曾逸夫（Zeng Yifu）
曾逸夫（Zeng Yifu）
@glenn-jocher @AyushExel I tested on torch==1.12.0 and everything works fine. Here is my configuration: YOLOv5 🚀 v6.1-370-g20f1b7e Python-3.9.12 torch-1.12.0+cu116 CUDA:0 (NVIDIA GeForce RTX 3090, 24260MiB) 
> 试下feature_demo_stream那个分支，examples/base/object_detector_yolo.cc，ProcessSDKOutput这个函数里的output换成你自己模型的output #### 我试了，5维的tnn模型和4维的tnn模型都试了，在feature_demo_stream这个分支，然而用手机摄像头检测没有任何反应。 #### 于是我用`物体检测-图片输入`进行测试，结果提示这个错误： ```shell E/tnn: virtual tnn::Status tnn::TNNSDKSample::Init(std::shared_ptr) [File /media/zyf/document/TNN/examples/base/tnn_sdk_sample.cc][Line 570] instance.net init failed 4098 jint Java_com_tencent_tnn_demo_ObjectDetector_init(JNIEnv *, jobject, jstring, jint, jint, jfloat, jfloat, jint, jint) [File...
@ouxiand 这个我很久没研究了，我当时使用的是YOLOv5 v5.0，你可以尝试一下最新的YOLOv5 v6.1。
@DamirPorobic Click this shortcut icon, there is no response. The previous version would have a reduced icon in the upper right corner, now there is nothing. 
@DamirPorobic Sorry I didn't find the file `~/.config/ksnip`, I searched the entire home directory and there is no ksnip related information.
@DamirPorobic I downgraded the installation one by one until the version `1.9.0` was used normally. I hope you can solve the problem of incompatibility between the higher version and Ubuntu....
> yolo.py -> class Detect(nn.Module): 大神，请问这个yolo.py文件是在什么操作下使用的？ 是运行export.py文件导出onnx模型的时候还是运行train.py文件生成pt模型时？
> 只有在导出的时候改一下就可以了，训练时不用改使用默认的就行。 谢谢大神
@glenn-jocher Thank you very much.
@dawoodkhan82 You try to use gradio 3.1.3, the problem is still not solved. You check the version of gradio you are using.