A SigLIP (Sigmoid Loss for Language-Image Pre-training) vision transformer model with 400 million parameters. Hosted in the timm library, it supports zero-shot image classification and produces high-quality image embeddings. The model was trained on the WebLI dataset and is compatible with the OpenCLIP ecosystem.
In the Other AI space, ViT SO400M 14 SigLIP takes a focused approach. It focuses on performing zero-shot image classification and vision-language tasks without task-specific training data. It is built as an open-source project for developers. ViT SO400M 14 SigLIP is open source under the Open Source license. It runs on the web and API.
timm builds and maintains ViT SO400M 14 SigLIP, and the product first shipped in 2023. Key capabilities include zero-shot classification, sigLIP architecture, and openCLIP compatible.
Latest indexed changes and source events
timm/ViT-SO400M-14-SigLIP verified by the PulseGate indexer
