paddledet-cuda-pathfix 0.1.4

Auto-load CUDA runtime libraries for PaddleDetection.

Highlights:

• Works with CUDA 12.x and 13.x (detects cu12/cu13 at runtime).
• Prevents import errors like libnvrtc.so / libcudart.so missing.

Asset:

• paddledet_cuda_pathfix-0.1.4-py3-none-any.whl

ext_op 0.1.2 (CUDA 12.6, cp313)

CUDA 12.6 build of ext_op for PaddleDetection rotated bbox ops.

Asset:

• ext_op-0.1.2+cuda12.6-cp313-cp313-linux_x86_64.whl

Requirements:

• CUDA 12.6 runtime (cu12 packages)
• Python 3.13
• paddlepaddle-gpu 3.2.2 (cu12)

paddledet-cuda-pathfix 0.1.3

Auto-load CUDA runtime libraries at Python startup to avoid missing .so errors (libnvrtc, libcudart, cublas, cudnn, etc.) when importing Paddle.

Install:
pip install https://github.com/yiluzhou/PaddleDetection/releases/download/v0.1.3/paddledet_cuda_pathfix-0.1.3-py3-none-any.whl

Notes:

• Linux only
• Intended for CUDA 13.x Python environments (cu13 packages)

ext_op 0.1.2 (CUDA 13.0, Python 3.13)

Custom PaddleDetection ext_op wheel (rotated bbox ops).
Build: CUDA 13.0, Python 3.13, Linux x86_64.
Install: pip install https://github.com/yiluzhou/PaddleDetection/releases/download/v0.1.2/ext_op-0.1.2+cuda13_0-cp313-cp313-linux_x86_64.whl