# Datasets, Page 56 of 132
Source: https://savrn.com/datasets/page/56
Markdown alternate of the page above; the site index is https://savrn.com/llms.txt

---

SAVRN Model Hub

# AI Training Datasets

Each dataset with its card, its structure (every split, row count and column type), its files, its license, and the models that disclose training on it.

11,122Models

3,164Datasets

285Papers

6,792Publishers

11,852Sourced relationships

Updated 2026-10-18 · [How the library is built](https://savrn.com/model-hub/methodology)

## Filter Datasets

Task

- [Text generation127](https://savrn.com/datasets?task=Text+generation)
- [Robotics106](https://savrn.com/datasets?task=Robotics)
- [Other48](https://savrn.com/datasets?task=Other)
- [Text classification47](https://savrn.com/datasets?task=Text+classification)
- [Text retrieval44](https://savrn.com/datasets?task=Text+retrieval)
- [Question answering36](https://savrn.com/datasets?task=Question+answering)
- [Speech recognition33](https://savrn.com/datasets?task=Speech+recognition)
- [Tabular classification21](https://savrn.com/datasets?task=Tabular+classification)
- [Time series forecasting21](https://savrn.com/datasets?task=Time+series+forecasting)
- [Image to text17](https://savrn.com/datasets?task=Image+to+text)
- [Image classification16](https://savrn.com/datasets?task=Image+classification)
- [Object detection14](https://savrn.com/datasets?task=Object+detection)
- [Text to speech14](https://savrn.com/datasets?task=Text+to+speech)
- [Visual question answering14](https://savrn.com/datasets?task=Visual+question+answering)
- [Image to image11](https://savrn.com/datasets?task=Image+to+image)
- [Image segmentation9](https://savrn.com/datasets?task=Image+segmentation)
- [Image to 3d9](https://savrn.com/datasets?task=Image+to+3d)
- [Translation9](https://savrn.com/datasets?task=Translation)
- [Video classification9](https://savrn.com/datasets?task=Video+classification)
- [Audio classification7](https://savrn.com/datasets?task=Audio+classification)
- [Feature extraction7](https://savrn.com/datasets?task=Feature+extraction)
- [Reinforcement learning7](https://savrn.com/datasets?task=Reinforcement+learning)
- [Text to image7](https://savrn.com/datasets?task=Text+to+image)
- [Text to video7](https://savrn.com/datasets?task=Text+to+video)

License

- [other250](https://savrn.com/datasets?license=other)
- [apache-2.0236](https://savrn.com/datasets?license=apache-2.0)
- [mit232](https://savrn.com/datasets?license=mit)
- [cc-by-4.0224](https://savrn.com/datasets?license=cc-by-4.0)
- [cc-by-nc-4.077](https://savrn.com/datasets?license=cc-by-nc-4.0)
- [cc0-1.030](https://savrn.com/datasets?license=cc0-1.0)
- [cc-by-sa-4.029](https://savrn.com/datasets?license=cc-by-sa-4.0)
- [odc-by24](https://savrn.com/datasets?license=odc-by)
- [cc-by-nc-sa-4.020](https://savrn.com/datasets?license=cc-by-nc-sa-4.0)
- [unknown14](https://savrn.com/datasets?license=unknown)
- [cc-by-sa-3.06](https://savrn.com/datasets?license=cc-by-sa-3.0)
- [odbl6](https://savrn.com/datasets?license=odbl)
- [gpl-3.05](https://savrn.com/datasets?license=gpl-3.0)
- [cc-by-3.04](https://savrn.com/datasets?license=cc-by-3.0)
- [cc3](https://savrn.com/datasets?license=cc)
- [cc-by-nc-nd-4.03](https://savrn.com/datasets?license=cc-by-nc-nd-4.0)
- [gpl3](https://savrn.com/datasets?license=gpl)
- [cc-by-2.02](https://savrn.com/datasets?license=cc-by-2.0)
- [cc-by-nd-4.02](https://savrn.com/datasets?license=cc-by-nd-4.0)
- [cdla-sharing-1.02](https://savrn.com/datasets?license=cdla-sharing-1.0)
- [agpl-3.01](https://savrn.com/datasets?license=agpl-3.0)
- [bsl-1.01](https://savrn.com/datasets?license=bsl-1.0)
- [c-uda1](https://savrn.com/datasets?license=c-uda)
- [cc-by-nc-3.01](https://savrn.com/datasets?license=cc-by-nc-3.0)

Access

- [Publicly accessible2,933](https://savrn.com/datasets?access=open_access)
- [Access requested at publisher231](https://savrn.com/datasets?access=gated)

Publisher

- [Benjamin J Barber35](https://savrn.com/datasets?publisher=endomorphosis)
- [Chang Yu24](https://savrn.com/datasets?publisher=g1n0st)
- [Ji Xie23](https://savrn.com/datasets?publisher=sanaka87)
- [Justicedao18](https://savrn.com/datasets?publisher=justicedao)
- [LeDuy13](https://savrn.com/datasets?publisher=duyle2408)
- [Nguyen13](https://savrn.com/datasets?publisher=icdn11)
- [Nguyen13](https://savrn.com/datasets?publisher=icdn12)
- [Nguyen13](https://savrn.com/datasets?publisher=icdn13)
- [National Engineering Research Center for Big Data Software13](https://savrn.com/datasets?publisher=thulab)
- [Nguyen12](https://savrn.com/datasets?publisher=icdn10)
- [Nguyen12](https://savrn.com/datasets?publisher=icdn14)
- [Byte I12](https://savrn.com/datasets?publisher=icdn15)
- [Rebrowser.net12](https://savrn.com/datasets?publisher=rebrowser)
- [Shufan Jiang12](https://savrn.com/datasets?publisher=tsumugii)
- [Ai210](https://savrn.com/datasets?publisher=allenai)
- [Ps Yo10](https://savrn.com/datasets?publisher=atokforps)
- [Dougall DeepMind10](https://savrn.com/datasets?publisher=dougalldeepmind)
- [Deesced9](https://savrn.com/datasets?publisher=ameluce)
- [Jiahao Zhang9](https://savrn.com/datasets?publisher=hi-todayis-jh)
- [Julia Dollis9](https://savrn.com/datasets?publisher=juliadollis)
- [Massive Text Embedding Benchmark9](https://savrn.com/datasets?publisher=mteb)
- [Pranab Pattanaik9](https://savrn.com/datasets?publisher=p2samapa)
- [Sota9](https://savrn.com/datasets?publisher=sotaagi2030)
- [Sunsetcanyon9](https://savrn.com/datasets?publisher=sunsetcanyon)

Size

- [10K<n<100K125](https://savrn.com/datasets?size=10K%3Cn%3C100K)
- [1K<n<10K106](https://savrn.com/datasets?size=1K%3Cn%3C10K)
- [n<1K100](https://savrn.com/datasets?size=n%3C1K)
- [100K<n<1M70](https://savrn.com/datasets?size=100K%3Cn%3C1M)
- [1M<n<10M51](https://savrn.com/datasets?size=1M%3Cn%3C10M)
- [10M<n<100M33](https://savrn.com/datasets?size=10M%3Cn%3C100M)
- [100M<n<1B21](https://savrn.com/datasets?size=100M%3Cn%3C1B)
- [n>1T16](https://savrn.com/datasets?size=n%3E1T)
- [1B<n<10B8](https://savrn.com/datasets?size=1B%3Cn%3C10B)
- [100B<n<1T5](https://savrn.com/datasets?size=100B%3Cn%3C1T)
- [10B<n<100B2](https://savrn.com/datasets?size=10B%3Cn%3C100B)
- [10GB<n<100GB1](https://savrn.com/datasets?size=10GB%3Cn%3C100GB)
- [120,8981](https://savrn.com/datasets?size=120%2C898)
- [1B-tokens1](https://savrn.com/datasets?size=1B-tokens)
- [1T<n<10T1](https://savrn.com/datasets?size=1T%3Cn%3C10T)
- [n1K<n10K1](https://savrn.com/datasets?size=n1K%3Cn10K)
- [n<100K1](https://savrn.com/datasets?size=n%3C100K)
- [n<10K1](https://savrn.com/datasets?size=n%3C10K)
- [n<1GB1](https://savrn.com/datasets?size=n%3C1GB)
- [n>1M1](https://savrn.com/datasets?size=n%3E1M)

Domain

- [LeRobot31](https://savrn.com/datasets?domain=LeRobot)
- [legal, law, graphrag12](https://savrn.com/datasets?domain=legal%2C+law%2C+graphrag)
- [tsfile, timeseries, time-series12](https://savrn.com/datasets?domain=tsfile%2C+timeseries%2C+time-series)
- [open-data, data-science, web-scraping9](https://savrn.com/datasets?domain=open-data%2C+data-science%2C+web-scraping)
- [seo-cluster, web-nodes9](https://savrn.com/datasets?domain=seo-cluster%2C+web-nodes)
- [dpo, rlhf, ultrafeedback8](https://savrn.com/datasets?domain=dpo%2C+rlhf%2C+ultrafeedback)
- [mteb, text, image7](https://savrn.com/datasets?domain=mteb%2C+text%2C+image)
- [nepali, instruction-tuning, agriculture7](https://savrn.com/datasets?domain=nepali%2C+instruction-tuning%2C+agriculture)
- [science, stackexchange, provenance5](https://savrn.com/datasets?domain=science%2C+stackexchange%2C+provenance)
- [LeRobot, robotics, lerobot4](https://savrn.com/datasets?domain=LeRobot%2C+robotics%2C+lerobot)
- [ml-intern4](https://savrn.com/datasets?domain=ml-intern)
- [AMASS, Retarget, Robotics3](https://savrn.com/datasets?domain=AMASS%2C+Retarget%2C+Robotics)
- [benchmark, agents, terminal3](https://savrn.com/datasets?domain=benchmark%2C+agents%2C+terminal)
- [climate3](https://savrn.com/datasets?domain=climate)
- [code3](https://savrn.com/datasets?domain=code)
- [crypto, cryptocurrency, finance3](https://savrn.com/datasets?domain=crypto%2C+cryptocurrency%2C+finance)
- [data-attribution, influence-functions, interpretability3](https://savrn.com/datasets?domain=data-attribution%2C+influence-functions%2C+interpretability)
- [fireviewer, fire, smoke3](https://savrn.com/datasets?domain=fireviewer%2C+fire%2C+smoke)
- [video3](https://savrn.com/datasets?domain=video)
- [LeRobot, lerobot, grabette2](https://savrn.com/datasets?domain=LeRobot%2C+lerobot%2C+grabette)
- [LeRobot, ur7e, robotiq2](https://savrn.com/datasets?domain=LeRobot%2C+ur7e%2C+robotiq)
- [agentenv, rl-environment, agents2](https://savrn.com/datasets?domain=agentenv%2C+rl-environment%2C+agents)
- [ai-governance, measurement2](https://savrn.com/datasets?domain=ai-governance%2C+measurement)
- [autonomous, agents, evolution2](https://savrn.com/datasets?domain=autonomous%2C+agents%2C+evolution)

[Most downloaded](https://savrn.com/datasets?sort=downloads)[Most liked](https://savrn.com/datasets?sort=likes)[Recently updated](https://savrn.com/datasets?sort=updated)[Name](https://savrn.com/datasets?sort=name)

3,164 datasets, sorted by most downloaded.

Dataset

### [satdashboard-db](https://savrn.com/datasets/satdashboard-db)

[Rhyno Wu](https://savrn.com/model-publishers/rhynowu)

Publicly accessible

[View dataset](https://savrn.com/datasets/satdashboard-db)

Dataset

### [bible-sphere-stats](https://savrn.com/datasets/bible-sphere-stats)

[Zhao](https://savrn.com/model-publishers/stephenzao)

Publicly accessible

[View dataset](https://savrn.com/datasets/bible-sphere-stats)

Dataset

### [telematics](https://savrn.com/datasets/telematics)

[Do Gia Ngan](https://savrn.com/model-publishers/ngando49081)

Publicly accessible

[View dataset](https://savrn.com/datasets/telematics)

Dataset

### [EngiWorld-Images](https://savrn.com/datasets/engiworld-images)

[Hongcheng Gao](https://savrn.com/model-publishers/hongchenggao)

This repository contains virtual machine images for the EngiWorld benchmark, covering 15 open-source engineering software environments. You can find more information in our paper, EngiWorld: What Can Frontier Agents Deliver in Professional Engineering Environments? Due to software licensing considerations, we currently provide public environment images only for open-source software. For environment images containing proprietary software, please contact us; access is subject to the applicable licenses.

Publicly accessible

[View dataset](https://savrn.com/datasets/engiworld-images)

Dataset

### [k2-dashboard](https://savrn.com/datasets/k2-dashboard)

[Martin Bollen](https://savrn.com/model-publishers/blattellus)

Publicly accessible

[View dataset](https://savrn.com/datasets/k2-dashboard)

Dataset

### [zeamed-app-log-backups](https://savrn.com/datasets/zeamed-app-log-backups)

[ZeaHealth, Inc](https://savrn.com/model-publishers/zeahealth)

Publicly accessible

[View dataset](https://savrn.com/datasets/zeamed-app-log-backups)

Dataset

### [fshare-backup-db](https://savrn.com/datasets/fshare-backup-db)

[Diva](https://savrn.com/model-publishers/divaashley9283)

Publicly accessible

[View dataset](https://savrn.com/datasets/fshare-backup-db)

Dataset

### [dublin-bikes-weather-raw](https://savrn.com/datasets/dublin-bikes-weather-raw)

[Malav naik](https://savrn.com/model-publishers/malavnaik)

Raw, untouched snapshots collected every ~15 minutes by dublin-data-pipeline. Files are only ever added, never changed. fetchedat is UTC. Weather rows repeat across hourly files; deduplicate on date + reportTime.

Publicly accessible cc-by-4.0

[View dataset](https://savrn.com/datasets/dublin-bikes-weather-raw)

Dataset

### [abyssal](https://savrn.com/datasets/abyssal)

[Vo Huu Phuong](https://savrn.com/model-publishers/phuongvo92194)

Publicly accessible

[View dataset](https://savrn.com/datasets/abyssal)

Dataset · Video text to text

### [openvid-256](https://savrn.com/datasets/openvid-256)

[Tristen Jackson](https://savrn.com/model-publishers/tristenjackson)

Derived from nkp37/OpenVid-1M (https://huggingface.co/datasets/nkp37/OpenVid-1M), released under CC-BY-4.0 for research and non-commercial purposes. Videos were collected by its authors from other publicly available datasets, whose own terms may also apply. Attribution: "OpenVid-1M: A Large-Scale High-Quality Dataset for Text-to-video Generation" (ICLR 2025), NJU-PCALab. License: https://creativecommons.org/licenses/by/4.0/ Modifications: cut/cropped into short clips, resized, re-encoded (H.264), audio removed.

Publicly accessible cc-by-4.0

[View dataset](https://savrn.com/datasets/openvid-256)

Dataset

### [role-play-data](https://savrn.com/datasets/role-play-data)

[Dasodefa](https://savrn.com/model-publishers/dasodefa)

Publicly accessible

[View dataset](https://savrn.com/datasets/role-play-data)

Dataset

### [kinex-v0.10.0-benchmark](https://savrn.com/datasets/kinex-v0-10-0-benchmark)

[RLE-Bench](https://savrn.com/model-publishers/rle-bench)

instructions. 82 selected tasks across 5 environments. 72 recorded results and 10 pending tasks. Each evaluated task shows its latest result. Modified instructions are labeled and edits highlighted. Native outcome, execution, evidence validity and usage coverage are reported separately. Pending tasks have no synthetic scores.

Publicly accessible n<1K

[View dataset](https://savrn.com/datasets/kinex-v0-10-0-benchmark)

Dataset

### [maxrl_dppo_qwen3_4B_maxrl_polaris1000_knk_rollouts](https://savrn.com/datasets/maxrl-dppo-qwen3-4b-maxrl-polaris1000-knk-rollouts)

[Fahim Tajwar](https://savrn.com/model-publishers/ftajwar)

Training rollouts of one verl run, one parquet per trainer step (data/step-.parquet), rows in the order the trainer consumed them; the rollout.n samples of a prompt are contiguous and share uid. prompttokenids / responsetokenids are the exact token ids (special tokens kept, no padding) and rolloutlogprobs the sampling-time log-prob of every response token. input / output are decoded with skipspecialtokens=True and are for reading only. score is the raw reward.

Publicly accessible

[View dataset](https://savrn.com/datasets/maxrl-dppo-qwen3-4b-maxrl-polaris1000-knk-rollouts)

Dataset

### [grpo_dppo_qwen3_4B_base_knk_rollouts](https://savrn.com/datasets/grpo-dppo-qwen3-4b-base-knk-rollouts)

[Fahim Tajwar](https://savrn.com/model-publishers/ftajwar)

Training rollouts of one verl run, one parquet per trainer step (data/step-.parquet), rows in the order the trainer consumed them; the rollout.n samples of a prompt are contiguous and share uid. prompttokenids / responsetokenids are the exact token ids (special tokens kept, no padding) and rolloutlogprobs the sampling-time log-prob of every response token. input / output are decoded with skipspecialtokens=True and are for reading only. score is the raw reward.

Publicly accessible

[View dataset](https://savrn.com/datasets/grpo-dppo-qwen3-4b-base-knk-rollouts)

Dataset · Tabular regression

### [metabolome-phenome-atlas](https://savrn.com/datasets/metabolome-phenome-atlas)

[Shiroha](https://savrn.com/model-publishers/shir0ha)

Public GWAS.linear summary files originally hosted on Alibaba Cloud OSS (metabolome-phenome-atlas, oss-cn-hongkong). Migrated to Hugging Face so downloads no longer consume OSS egress. Each file is one phenotype, ~631 MB. Paths keep the original Summarydata/ prefix.

Publicly accessible other 100M<n<1B

[View dataset](https://savrn.com/datasets/metabolome-phenome-atlas)

Dataset

### [brats-qos-hrgn-graphs](https://savrn.com/datasets/brats-qos-hrgn-graphs)

[Adarsh Binu](https://savrn.com/model-publishers/marvelpokemaster)

Publicly accessible

[View dataset](https://savrn.com/datasets/brats-qos-hrgn-graphs)

Dataset · Image to 3d

### [Alchemy3D-1M](https://savrn.com/datasets/alchemy3d-1m)

[Badi Li](https://savrn.com/model-publishers/libadi)

Alchemy3D-1M is the training corpus for Alchemy3D: source–target 3D editing pairs on PBR-ready assets, with text instructions and image conditions. It contains 1.25M unique assets and 1.38M editing pairs across 7 edit types. Due to Hugging Face storage limits, this repository includes only training metadata, structured latents at resolution 512, and rendered conditions. For GLBs and latents at resolution 1024, please use the ModelScope counterpart Alchemy3D-1M. The dataset contain encoded latents for sparse structure (sslatents), geometry (shapelatents/512), PBR material (pbrlatents/512). csv files are metadata. Latents are packed as partXXX.zip shards. Then unzip all zip files. Apache…

Publicly accessible apache-2.0 1M<n<10M

[View dataset](https://savrn.com/datasets/alchemy3d-1m)

Dataset

### [11](https://savrn.com/datasets/11)

[Anderson Freitas Torres](https://savrn.com/model-publishers/bket11)

Publicly accessible

[View dataset](https://savrn.com/datasets/11)

Dataset · Text retrieval

### [ipfs_albania_laws_ir](https://savrn.com/datasets/ipfs-albania-laws-ir)

[Justicedao](https://savrn.com/model-publishers/justicedao)

Research retrieval release of endomorphosis/ipfsalbanialaws (revision 35e648331c43d903f9008d4cbdde47982eb46996) packaged as country-laws-ir-graphrag/v1 (layout family skillcenter-huggingface-release/v3 / publicus-ir). Not legal advice. This is a research snapshot. The official gazette / authentic source of Albania prevails over this corpus. Retrieved documents and graph edges are retrieval evidence only. No legal text was invented. Primary key: entrycid (CIDv1 raw sha2-256 of a canonical identity record). Integer documentindex values are compact shard pointers, not identities. Target Hub id (packaging metadata only): justicedao/ipfsalbanialawsir. entrycid, lawcid, recordtype, jurisdiction…

Publicly accessible other 100K<n<1M

[View dataset](https://savrn.com/datasets/ipfs-albania-laws-ir)

Dataset

### [Niji_1_Boy-control](https://savrn.com/datasets/niji-1-boy-control)

[D16](https://savrn.com/model-publishers/dsixteen)

Publicly accessible

[View dataset](https://savrn.com/datasets/niji-1-boy-control)

Dataset

### [coast](https://savrn.com/datasets/coast)

[Raymond Louis Murray](https://savrn.com/model-publishers/raymondlmurray9621)

Publicly accessible

[View dataset](https://savrn.com/datasets/coast)

Dataset

### [0a20826dfc1b1132](https://savrn.com/datasets/0a20826dfc1b1132)

[Howler](https://savrn.com/model-publishers/sam04)

Publicly accessible

[View dataset](https://savrn.com/datasets/0a20826dfc1b1132)

Dataset · Text retrieval

### [genai-image-tag-db-CC4](https://savrn.com/datasets/genai-image-tag-db-cc4)

[NEXTAltair](https://savrn.com/model-publishers/nextaltair)

This repository contains the cc-by-4.0 build of the tag database. The main artifact is the SQLite database. The parquetdanbooru/ directory is a derived export so the Hugging Face Dataset Viewer can preview a subset of rows (Danbooru-only). - genai-image-tag-db-cc4.sqlite: SQLite database - parquetdanbooru/.parquet: Parquet export for Dataset Viewer - buildmanifest.json: Build manifest (revisions and stats) - report/: Source effects and health checks Only sources with dbchanges > 0 are listed. - sitetags (cc-by-4.0) https://huggingface.co/datasets/deepghs/sitetags (externalsources/sitetags/anime-pictures.net/tags.sqlite) - sitetags (cc-by-4.0) https://huggingface.co/datasets/deepghs/sitetags…

Publicly accessible cc-by-4.0 1M<n<10M

[View dataset](https://savrn.com/datasets/genai-image-tag-db-cc4)

Dataset

### [Map](https://savrn.com/datasets/map)

[Enyu Chen](https://savrn.com/model-publishers/enyu913)

Publicly accessible

[View dataset](https://savrn.com/datasets/map)

## Model Collections

Hand-picked starting points, each with the reason it exists.

Collection · 4 entries

### [Embedding models for retrieval](https://savrn.com/collections/embedding-models-for-retrieval)

Sentence and document embedding models used to build retrieval systems. Dimension and sequence length matter more than size here, and both come from the publisher.

[View collection](https://savrn.com/collections/embedding-models-for-retrieval)

Collection · 4 entries

### [Models that fit on one accelerator](https://savrn.com/collections/models-that-run-on-one-accelerator)

Models whose publisher-reported parameter count puts them within reach of a single accelerator at common precisions. Memory needed depends on precision and serving configuration, so treat the parameter count as the starting point, not the answer.

[View collection](https://savrn.com/collections/models-that-run-on-one-accelerator)

Collection · 6 entries

### [Open-weight text models worth knowing](https://savrn.com/collections/open-weight-text-models-worth-knowing)

Widely used open-weight language models, chosen because each one is a distinct family rather than a variant of the one above it. Selection, not a ranking.

[View collection](https://savrn.com/collections/open-weight-text-models-worth-knowing)

Collection · 3 entries

### [Speech and audio models](https://savrn.com/collections/speech-and-audio)

Recognition and synthesis models, grouped so the two directions are easy to compare.

[View collection](https://savrn.com/collections/speech-and-audio)

## Related SAVRN Research

The hub sits beside SAVRN's market data and infrastructure research: what models cost to run, and what it takes to run them.

SAVRN Index

### [What open models cost to run](https://savrn.com/ai-index/pricing/open-models)

The same open-weight model priced by every host that serves it, per million tokens.

Research Hub

### [Data center trackers and maps](https://savrn.com/data-center-research)

Moratoriums, permits, power, water and capital behind the facilities that run these models.

Method

### [How the Model Hub is built](https://savrn.com/model-hub/methodology)

Sources, evidence labels, refresh behaviour, and the limits of every comparison here.
