v5.5.1
📦 sentence-transformersView on GitHub →
🐛 1 fixes🔧 1 symbols
Summary
This patch release resolves a quirk in multimodal inference where single-key dictionary inputs were assigned an incorrect tuple modality, ensuring consistency with bare modality inputs.
🐛 Bug Fixes
- Fixed an issue where single-key multimodal inputs like model.encode({"image": ...}) were incorrectly inferred as the modality ("image",) instead of the bare modality "image", which caused errors if the model lacked a config for the tuple modality.