tensorbase / datasets
Datasets
Versioned, sharded datasets with the same chunked storage as model weights. Every revision is immutable and addressable by hash.
| Dataset | Modality | Format | Rows / items | Size | Revision |
|---|---|---|---|---|---|
| tb-labs/ru-web-corpus-2026 | Text | parquet · zstd | 412.6M | 1.84 TB | v3 · 8c1e0a2 |
| volga-ai/callcenter-speech-ru | Audio | webdataset · flac | 12 040 h | 917 GB | v1.2 · 44b9f71 |
| northgrid/industrial-defects-hd | Computer Vision | webdataset · jpeg | 2.3M | 684 GB | v2 · d07a3be |
| finlab/ifrs-reports-instruct | Text | jsonl · gzip | 1.1M | 6.2 GB | v5 · 19e6c4d |
| medsight/ct-segmentation-anon | Computer Vision | nifti · arrow index | 38 450 | 2.96 TB | v1 · a5f20c8 |
| tb-labs/ru-docvqa-scans | Multimodal | parquet · png | 286 000 | 143 GB | v2.1 · 6b3d9e0 |
stream a dataset
$ tb pull tb-labs/ru-web-corpus-2026 --include 'data/train-000*.parquet'Need to host a private dataset larger than 10 TB?
Contact sales