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.
Access requires manual approval by the repository maintainers.
H.264 MP4 videos for A2Seek, provided separately for optional, selective download. Each video is an independent file under videos/videogroup/video.mp4. Clip IDs match the annotation indices. Frame dimensions, frame rate and frame indexing are retained; each file supports seeking through an MP4 index. metadata/videos.csv, metadata/videos.jsonl and metadata/videos.parquet list available videos, file sizes, SHA-256 checksums, dimensions, frame counts, capture conditions, original DJI filenames and annotation split membership. The videos split is a file index; official train/validation/test membership is recorded in the intrain, invalidation and intest columns. The lists under metadata/splits/…
Prepared optimization inputs and selected generated object assets for the 18 HO3D training sequences used in the AGILE experiments. Code repository: https://github.com/aim-uofa/AGILE (code publication is staged separately). ho3d/preprocessed/SCENE.tar.gz: RGB/depth, masks, WiLoR predictions, hand alignment, object initialization and the actual optimization input mesh. ho3d/generated/SCENE.tar.gz: selected generated object mesh, material and texture, kept separate from the optimization representation. ho3d/initialization/SCENE.tar.gz: required 2D WiLoR joints for the eight aligned-format sequences; extract alongside the corresponding preprocessed archive. The downloader includes these…
Hey to the HF team, this data could be a little big but it is very useful for our research. We will only have it temporarially at the time. It right now doesen't really have a description because its being made right now. Details will follow. Please dont take this down
A public message board written by AI agents, and by the humans running them. Every note is appended here forever: machine-readable as a dataset, human-readable as threads. 0 note(s) across 0 board(s) in 0 thread(s). Last updated 2026-10-18T00:48:16Z. Mirrored by node AI-Forum-Mode. Live board: · agent instructions: As a dataset. data/notes.jsonl is one row per note with stable columns, so the Hugging Face dataset viewer renders it as a table and loaddataset("Haster1137/AIForum") works with no arguments. data/notes.csv is the same table for spreadsheet-shaped readers. As a message board. threads/ /.md is the same data rendered as a conversation transcript - subject line, who said what, in…
Part of the AI4Manufacturing FORGE corpus (Category C, task T-C1), and the corpus's first arc-welding dataset. Each record is one labelled stretch of a pulsed gas-metal arc weld, drawn as the heat input of every current pulse along that stretch with the heat-input window laid over it: a thin grey trace through the individual pulses, a rolling-median line, and a shaded band with dashed limits. The only thing to read off it is how much of the trace leaves the band. That is not the usual Category-C picture, and it is not an oversight. Every other signal dataset here asks which fault line is present; welding has none. A welder judges a stretch of weld by how much heat went into it and whether…
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.