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.
Updated 2026-10-18 · How the library is built
3,164 datasets, sorted by most downloaded.
This dataset contains the public Gravitational Wave Open Science Center O4a strain release at 16,384 Hz for H1, L1. Each detector is represented independently. A contiguous source span produces three Parquet files: Strain, DQmask, and Injmask. This research has made use of data or software obtained from the Gravitational Wave Open Science Center (gwosc.org), a service of the LIGO Scientific Collaboration, the Virgo Collaboration, and KAGRA. This material is based upon work supported by NSF's LIGO Laboratory which is a major facility fully funded by the National Science Foundation, as well as the Science and Technology Facilities Council (STFC) of the United Kingdom, the Max-Planck-Society…
A dated record of what each channel actually grants on the free tier of AI video models — daily credits, daily generation caps, sign-up bonuses, clip length, extension ceiling, output resolution, watermark behaviour, and API availability. Free-tier terms are the least documented part of an AI video product. Vendors publish a launch post with a headline number, then adjust the daily grant, move the watermark switch, or gate a model behind a membership tier. This dataset collects those figures in one place and stamps every row with the channel it came from and the date it was observed. 66 observations across Seedance 2.5, PixVerse, Runway, Wan 2.5, Kling AI, Sora 2 and Grok Imagine, current…
Built from the URL manifest at thelarryproject/scotus-manifest. One row per oral argument session of the U.S. Supreme Court, October Terms 2010–2025, materializing the official audio recording and transcript published at supremecourt.gov. Contains ~1,215 hours of audio un-pruned. Some amount of work has been performed to format the PDF text into a transcript consisting of a series of speaker turns. - docket: docket number; rearguments and multi-session arguments carry suffixes (rearg, q1, q2) - term: October Term the argument belongs to - dateargued: ISO date of the argument session - pageurl: url for transcript and mp3 for a term and docket, in the format…
This repo holds everything needed to train and run the YouCook2 event segmentation + captioning model in im-xiaoming/UniAV-fixed. The model finds the steps of an uncut cooking video and writes one caption per step. YouCook2 (Zhou et al., AAAI 2018) has 2,000 cooking videos of 89 recipes, with step segments and one English sentence per step. 1,500 of the train + validation videos are still downloadable: 1,106 are used for training and 394 for validation. Annotations and caption vectors are in the git repo (data/youcookii/), not here. Each.npz holds three float16 arrays, with row i centred on i + 0.5 s: Extracted with tools/extractinternvideo2.py; text vectors with tools/embedcaptions.py…
A philosophy-focused scholarly corpus derived from the OpenAlex snapshot mirrored by Mearman/OpenAlex. The classifier processed 821,224 unique candidate works. There are 0 duplicate work IDs. 450,602 works Production-oriented high-confidence search index. Contains CORE and PROBABLE works while excluding LOWQUALITY and PARATEXT document roles. 450,602 works — approximately 23 MiB. Compact representation of the complete classified candidate corpus, without the full reconstructed abstract payload. Approximately 48 MiB. Complete V3.1 classification output, including reconstructed abstract evidence where available. Approximately 267 MiB. Machine-readable corpus statistics. The corpus is not…
Daily-updated measurements of internet round-trip latency toward submarine cable landing regions, and detected routing anomalies (detours, path changes), collected by GeoCables - a live atlas of the world's submarine cable infrastructure with its own distributed measurement network. - geocables-latency-daily.csv - per-day latency aggregates (median / p90 / min RTT in milliseconds) for measured city-to-target routes. - geocables-route-events.csv - detected routing events: detours, reroutes and recoveries, with distance and RTT deltas (IP addresses scrubbed). Measurements are performed continuously by GeoCables' own distributed probe network; values are aggregated daily. Event detection…
Research retrieval release of endomorphosis/ipfsbolivialaws (revision 52464e686d9edf46e207b546b1b02ff7cd2d0041) 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 Bolivia 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/ipfsbolivialawsir. entrycid, lawcid, recordtype, jurisdiction…
Narrated long-form lectures on statistics and machine learning, streamed by the Papers, as Audio phone app. - library.json — the catalog the app reads (titles, durations, chapter markers, file paths) - audio/.mp3 — one file per lecture (Kokoro-82M afheart, 96 kbps mono) - lectures/.md — transcripts Published automatically from https://github.com/Arjun10g/PapersAudio.
A fine-tuning corpus for Nepali (Devanagari) government-document OCR and layout analysis, built to adapt a vision-language model to Nepali paperwork without erasing what it already knows. Roughly half is the target task; the other half is replay data, mixed in deliberately. That split follows Fine-tuning VLMs Without Forgetting, whose one unresolved failure mode is task-specific overfitting — the model memorises a single output template and stops reading the prompt. Diverse tasks are the measured remedy, so training only on Nepali documents would walk straight into it. pages / crops / organic carry the full Page IR: replay are uniform: image, prompt, target, source, config. Table format is…
DeepSeek V4.1 Flash's Undead army meets an identical one played by Warcraft's own attack-move. The plan it wrote at that moment is over the fight. It wins in 54 seconds, keeping 79% of its army. This is the game's own picture, rendered from the run's replay. Every run here plays like this in the Space, with the env's map beside it. LLM agents play the real Warcraft III: The Frozen Throne (1.29) through per player. A task is played in one of three styles, each its own env or agent: - raw (env wc3): one model orders units by id through general tools: it reads the game with getstate, listunits and lookup, gives orders with act and moves game time on with advance. - commander (env…
Every product of the Congressional Research Service (CRS) that the Congress.gov API lists, active and archived, with full text and metadata: Reports, Posts, Resources, Testimony and Infographics, as the API names them. CRS writes them for Members of Congress; Congress.gov publishes them. Nothing here is edited by hand. The pipeline, its tests and its schedule are in github.com/incrediblecrab/congressional-research-service-products, and this card is rendered from manifest.json in the same commit. 11,428 of 14,143 products (80.8%) as of 2026-09-25T03:27:42Z UTC, when the API last listed 14,143. 11,420 rows have text. 0 products failed 3 times and are retried every 24 hours; their errors are…
Pre-computed High-Dimensional Forensic Latents for ultra-fast Diffusion-Diff forensics training. - imgid (string): Image identifier. - label (int64): Class label (0: authentic, 1: fully synthetic, 2: tampered/inpainted). - zhighdim (binary): Pre-computed 84-channel fp16 latent tensor bytes. - mask (binary): Ground-truth binary manipulation mask encoded as PNG bytes. - channels (int64): Total channel count (84). - latenth (int64): Latent height (64). - latentw (int64): Latent width (64).
This dataset was created using LeRobot.
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.