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.
CloneGate work: Idris models and native launchers, WAT holographic frontend, and browser demo. The Idris entry point checks the supplied key/challenge against its fixed fingerprint before executing the guarded body. The Hugging Face dataset requires manual access approval; unauthenticated file downloads are denied. The browser visualizes that check using public demo inputs and the supplied toy transformation. Current product entry point: 29-product source suite, production staging and architecture, current inventory and 83-commit mapping, and release readiness evidence. Start with Build Your Own Bank. The main license is the Sovereign Leviathan mutated AGPLv3 covenant, preserving component…
This dataset comprises 446 machine-checked theorems in mathematical finance, formalized using Lean 4 atop Mathlib and Rémy Degenne's BrownianMotion package. Each entry includes a theorem's formal statement, its proof, subject area, and a "faithfulness tier" indicating alignment between the mathematical and formal claims. Sourced from the formal-mathfin library, this collection serves as training and evaluation material for autoformalization and automated theorem proving focused on quantitative finance — a domain underrepresented in existing mathematical benchmarks. The dataset provides these fields for each theorem: - formalizationstatus: faithfulness category delivers less than the source…
jais Synthetic multi-object rigid-body videos generated with Kubric (PyBullet physics + Blender/Cycles rendering, GPU-rendered). Every scene has 4-6 fixed-mass bodies on a table; exactly one body (the subject) receives an initial velocity along a single heading and nothing else is actuated. Some bodies sit in the subject's corridor and get struck, others are bystanders that are never touched. Spheres roll, boxes slide, and a sphere launched without spin slides first and then
German European Parliament plenary speech paired with a transcript, acoustic measurements, discrete German descriptor tags and a free-text German description of the voice and recording conditions. The dataset is intended for training description-conditioned TTS models such as The data was built as part of research work. It is a random subset of the German unlabelled part of VoxPopuli (release v1, from the VoxPopuli GitHub repository), not the transcribed subset on the Hub (facebook/voxpopuli), which it does not overlap (see below). Most of the audio is simultaneous interpretation. VoxPopuli's German unlabelled data is the Parliament's German audio channel: when a member speaks German it…
A chrome-less, unrestricted Android browser with built-in AI, built on GeckoView (the engine behind Firefox for Android) with real WebExtensions support and a developer console. 1. No restrictions by default — OFFDIGITZ deliberately diverges from conventional browsers: - No tracking protection / content blocking (Gecko defaults stand) - Extension install permissions auto-granted (no dialogs) - about:config enabled — full engine configuration surface 2. Chrome-less UI — no URL bar, no toolbar, no tab strip. Edge-to-edge content; one floating ⌘ button summons the command palette that routes everything. 3. AI built in — chat with Hugging Face models or any OpenAI-compatible endpoint; keys live…
این مجموعه‌داده شامل 13 بخش اصلی و مستقل است که از منوی کشویی Subset در بالای صفحه هاگینگ‌فیس به صورت زنده قابل انتخاب و مشاهده است: 1. long: مقالات بلند و بیمه‌شده ویکی‌پدیای فارسی (JSONL). 2. short: مقالات خرد و کوتاه ویکی‌پدیای فارسی (JSONL). 3. enwikipedia1B: یک میلیارد توکن مقالات بلند ویکی‌پدیای انگلیسی. 4. finewebedu6B: مقالات آموزشی انگلیسی FineWeb-Edu. 5. syntheticpersian2B: دو میلیارد توکن استدلال و حل مسئله مصنوعی ابن‌سینا. 6. parallelmizan: یک میلیون جفت‌جمله رمان‌های چاپی میزان بسته‌بندی‌شده به صورت پاراگراف متوالی. 7. paralleltinystories: داستان‌های دوزبانه انگلیسی-فارسی کودکان TinyStories با گرامر روان. 8. pythoncode12B: ده میلیارد توکن کد پایتون ۳ با حداقل ۱ ستاره گیت‌هاب…
paper-with-me 서비스용 SQLite 스냅샷. Papers with Code 2025-07 아카이브(pwc-archive, CC-BY-SA 4.0) + arXiv/HF Daily Papers/GitHub 일일 수집분.
Runtime input data for the Search-SWE benchmark. The dataset provides task runtime assets such as corpora, fixed vectors, training data, and public validation inputs. File formats vary by task; use the benchmark download script to restore the expected layout. File inventories, sizes, and checksums are recorded in the asset manifests. The contents of each task's data/ directory are placed directly under tasks/ / here. For example: The dataset's manifest.json and the task manifests in the code repository record the expected sizes and SHA-256 checksums. Hidden evaluation queries and labels remain in the code packages' tests/data/ directories and are not included here. Fixed model weights are…
ChainSWE is a benchmark of sequential bug-fix tasks for evaluating coding agents on continuous software maintenance. Each row is one time-ordered sequence in one repository, with a shared base commit and Docker image; bugfixes contains its ordered tasks. The length-specific splits contain 376 sequences and 1504 bug-fixes. unique is a non-overlapping sequence union from the three length-specific splits, not additional data. Evaluation is run with the SWE-Edit.
Each row is one panel: the original black-and-white manga panel (bw) next to its colorized versions coloredv0... coloredv5, generated by Qwen-Image 2.1 Turbo from color-grading recipes. v0 has all tags null (generic prompt); v1..v5 use sampled recipes (mood, time of day, light tone, saturation, palette words). All variants of a panel share one noise seed, so differences come from the tags. Columns: bw, coloredv0..v5 (images), panel, series, chapter, and metav0..v5 (one struct per variant: style, mood, timeofday, lighting, saturation, palette, prompt, seed, meanchroma, grayssim, bubblechroma, swatcheslab = 4 Lab swatches, most frequent first, flattened to 12 floats). No quality filtering has…
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.