Command Palette
Search for a command to run...
Hebrew-CLIP Hebrew Dataset
Hebrew-CLIP is a dataset released by NVIDIA in 2024 for training Hebrew visual-language models, aiming to facilitate the application of visual-language models such as CLIP in Hebrew contexts.
The dataset contains approximately 7.78 million Hebrew image captions, primarily composed of machine-translated data from DataComp-1B and multilingual data from LAION-5B. The data is stored in Parquet format, does not include actual images, and provides only text descriptions along with references to corresponding image embeddings, making it suitable for training and evaluating multimodal models.
Dataset Composition
The dataset consists of two Parquet files, each primarily containing the following fields:
- key: The unique identifier for the description.
- heb_caption: The Hebrew caption text.
- file_name: The filename of the corresponding image embedding.
- file_index: The index position within the embedding file.
Build AI with AI
From idea to launch — accelerate your AI development with free AI co-coding, out-of-the-box environment and best price of GPUs.