#89 · Primary category： Foundation Models
DeepSeek-VL
DeepSeek-VL: Towards Real-World Vision-Language Understanding
Project last updated：04/24/24
GitHub Stars
4.2K
Forks
597
Contributors
7
License
MIT
Why we included this project
DeepSeek-VL is an open vision-language model you can download, run, and fine-tune yourself instead of only calling through an API. It combines a hybrid vision encoder with the DeepSeek language backbone, and the design leans toward real workloads: it accepts 1024x1024 images within a fixed token budget, and the training mix covers web screenshots, PDFs, OCR output, charts, and textbooks. That makes it a reasonable fit for document understanding, reading formulas and diagrams, and answering questions about images. Both 1.3B and 7B checkpoints are available, so a small team can start with the lighter one and move up later. The repo ships model downloads, a quick-start guide, and a hosted demo, which makes it easy to try before investing in a bigger setup.
Articles for this project
No articles for this project yet.
To suggest a topic or contribute an article, contact us.
Related projects in this category
transformers
🤗 Transformers: the model-definition framework for state-of-the-art machine learning models in text, vision, audio, and multimodal models, for both inference and training.
CLIP
CLIP (Contrastive Language-Image Pretraining), Predict the most relevant text snippet given an image
MiniCPM-V
A Pocket-Sized MLLM for Ultra-Efficient Image and Video Understanding on Your Phone
generative-models
Generative Models by Stability AI
unilm
Large-scale Self-supervised Pre-training Across Tasks, Languages, and Modalities