Research snapshot of official national legislation from laws.gov.ag / Official Gazette (gazette.laws.gov.ag) / Ministry of Justice & Legal Affairs. Not legal advice. The official gazette / authentic source prevails over this corpus. - data/laws.parquet — one row per instrument. - data/articles.parquet — article/section rows for the same snapshot. - scrapers/ — collectag.py (collector, splitter), agcatalog.py (official-source catalog), agbatch.py (Hub-aware batch runner), agfixv3.py (v3 approved tip fixes), textlayer.py (garbled text-layer detector), ctrlchars.py / ctrlaudit.py (control characters), bzocrgarbled.py (OCR noise-page filter), and the shared helpers common.py, worldlib.py…
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 dataset contains clean, structured public data exported directly from production runs of Apify actors. It serves as a benchmark and sample for lead qualification, market intelligence, research, and machine learning pipelines. Fresh job listings for AI, ML, and Software Engineering positions across the United States. Includes canonical job posting URLs, hiring company names, job titles, locations, raw posting age, and estimated posting dates. Generated via Apify Actor captainhandsome/linkedin-public-jobs-search. - title: (e.g. AI Engineer) - company: (e.g. P-1 AI) - location: (e.g. San Francisco Bay Area) - url: (e.g. https://www.linkedin.com/jobs/view/ai-engineer-at-p-1-ai-443)…
Research snapshot of official national legislation from AGC Legislation Portal — Acts currently in force (attorneygenerals.gov.sb). Not legal advice. The official gazette / authentic source prevails over this corpus. - data/laws.parquet — one row per instrument. - data/articles.parquet — article/section rows for the same snapshot. - scrapers/collectsb.py — research collector used to build this snapshot. Official Solomon Islands AGC texts; authentic text prevails; not legal advice. This dataset is not legal advice. The official source prevails. - https://attorneygenerals.gov.sb/legislation/legislation-portal/ Download Monitor Acts currently in force via official AGC portal. Polite ≤1 req/s.…
Code, worktrees, original corpora, experiment outputs and frozen evidence for sglangtest. Read RESTORE.md and verification.json before restoring. migration-inventory.json maps the included files; SHA256SUMS uses portable relative paths. Research-secure remote data is deferred at the owner request. Runtime KV stores and credentials are excluded.
Native GEPA, CLUE and ACE experiment logs and available actor trajectory evidence. Files require manual access approval. Request access with your Hugging Face account. - Combined experiment summary: GEPA, CLUE and ACE; method × site for WebArena. - Non-WebArena repeat scores and variance, including completed ACE ≤50k-token evaluations. - Systematic raw evidence snapshot: 590,861 file paths, source/export checksums, original and invalidated attempts, checkpoints and restoration script. This is a timestamped snapshot. WebArena training continues; incomplete groups are explicitly marked. Original ACE context-limit failures and separate reevaluations remain distinct. - Non-WebArena results…
Versioned geographic map packages and tiled viewer assets for FireViewer. The generic producer is the separate UWD Map Builder; historical FireViewer spatial identifiers remain part of the published compatibility and provenance contracts. These are geographic resources, not a conventional image-training dataset. Map geometry, semantic placements and display assets must remain distinct from dated fire observations, incident perimeters, simulations and model predictions. The current repository contains three runtime/viewer-tiled/catalog.json files: Each catalogue records its bounds, local origin, metre units, asset paths and hashes. Viewer coordinates use a local glTF Y-up frame; preserve the…
Public archive of /media/fulei/jlu for account fulei1232. Hugging Face cannot store a single file larger than 50GB, and it cannot store paths inside a directory named.cache. - 30 optimizer checkpoints are stored as 40GB parts named.optimizer.part000,.optimizer.part001,... - Restore one checkpoint with cat some.optimizer.part > some.optimizer after checking optimizerpartsmanifest.tsv. - robomme/robommepreprocesseddata/.cache is copied to robomme/robommepreprocesseddata/cache. - proxy.env is not in this repository.
This dataset contains Direct Preference Optimization (DPO) pairs translated and culturally adapted into Icelandic from the original argilla/ultrafeedback-binarized-preferences-cleaned dataset. The dataset contains the following fields: - prompt: The culturally adapted Icelandic prompt. - chosen: The high-quality, authentic Icelandic response. - rejected: The poorly translated, uncalibrated response. - language: The target language of the row. This dataset inherits the license from its source dataset argilla/ultrafeedback-binarized-preferences-cleaned (MIT).
TableVerse -> Gaussian splats Real Gaussian splats (.spz) converted from TableVerse-100K (ByteDance, CC BY 4.0) via this project's own analytic, training-free mesh-to-splat converter (splataverse.cli mesh-to-splat --split components, CPU/torch backend). Ingestion of the full ~100k-scene dataset is in progress; scenes are added incrementally as they finish. Source attribution Dataset: TableVerse-100K, ByteDance
Research retrieval release of endomorphosis/ipfsswedenmunicipallaws (revision 2f9d9e955fbbab941a0508e85d1e65f6e50cb623) 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 Sweden Municipal 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/ipfsswedenmunicipallawsir. entrycid, lawcid…
This dataset contains 92,313 verified, preprocessed Knee MRI volumes from the Osteoarthritis Initiative (OAI). The original DICOM/medical imaging files have been standardized, cleaned, and converted into highly efficient 3D.npy (NumPy) arrays to accelerate machine learning and computer vision workflows. To handle the massive scale of this data, the dataset has been strictly verified to remove missing files and non-MRI scans (e.g., X-rays), and is packaged into sequential.tar chunks (~82 GB each) for stable downloading and distributed training. This dataset is designed for researchers, data scientists, and medical AI developers building 3D computer vision models. Primary use cases include…
Held-out video eval set for SurgVLM-style surgical video tasks. Each QA row is paired with an MP4 clip (the listed frames encoded in order). - 36 tasks, 200 QA rows each = 7,200 rows - 6,512 MP4 clips (rows can share a clip) videopath is relative to this repo root, e.g. media/GraSP/GraSPCASE041002476-002485n10.mp4. Clip identity is mediauid (SHA-256 of {imagepaths, clipdurationsec, fps}). Multiple QA rows can share one clip. If you hit Xet Runtime Error / Task Panic, retry with HFHUBDISABLEXET=1.
Dataset files and annotations used in REVA-PO: Stabilizing Reinforcement Learning for Chest X-ray Report Generation, accepted to ECCV 2026. REVA-PO is a reinforcement learning framework for chest X-ray report generation. It uses Response-Weighted Regularization and Validation-Anchored Policy Reset to improve training stability and exploration. This repository contains the IU-Xray data package and the MIMIC-CXR annotation files used by the released code. The model checkpoints are available in the REVA-PO weights repository. Install the Hugging Face Hub client: Extract images.zip inside iuxraydataset/: Set the following fields in the REVA-PO configuration files: MIMIC-CXR images are not…
Curated by DOVEXAI LTD (Nigeria) · dovexai.africa · dovexai.io A provenance-tracked dataset of Nigerian-language voice recordings paired with human transcriptions and translations, built for evaluation and supervised fine-tuning (SFT) of speech and translation models. Every recording is cryptographically receipted, pseudonymised, and quality-checked before release. DOVEXAI builds training and evaluation data across every modality - text, audio, image, video, document and multimodal - including collection, transcription, translation, data generation, SFT and RLHF preference data, human evaluation and red-teaming packs. Our differentiated coverage is Nigerian and other African languages…
Evaluation logs of every model row reported in the ICLR 2027 submission MemGUI-RL (project page: https://memgui-rl-anonymous.github.io/). One zip archive per evaluation session: memguibench/.zip - per-task ConAct trajectories (traj.json, three attempts where applicable), MemGUI-Eval judge outputs (results.csv, per-attempt judgments, failure labels, information-retention analyses) and run metadata. Screenshots are omitted for size; the paper's case-study tasks are included under casestudies/ with down-scaled screenshots. mobileworld/.zip - per-task trajectories and the MobileWorld GUI-only evaluation report. manifest.json maps every archive to the row of the paper it supports.…
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.