of models. It contains the following subsets:
Kaggle (kaggle): Kaggle notebooks from Meta-Kaggle-Code dataset, converted to scripts and prefixed with information on the Kaggle datasets used in the notebook. The file headers have a similar format to Jupyter Structured but the code content is only one single script.
StackOverflow (stackoverflow): stackoverflow… See the full description on the dataset page: https://huggingface.co/datasets/bigcode/starcoder2data-extras.pa-warm-start-sft-xl-50b-mix
geodesic-research/pa-warm-start-sft-xl-50b-mix
Auto-generated by dataset-builder.
Each config below is a separate dataset produced from a versioned YAML build
config. Load with:
from datasets import load_dataset
ds = load_dataset("geodesic-research/pa-warm-start-sft-xl-50b-mix", "<config_name>", revision="<commit-sha>")
Pin revision= to the specific commit SHA you want; without it, you get the
current HEAD of the dataset repo, which may change when the builder re-pushes.… See the full description on the dataset page: https://huggingface.co/datasets/geodesic-research/pa-warm-start-sft-xl-50b-mix.mosaic-starcoder-filtered
Mosaic format for filtered starcoder dataset to train Malaysian LLM
This repository is to store dataset shards using mosaic format.
prepared at https://github.com/malaysia-ai/dedup-text-dataset/blob/main/pretrain-llm/combine-starcoder.ipynb
using tokenizer https://huggingface.co/malaysia-ai/bpe-tokenizer
4096 context length.
how-to
git clone,
git lfs clone https://huggingface.co/datasets/malaysia-ai/mosaic-starcoder-filtered
load it,
from streaming import… See the full description on the dataset page: https://huggingface.co/datasets/malaysia-ai/mosaic-starcoder-filtered.plurel-preprocessedstarcoderdata-python-edu-lang-score
Dataset Card for Starcoder Data with Python Education and Language Scores
Dataset Summary
The starcoderdata-python-edu-lang-score dataset contains the Python subset of the starcoderdata dataset. It augments the existing Python subset with features that assess the educational quality of code and classify the language of code comments. This dataset was created for high-quality Python education and language-based training, with a primary focus on facilitating models that can… See the full description on the dataset page: https://huggingface.co/datasets/JanSchTech/starcoderdata-python-edu-lang-score.FUSION-Finetune-12M
FUSION-12M Dataset
Please see paper & website for more information:
https://arxiv.org/abs/2504.09925
https://github.com/starriver030515/FUSION
Overview
FUSION-12M is a large-scale, diverse multimodal instruction-tuning dataset used to train FUSION-3B and FUSION-8B models. It builds upon Cambrian-1 by significantly expanding both the quantity and variety of data, particularly in areas such as OCR, mathematical reasoning, and synthetic high-quality Q&A data. The goal is… See the full description on the dataset page: https://huggingface.co/datasets/starriver030515/FUSION-Finetune-12M.stack-v2-starcoder2-3bstarrail-voice
StarRail Voice
StarRail Voice is a dataset of voice lines from the popular game Honkai: Star Rail.
Hugging Face 🤗 StarRail-Voice
ModelScope StarRail-Voice
Last update at 2026-07-16, game version 4.4.0
403437 wavs
60164 without speaker (15%)
61375 without transcription (15%)
57869 without inGameFilename (14%)
Dataset Details
Dataset Description
The dataset contains voice lines from the game's characters in multiple languages, including Chinese… See the full description on the dataset page: https://huggingface.co/datasets/simon3000/starrail-voice.pixelrag-tiles
PixelRAG tile corpus
Rendered screenshot tiles for PixelRAG, a visual retrieval-augmented-generation system that retrieves over page images instead of parsed text. Each Wikipedia page is rendered to an image and cut into fixed-height tiles; retrieval runs on the tiles directly with a Qwen3-VL embedding model.
This repository holds the full tile corpus that the published FAISS indexes and embeddings were built from, so the whole pipeline (tiles → embeddings → index → search) can… See the full description on the dataset page: https://huggingface.co/datasets/StarTrail-org/pixelrag-tiles.stark
STaRK
Website | Github | Paper
STaRK is a large-scale semi-structure retrieval benchmark on Textual and Relational Knowledge Bases
Downstream Task
Retrieval systems driven by LLMs are tasked with extracting relevant answers from a knowledge base in response to user queries. Each knowledge base is semi-structured, featuring large-scale relational data among entities and comprehensive textual information for each entity. We have constructed three knowledge bases: Amazon SKB… See the full description on the dataset page: https://huggingface.co/datasets/snap-stanford/stark.starcoderdata-python-edu
starcoderdata-python-edu
StarCoder Training Dataset Cleaned and Scored
Dataset Details
Dataset Description
This dataset is a filtered version of StarCoder Training Dataset
that has been scored with the python-edu-scorer.
Dataset Sources
Repository: https://huggingface.co/collections/HuggingFaceTB/smollm-6695016cad7167254ce15966
Paper: SmolLM - blazingly fast and remarkably powerful
Citation
@misc{allal2024SmolLM,
title={SmolLM… See the full description on the dataset page: https://huggingface.co/datasets/jon-tow/starcoderdata-python-edu.plurel
PluRel
A collection of 2000 synthetic relational databases (plurel-3000 … plurel-4999)
generated with PluRel. This is the collection used to pretrain the
Relational Transformer and obtain the
stanford-star/rt-plurel classification
and regression checkpoints.
Each database is a self-describing folder in the RelBench v3
manifest layout, so it loads with no PluRel dependency:
plurel-<seed>/
manifest.yaml # tables, primary keys, foreign-key graph, val/test… See the full description on the dataset page: https://huggingface.co/datasets/stanford-star/plurel.relbench-v2-extra
RelBench v2 extras
Everything RelBench v2 added on top of v1: the v2-only
databases (with their tasks), and the v2-only tasks defined on the v1 databases. The v1
databases and tasks stay in
stanford-star/relbench.
Same self-describing manifest layout (plain parquet + manifest.yaml):
<dataset>/
manifest.yaml # tables, primary keys, foreign-key graph, val/test timestamps
schema.svg # ER diagram
db/*.parquet # relational… See the full description on the dataset page: https://huggingface.co/datasets/stanford-star/relbench-v2-extra.starcoder_cleanedThis is starcoderdata, but with leading boilerplate text/license text removed, and with short sequences filtered out. It also removes the extra tags at the beginning of some of the files, like <reponame>.
FUSION-Pretrain-10M
FUSION-10M Dataset
Please see paper & website for more information:
https://arxiv.org/abs/2504.09925
https://github.com/starriver030515/FUSION
Overview
FUSION-10M is a large-scale, high-quality dataset of image-caption pairs used to pretrain FUSION-3B and FUSION-8B models. It builds upon established datasets such as LLaVA, ShareGPT4, and PixelProse. In addition, we synthesize 2 million task-specific image-caption pairs to further enrich the dataset. The goal of… See the full description on the dataset page: https://huggingface.co/datasets/starriver030515/FUSION-Pretrain-10M.starcoder-python-instruct
StarCoder-Python-Qwen-Instruct
Dataset Description
This dataset contains Python code samples paired with synthetically generated natural language instructions. It is designed for supervised fine-tuning of language models for code generation tasks. The dataset is derived from the Python subset of the bigcode/starcoderdata corpus, and the instructional text for each code sample was generated using the Qwen/Qwen3-Coder-30B-A3B-Instruct-FP8 model.
Creation… See the full description on the dataset page: https://huggingface.co/datasets/OLMo-Coding/starcoder-python-instruct.starlink-tle-latest
Latest Starlink & GPS TLEs
Credit: NASA
Part of the Orbital Mechanics Datasets collection on Hugging Face.
Dataset description
Latest Two-Line Element (TLE) orbital data for the Starlink and GPS constellations, sourced daily from CelesTrak.
Two-Line Element sets (TLEs) are the standard format for representing satellite orbital elements, developed by NORAD in the 1960s and still used universally today. Each TLE encodes six Keplerian orbital elements plus… See the full description on the dataset page: https://huggingface.co/datasets/juliensimon/starlink-tle-latest.STAR
STAR Scene Graph Dataset
STAR（Scene graph generaTion in lArge-size satellite imageRy）是面向大幅高分辨率卫星影像场景图生成的数据集。本仓库依据官方 SGG-ToolKit 发布的 STAR-SGG-with-attri.h5、taxonomy 和图片元数据构建。
图片作为仓库普通文件保存在 images/，Parquet 中的 image 字段是仓库相对路径，不包含图片字节，也不会自动解码为 Pillow 对象。完整仓库约 127 GB，下载前请确认磁盘空间。
Split
本仓库使用项目内固定的 H5 行号索引划分数据，而不是使用 H5 的 split 数值：
Hugging Face split
图片
对象
关系
train
771
131,470
203,889
validation
245
43,800
100,932
test
264
43,828
99,771
split 行总计
1,280… See the full description on the dataset page: https://huggingface.co/datasets/wliafe/STAR.pa-warm-start-sft-xl-smokechartverse-allSTARK_1k
Benchmarking Spatiotemporal Reasoning in Large Language Models: Capabilities and Challenges
Dataset for our paper: Benchmarking Spatiotemporal Reasoning in Large Language Models: Capabilities and Challenges
Contact Information
If you have any questions or feedback, feel free to reach out:
Name: Pengrui Quan
Email: prquan@ucla.edu
License
Copyright (c) 2025, UCLA Networked and Embedded Systems Laboratory (NESL)
All rights reserved.
Redistribution and use in… See the full description on the dataset page: https://huggingface.co/datasets/prquan/STARK_1k.starlink-ground-stations
Starlink Ground Stations & Points of Presence
Credit: NASA
Part of the Orbital Mechanics Datasets collection on Hugging Face.
Dataset description
Starlink ground infrastructure data: gateway earth stations and internet Points of Presence (PoPs). Gateway earth stations maintain continuous Ka-band and Ku-band links with the overhead satellite constellation -- when a user terminal communicates with a Starlink satellite, the signal is relayed down to the… See the full description on the dataset page: https://huggingface.co/datasets/juliensimon/starlink-ground-stations.svg-stack
Dataset Card for svg-stack
Dataset Description
This dataset contains SVG code examples for training and evaluating SVG models for image vectorization.
Dataset Structure
Features
The dataset contains the following fields:
Field Name
Description
Filename
Unique ID for each SVG
Svg
SVG code
Usage
from datasets import load_dataset
dataset = load_dataset("starvector/svg-stack")… See the full description on the dataset page: https://huggingface.co/datasets/starvector/svg-stack.