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
3,164 datasets, sorted by most downloaded.
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.
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.
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.
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.
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.
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.
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.
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…
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…
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…
Model Collections
Hand-picked starting points, each with the reason it exists.
Collection · 4 entries
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.
Collection · 4 entries
Models that fit 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.
Collection · 6 entries
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.
Collection · 3 entries
Speech and audio models
Recognition and synthesis models, grouped so the two directions are easy to compare.
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
The same open-weight model priced by every host that serves it, per million tokens.
Research Hub
Data center trackers and maps
Moratoriums, permits, power, water and capital behind the facilities that run these models.
Method
How the Model Hub is built
Sources, evidence labels, refresh behaviour, and the limits of every comparison here.