Datasets
High-quality open-source AI training datasets
Hugging Face Datasets Collection
The largest open-source dataset hub on the web, hosted by Hugging Face. It spans NLP, computer vision, audio, and multimodal tasks, offering hundreds of thousands of community-contributed datasets with efficient streaming and loading utilities. Whether you need text classification benchmarks, image-caption pairs, or speech-to-text corpora, this is the go-to infrastructure for sourcing quality training data.
Kaggle Datasets Platform
Kaggle hosts tens of thousands of public datasets spanning machine learning competitions, real-world business problems, and academic research. All datasets are available for online analysis through built-in notebooks and can be downloaded directly for local experimentation and model training.
Papers with Code Datasets
Papers with Code links benchmark datasets from academic papers to their corresponding state-of-the-art models and open-source implementations. This makes it easy to compare methods, reproduce results, and find the right dataset and code baseline for your research.
Google Dataset Search
Google Dataset Search is a dedicated search engine that indexes public datasets across thousands of sources, letting researchers and developers quickly locate the specific data they need for training, analysis, or validation regardless of where it is hosted.
Common Crawl Web Corpus
Common Crawl is an open, petabyte-scale web crawl corpus that serves as a foundational pre-training data source for many large language models. The corpus is freely available for download and widely used in NLP research, data mining, and large-scale machine learning experiments.
LAION Open Multimodal Datasets
LAION (Large-Scale Artificial Intelligence Open Network) provides massive open-source image-text pair datasets that have been widely used to train multimodal models such as CLIP and Stable Diffusion. The datasets are freely available and serve as key infrastructure for open-source generative AI research.
UCI Machine Learning Repository
The UCI Machine Learning Repository is a classic collection of hundreds of structured datasets maintained since the late 1990s, widely used for hands-on practice, algorithm benchmarking, and teaching fundamental machine learning concepts in academic courses worldwide.
The Stack Source Code Dataset
The Stack is a large, permissively licensed source code dataset from the BigCode project, covering hundreds of programming languages. It is widely used to pre-train and evaluate code generation models, and includes opt-out mechanisms and deduplication tooling to support responsible data practices.
ImageNet
ImageNet is a landmark large-scale image dataset with millions of labeled images spanning thousands of categories. It powered the deep learning revolution in computer vision through the ILSVRC benchmark competition and remains foundational for pre-training, transfer learning, and evaluating image classification models.