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.
Episodes 0 to 9 of DreamMachines/actuatorunboxingfullboxcropDall (revision 6a8bded63372), with the street seen behind the robot cell replaced by a measuring lab. The robot, table, box, tray and actuators are the original pixels. Actions, joint states, timestamps, task and the intervention columns are copied unchanged. Only the three camera videos differ. - 10 episodes, 5,308 frames at 50 fps Same ten episodes with other backgrounds: - kiteml/dm-actuator-unboxing-bg-aircraft-workshop: an aircraft workshop - kiteml/dm-actuator-unboxing-bg-hangar: an aircraft hangar 1. SAM 3 finds the background (street, buildings, sky, cars, passers-by) and the foreground (robot, gripper, box, tray, actuator…
Preparation notes and schema examples for Document OCR tasks using Multimodal3 data. Full source material is intentionally not bundled, so provenance and licensing remain explicit. - dataset.py — loading, cleaning, and split preparation code. - datasetinfos.json — schema and split metadata. - metadatasample.jsonl — small, human-readable records for checking the schema. - README.md — data card and usage notes. Before using the prepared data, verify source licenses, duplicates across splits, missing values, label balance, and modality-specific corruption. Record the source version and every filtering rule so a later run can reproduce the same rows. The repository is suitable for testing the…
This dataset was created using LeRobot.
Tooling that builds, qualifies, packages and publishes the portable B200 runtime for the DPCA reproduction of On-Policy Distillation (ivanniu/On-Policy-Distill). This repo is the harness, not the runtime. The runtime payload (venv, source tree, node helpers) and the model weights live in a separate dataset repo, because they are ~56 GB and the tooling is 61 KB. The overlay is 4 commits on top of the upstream commit. That count is a build-time assertion (OVERLAYCOMMITSAHEAD = 4), not a claim in this README. The tooling proves the runtime executes the paper's objective on the paper's Table 4 model pair, on B200, from a portable payload that another machine can verify byte-for-byte. It does…
This dataset contains Direct Preference Optimization (DPO) pairs translated and culturally adapted into Finnish from an English source dataset (e.g. ultrafeedback or safe-rlhf). The dataset contains the following fields: - prompt: The culturally adapted Finnish prompt. - chosen: The high-quality, authentic Finnish response. - rejected: The poorly translated, uncalibrated response. - targetlanguage: The target language of the row. - datasetsource: The source dataset name. - issafetyfocused: Boolean indicating if this is a safety alignment sample. This dataset inherits the license from its source dataset (MIT or Apache 2.0).
This dataset contains Direct Preference Optimization (DPO) pairs translated and culturally adapted into French from an English source dataset (e.g. ultrafeedback or safe-rlhf). The dataset contains the following fields: - prompt: The culturally adapted French prompt. - chosen: The high-quality, authentic French response. - rejected: The poorly translated, uncalibrated response. - targetlanguage: The target language of the row. - datasetsource: The source dataset name. - issafetyfocused: Boolean indicating if this is a safety alignment sample. This dataset inherits the license from its source dataset (MIT or Apache 2.0).
This dataset contains Direct Preference Optimization (DPO) pairs translated and culturally adapted into German from an English source dataset (e.g. ultrafeedback or safe-rlhf). The dataset contains the following fields: - prompt: The culturally adapted German prompt. - chosen: The high-quality, authentic German response. - rejected: The poorly translated, uncalibrated response. - targetlanguage: The target language of the row. - datasetsource: The source dataset name. - issafetyfocused: Boolean indicating if this is a safety alignment sample. This dataset inherits the license from its source dataset (MIT or Apache 2.0).
This dataset contains Direct Preference Optimization (DPO) pairs translated and culturally adapted into Norwegian from an English source dataset (e.g. ultrafeedback or safe-rlhf). The dataset contains the following fields: - prompt: The culturally adapted Norwegian prompt. - chosen: The high-quality, authentic Norwegian response. - rejected: The poorly translated, uncalibrated response. - targetlanguage: The target language of the row. - datasetsource: The source dataset name. - issafetyfocused: Boolean indicating if this is a safety alignment sample. This dataset inherits the license from its source dataset (MIT or Apache 2.0).
This dataset contains Direct Preference Optimization (DPO) pairs translated and culturally adapted into Polish from an English source dataset (e.g. ultrafeedback or safe-rlhf). The dataset contains the following fields: - prompt: The culturally adapted Polish prompt. - chosen: The high-quality, authentic Polish response. - rejected: The poorly translated, uncalibrated response. - targetlanguage: The target language of the row. - datasetsource: The source dataset name. - issafetyfocused: Boolean indicating if this is a safety alignment sample. This dataset inherits the license from its source dataset (MIT or Apache 2.0).
This dataset contains Direct Preference Optimization (DPO) pairs translated and culturally adapted into Spanish from an English source dataset (e.g. ultrafeedback or safe-rlhf). The dataset contains the following fields: - prompt: The culturally adapted Spanish prompt. - chosen: The high-quality, authentic Spanish response. - rejected: The poorly translated, uncalibrated response. - targetlanguage: The target language of the row. - datasetsource: The source dataset name. - issafetyfocused: Boolean indicating if this is a safety alignment sample. This dataset inherits the license from its source dataset (MIT or Apache 2.0).
This repository is a mirror for reproducible builds: it holds the files produced once from the published originals listed below by the processing described here, so that later builds get exactly the same bytes (SHA256SUMS gives the SHA-256 of each file). Please cite and credit the original authors. For 40 of the 62 sessions, in sessions/: -.npy: float16 microvolts [30 channels, samples] at 250 Hz: the 30 scalp channels (A1/A2 references and the vehicle-position channel dropped), consecutive sample pairs averaged from 500 Hz, clipped to +-30000; - trials.csv: each lane drift's deviation onset, reaction time (response onset minus deviation onset) and response offset, in seconds from the start…
Synthetic dataset used for an authorized Hugging Face dataset-viewer cache-key test.
Synthetic dataset used for an authorized Hugging Face dataset-viewer cache-key test.
DVL paper structure archives, prepared 2026-10-05. Total: 59890049555 compressed bytes in ten lossless.tar.zst archives. Upload is in progress. Two archives are already verified on HF; the remaining archives are being transferred from the storage server through the local machine. Five A1002 Bvjp pullbacks are absent: Llama-3.2-1B-Instruct, Qwen2.5-0.5B-Instruct, Qwen2.5-1.5B-Instruct, Qwen2.5-3B-Instruct, Qwen3-1.7B. Available paper-selected48 A31 structures are complete. See TOTALSUMMARY.json for sizes and SHA256SUMS for archive checksums. Full provenance and restoration notes are in DVLstorageaudit.tar.zst.
A documented Ecommerce data-preparation workflow for Pointcloud Text records. The bundled rows demonstrate the schema and validation path rather than pretending to be a full training corpus. - builddataset.py — loading, cleaning, and split preparation code. - datasetinfos.json — schema and split metadata. - metadatasample.jsonl — small, human-readable records for checking the schema. - README.md — data card and usage notes. Before using the prepared data, verify source licenses, duplicates across splits, missing values, label balance, and modality-specific corruption. Record the source version and every filtering rule so a later run can reproduce the same rows. The repository is suitable…
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.