wikitext
DatasetFreeDataset by Salesforce. 12,11,500 downloads.
Capabilities5 decomposed
large-scale language modeling pretraining dataset with wikipedia source material
Medium confidenceProvides a curated corpus of 100M+ tokens extracted from Wikipedia articles, preprocessed into train/validation/test splits optimized for causal language modeling and masked language modeling tasks. The dataset is distributed via HuggingFace Datasets library with native support for streaming, lazy loading, and multi-format export (Parquet, Arrow, CSV), enabling efficient batch processing at scale without requiring full dataset materialization in memory.
Combines Wikipedia's high-quality, encyclopedic text with HuggingFace's streaming infrastructure, enabling researchers to load and iterate on 100M+ tokens without local storage constraints; native support for Parquet, Arrow, and Dask enables distributed preprocessing across clusters without custom ETL pipelines
Larger and more curated than raw Wikipedia dumps (removes boilerplate, metadata, markup) while maintaining reproducibility through versioned HuggingFace hosting, unlike ad-hoc Wikipedia snapshots that require custom preprocessing and deduplication
train-validation-test split management with stratified sampling
Medium confidenceAutomatically partitions the Wikipedia corpus into three disjoint subsets (train: ~90%, validation: ~5%, test: ~5%) with stratified sampling to ensure consistent article-level distribution across splits. The splits are deterministically generated using seeded random sampling, enabling reproducible train/eval workflows and preventing data leakage between model development and evaluation phases.
Provides deterministic, article-level stratified splits baked into the HuggingFace dataset versioning system, eliminating the need for custom train-test-split scripts and ensuring all researchers using WikiText use identical splits for fair benchmarking
More reproducible than raw Wikipedia dumps requiring manual splitting, and more transparent than proprietary datasets with undisclosed split methodologies; enables direct comparison with published results using WikiText
streaming-compatible lazy loading with memory-efficient batch iteration
Medium confidenceImplements HuggingFace Datasets' streaming protocol, enabling on-the-fly data loading without downloading the full corpus. Users iterate over batches via a generator interface that fetches and caches chunks from remote storage (Hugging Face Hub CDN), supporting distributed training on clusters with limited local storage. Integrates with PyArrow and Polars for columnar processing, enabling efficient filtering, grouping, and transformation without materializing the entire dataset in memory.
Leverages HuggingFace's distributed CDN infrastructure and streaming protocol to enable training without local materialization; integrates with PyArrow columnar format for zero-copy filtering and transformation, avoiding redundant data copies during preprocessing
More efficient than downloading full Wikipedia dumps and storing locally; more flexible than fixed-size sharded datasets because streaming adapts to available bandwidth and enables dynamic filtering without re-downloading
multi-format export with native parquet and arrow serialization
Medium confidenceExports dataset content to multiple columnar and row-based formats (Parquet, Arrow, CSV) via HuggingFace Datasets' native serialization layer. Parquet export enables efficient compression and columnar storage for analytics workflows, while Arrow enables zero-copy in-memory processing for PyArrow and Polars. Metadata (split information, article IDs, token counts) is preserved across formats, enabling downstream tools to reconstruct dataset provenance.
Provides native, zero-copy export to Arrow and Parquet via HuggingFace's integrated serialization, avoiding custom ETL scripts; preserves dataset metadata and versioning across formats, enabling reproducible downstream workflows
More efficient than manual CSV generation or custom Parquet writers; native HuggingFace integration ensures schema consistency and metadata preservation, unlike ad-hoc export scripts that often lose provenance information
dataset versioning and reproducibility tracking via huggingface hub
Medium confidenceMaintains immutable dataset versions on HuggingFace Hub with Git-based version control, enabling users to pin specific dataset versions in code and reproduce results across time. Each version includes metadata (creation date, preprocessing steps, source Wikipedia dump date) and is accessible via semantic versioning (e.g., 'wikitext-3.1.0'). Dataset cards document preprocessing decisions, licensing, and known limitations, enabling transparent auditing of data provenance.
Integrates Git-based version control with HuggingFace Hub's immutable dataset storage, enabling semantic versioning and reproducible pinning without custom version management infrastructure; dataset cards provide transparent documentation of preprocessing and licensing
More reproducible than raw Wikipedia snapshots or ad-hoc dataset distributions; more transparent than proprietary datasets with opaque versioning; enables direct reproducibility of published results via version pinning
Capabilities are decomposed by AI analysis. Each maps to specific user intents and improves with match feedback.
Related Artifactssharing capabilities
Artifacts that share capabilities with wikitext, ranked by overlap. Discovered automatically through the match graph.
FineFineWeb
Dataset by m-a-p. 5,55,725 downloads.
StarCoderData
250GB curated code dataset for StarCoder training.
Build a Large Language Model (From Scratch)
A guide to building your own working LLM, by Sebastian Raschka.
Image as a Foreign Language: BEiT Pretraining for All Vision and Vision-Language Tasks (BEiT)
* ⭐ 09/2022: [PaLI: A Jointly-Scaled Multilingual Language-Image Model (PaLI)](https://arxiv.org/abs/2209.06794)
happy-llm
📚 从零开始构建大模型
Dolma
Allen AI's 3T token dataset for fully reproducible LLM training.
Best For
- ✓NLP researchers validating language model architectures against standardized benchmarks
- ✓ML engineers building production language models requiring reproducible, versioned training data
- ✓Teams fine-tuning pretrained models on domain-specific tasks with Wikipedia as initialization corpus
- ✓Academic institutions with limited compute budgets needing efficient, streaming-compatible datasets
- ✓ML researchers publishing results and requiring reproducible, auditable data splits
- ✓Teams implementing rigorous model evaluation protocols with separate hyperparameter tuning and final test phases
- ✓Practitioners benchmarking against published results that use the same WikiText splits
- ✓Teams with limited local storage or GPU memory training on large-scale datasets
Known Limitations
- ⚠English-only monolingual dataset — no multilingual coverage or cross-lingual transfer capabilities
- ⚠Wikipedia bias toward encyclopedic, formal writing style — poor representation of conversational, technical, or domain-specific language patterns
- ⚠Fixed snapshot from specific Wikipedia dump date — does not reflect real-time Wikipedia updates or evolving language use
- ⚠No built-in deduplication across Wikipedia versions or article revisions — may contain near-duplicate content
- ⚠Requires external preprocessing for tokenization, vocabulary building, and sequence packing — dataset provides raw text only
- ⚠Fixed splits cannot be customized per user — no support for k-fold cross-validation or stratified sampling by article category
Requirements
Input / Output
UnfragileRank
UnfragileRank is computed from adoption signals, documentation quality, ecosystem connectivity, match graph feedback, and freshness. No artifact can pay for a higher rank.
About
wikitext — a dataset on HuggingFace with 12,11,500 downloads
Categories
Alternatives to wikitext
Are you the builder of wikitext?
Claim this artifact to get a verified badge, access match analytics, see which intents users search for, and manage your listing.
Get the weekly brief
New tools, rising stars, and what's actually worth your time. No spam.
Data Sources
Looking for something else?
Search →