rknn-toolkit
rknn-toolkit copied to clipboard
rockchip-linux
运行adb devices输出 List of devices attached baadda55a7a0319a device ，但是运行rknn程序时报错说设备未连接 “E Device(id: baadda55a7a0319a) is not connected, please using device connected. E Catch exception when init runtime! E Traceback (most recent call...
I rknn-toolkit2 version: 2.3.2 [error_analysis.txt](https://github.com/user-attachments/files/20722845/error_analysis.txt)
rpdzkj@localhost:~/Desktop$ sudo python3 Yolov8m_detect.py W rknn-toolkit-lite2 version: 2.2.0 I RKNN: [08:21:53.965] RKNN Runtime Information: librknnrt version: 1.1.0 (8a9d6f66@2021-08-19T15:09:11) E RKNN: [08:21:53.967] failed to open rknpu module, need to insmod rknpu...
使用提供的预编译代码，对自己训练的yolov11m.rknn(fp16)进行预编译报错 E RKNNAPI: rknn_query, recv(MsgIoctl IOCTL_CMD_RKNN_PRECOMPILE) fail! E Catch exception when get precompile model E Traceback (most recent call last): E File "rknn/api/rknn_base.py", line 1317, in rknn.api.rknn_base.RKNNBase.get_precompile_model_on_hardware E File "rknn/api/rknn_runtime.py",...
在转换relu-yolov5s时，配置了force_builtin_perm=add_perm，然后使用零拷贝方式，结果推理一帧数据需要100多秒，结果是正确的。版本是1.7.3
**背景：** paddleOCR转RKNN后，使用python fastdeploy推理，官网要求1.4.2b3版本工具转的rk，否则报错如下。但网上查找不到该版本，大佬们是否有，或者有其他解决方案？ **报错信息：** FastDeploy_release_1_0_7/examples/vision/ocr/PP-OCR/rockchip/python/infer.py E RKNN: [09:48:47.957] 6, 1 E RKNN: [09:48:47.957] Invalid RKNN model version 6 E RKNN: [09:48:47.957] rknn_init, load model failed! [ERROR] fastdeploy/runtime/backends/rknpu2/rknpu2_backend.cc(180)::LoadModel The function(rknn_init) failed!...
在pytorch模型中写深度卷积时，按照分组卷积的方式实现，但是转出来的rknn模型中并没有Depthwise Convolution的op id，请问如何才能正确转出深度可分离卷积
