| --- |
| language: |
| - ar |
| - de |
| - en |
| - es |
| - hi |
| - ja |
| - ru |
| - tr |
| - vi |
| - zh |
| task_categories: |
| - text-generation |
| configs: |
| - config_name: qwen3_0.6b_base_ar_10B |
| data_files: |
| - split: train |
| path: subsets/qwen3_0.6b_base_ar_10B/train/shard_*/data-*.arrow |
| - split: validation |
| path: subsets/qwen3_0.6b_base_ar_10B/validation/shard_*/data-*.arrow |
| - config_name: qwen3_0.6b_base_de_10B |
| data_files: |
| - split: train |
| path: subsets/qwen3_0.6b_base_de_10B/train/shard_*/data-*.arrow |
| - split: validation |
| path: subsets/qwen3_0.6b_base_de_10B/validation/shard_*/data-*.arrow |
| - config_name: qwen3_0.6b_base_en_30B |
| data_files: |
| - split: train |
| path: subsets/qwen3_0.6b_base_en_30B/train/shard_*/data-*.arrow |
| - split: validation |
| path: subsets/qwen3_0.6b_base_en_30B/validation/shard_*/data-*.arrow |
| - config_name: qwen3_0.6b_base_es_10B |
| data_files: |
| - split: train |
| path: subsets/qwen3_0.6b_base_es_10B/train/shard_*/data-*.arrow |
| - split: validation |
| path: subsets/qwen3_0.6b_base_es_10B/validation/shard_*/data-*.arrow |
| - config_name: qwen3_0.6b_base_hi_10B |
| data_files: |
| - split: train |
| path: subsets/qwen3_0.6b_base_hi_10B/train/shard_*/data-*.arrow |
| - split: validation |
| path: subsets/qwen3_0.6b_base_hi_10B/validation/shard_*/data-*.arrow |
| - config_name: qwen3_0.6b_base_ja_10B |
| data_files: |
| - split: train |
| path: subsets/qwen3_0.6b_base_ja_10B/train/shard_*/data-*.arrow |
| - split: validation |
| path: subsets/qwen3_0.6b_base_ja_10B/validation/shard_*/data-*.arrow |
| - config_name: qwen3_0.6b_base_ru_10B |
| data_files: |
| - split: train |
| path: subsets/qwen3_0.6b_base_ru_10B/train/shard_*/data-*.arrow |
| - split: validation |
| path: subsets/qwen3_0.6b_base_ru_10B/validation/shard_*/data-*.arrow |
| - config_name: qwen3_0.6b_base_tr_10B |
| data_files: |
| - split: train |
| path: subsets/qwen3_0.6b_base_tr_10B/train/shard_*/data-*.arrow |
| - split: validation |
| path: subsets/qwen3_0.6b_base_tr_10B/validation/shard_*/data-*.arrow |
| - config_name: qwen3_0.6b_base_vi_10B |
| data_files: |
| - split: train |
| path: subsets/qwen3_0.6b_base_vi_10B/train/shard_*/data-*.arrow |
| - split: validation |
| path: subsets/qwen3_0.6b_base_vi_10B/validation/shard_*/data-*.arrow |
| - config_name: qwen3_0.6b_base_zh_10B |
| data_files: |
| - split: train |
| path: subsets/qwen3_0.6b_base_zh_10B/train/shard_*/data-*.arrow |
| - split: validation |
| path: subsets/qwen3_0.6b_base_zh_10B/validation/shard_*/data-*.arrow |
| --- |
| # CulturaX sampled text pools |
|
|
| Use `UPLOAD_COMPLETE.json` before consuming this release. Its absence means the |
| upload is incomplete. Pin the completed repository revision for experiments. |
|
|
| Original, unpacked document text from [uonlp/CulturaX](https://huggingface.co/datasets/uonlp/CulturaX). |
| This is a fresh sample, independent of nguyenhuuthuat09/CulturaX_sampled. |
| No token sequences or training caches are distributed. Each subset has a fixed |
| validation set and a nested training prefix suitable for smaller token budgets. |
| |
| Token counts use **Qwen/Qwen3-0.6B-Base**, revision |
| `da87bfb608c14b7cf20ba1ce41287e8de496c0cd`. Sampling seed: **42**. |
| Source revision: `6a8734bc69fefcbb7735f4f9250f43e4cd7a442e`. |
| Each count includes one explicit `<|endoftext|>` (151643) after the document; |
| that marker is not appended to the stored text. Training must append it. |
| Tokenization has no truncation or added special tokens. Budgets round up to |
| whole documents and do not include packing-loss headroom. Other tokenizers can |
| use the text, but these counts and token-budget anchors will not apply to them. |
| |
| | Subset | Training tokens including EOD | Validation tokens including EOD | |
| |---|---:|---:| |
| | qwen3_0.6b_base_ar_10B | 10,000,002,178 | 10,000,061 | |
| | qwen3_0.6b_base_de_10B | 10,000,001,061 | 10,000,620 | |
| | qwen3_0.6b_base_en_30B | 30,000,000,081 | 10,000,244 | |
| | qwen3_0.6b_base_es_10B | 10,000,001,093 | 10,001,384 | |
| | qwen3_0.6b_base_hi_10B | 10,000,001,763 | 10,008,895 | |
| | qwen3_0.6b_base_ja_10B | 10,000,002,841 | 10,000,048 | |
| | qwen3_0.6b_base_ru_10B | 10,000,001,931 | 10,009,101 | |
| | qwen3_0.6b_base_tr_10B | 10,000,004,885 | 10,001,348 | |
| | qwen3_0.6b_base_vi_10B | 10,000,000,013 | 10,000,396 | |
| | qwen3_0.6b_base_zh_10B | 10,000,000,321 | 10,000,728 | |
| |
| ```python |
| from datasets import load_dataset |
| data = load_dataset("nht10/cx_sampled", "qwen3_0.6b_base_ar_10B", revision="<completed-commit>") |
| ``` |
| |
| Each `subsets/<name>/train/shard_*` and `validation/shard_*` is also a complete |
| `save_to_disk` folder: use `load_from_disk` and concatenate shards in sorted |
| order. This preserves compatibility with the existing local training loader. |
| Download all Arrow files referenced by each shard's `state.json`, plus |
| `state.json` and `dataset_info.json`. `document_index.parquet` is a row-aligned |
| sidecar, explicitly excluded from HF text split definitions. |
| |
| `subsets/<name>/manifest.json` records ordered shard paths, counts, cumulative |
| endpoints, checksums, source-file identities, and every 100M-token anchor. |
| For an arbitrary smaller budget, select the prefix of training shards and use |
| the boundary shard's cumulative token index to round up to a whole document. |
| Keep its fixed validation split. The root manifest lists every published file. |
| Individual files can be downloaded using pinned HF resolve URLs; a whole-repo |
| download retrieves all ten subsets. Downloading files does not load a Dataset. |
| |
| Sampling shuffles source files, row groups, and rows within groups with seeded |
| RNGs; this is not a globally uniform document shuffle. Validation is selected |
| first; its exact text hashes are excluded from training within each language. |
| There is no additional global or near-duplicate deduplication. See provenance |
| for the frozen sampling implementation, complete file choices and tokenizer. |
| |
| ## Source and license |
| |
| This sample inherits the source terms. CulturaX states that its licensing |
| follows mC4 and OSCAR; see the preserved upstream README under provenance and |
| the [CulturaX dataset card](https://huggingface.co/datasets/uonlp/CulturaX). |
| No new license or ownership of the underlying web documents is claimed here. |
| |
