PP-OCRv6 AI Pod runtime archives and engine caches
Prebuilt Docker-load archives for AGX Orin and AGX Thor, runtime version 0.1.12. Orin: Jetson Linux R36.4.3 / TensorRT 10.7 / SM87. Thor: Jetson Linux R39.2.0 / TensorRT 10.16.2 / CUDA 13.2 / SM110. The runtime images contain inference dependencies without ONNX model weights. The six separate FP16 TensorRT engine caches contain weights derived from the Apache-2.0 PaddlePaddle PP-OCRv6 Tiny, Small and Medium ONNX models. Runtime dependencies retain their respective third-party licenses, including NVIDIA runtime license terms. This repository does not relicense those files.
Each cache is specific to its hardware and runtime image identity. Verify size and SHA256 before importing; never mix hardware targets or runtime versions. The installer downloads official ONNX models separately from PaddlePaddle. See checksums.json for the eight released archive sizes and SHA256 digests.
- Downloads last month
- -