Update to use Dataflow Compiler v3.33.1
New Models:
- YOLO26 - End-to-end NMS-free object detection models - yolo26n, yolo26s, yolo26m
- DepthAnything - Zero-shot depth estimation models - depthanything_vits, depthanything_v2_vits.
- Face landmarks detection - release face_landmarks_lite.
- TinyCLIP - TinyCLIP family - Contrastive Language-Image Pre-training models (image & text encoders) tinyclip_vit_40m_32_text_19m_laion400m
- SigLip - SigLip family - Contrastive Language-Image Pre-training model (image & text encoders)- siglip2_b_32_256, siglip_b_16
- PaddleOCR-v5 - Text detection and recognition models - released paddle_ocr_v5_mobile_detection, paddle_ocr_v5_mobile_recognition
- DaViT - tiny - Dual Attention Vision Transformer classification model - davit_tiny.
- Whisper - Whisper family - Automatic Speech Recognition models - whisper_tiny, whisper_base