-
Notifications
You must be signed in to change notification settings - Fork 143
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
end2end C++推理报错,[pluginV2DynamicExtRunner.cpp::execute::115] Error Code 2: Internal Error (Assertion status == kSTATUS_SUCCESS failed. ) #76
Comments
我的trt版本是8.4.0.6 |
建议使用netron检查模型输出 |
int8 效果取决于你的校准集,增加校准集数据量 @leayz-888 |
|
我当时验证过v6得代码,是没有问题的,v7你是哪个模型,可能是v7的重参数化会影响量化效果。 |
v7/v7-tiny都试了,都是这个问题 |
你可以在v7的repo里看一下,有没有提到这个问题,或者验证一下v6是不是也存在这个问题。 |
好的,我验证了v7、v7-tiny带nms插件的fp16的engine推理精度是没问题的,感觉是int8量化过程出问题了 |
是的,我的意思是可能是v7的重参数化,本身就对int8量化不友好。 @leayz-888 |
|
我用的yolov5l模型在onnx转engin的时候能成功的。但是推理出来的结果是空白的。 |
v5 目前更新了仓库, 需要保证输出是一个, 建议使用netron 查看onnx模型,修改代码使得导出只有一个输出! |
您好,我按照指令:python export.py --weights ./yolov7-tiny.pt --grid --simplify将pt模型转为onnx
然后使用:python export.py -o ./yolov7-tiny.onnx -e yolov7-tiny.engine -p int8 --calib_input ./calibration_img --end2end --calib_num_images 500 进行量化,成功得到了engine文件,但是我发现engine文件有7个输出,分别是:278/336/394、num/boxes/scores/classes,前面3个是小、中、大检测头的输出,后4个是端到端的输出,我确认了main.cpp中的输入输出节点名称是一样的,但是在使用engine文件推理的时候报错:
[pluginV2DynamicExtRunner.cpp::execute::115] Error Code 2: Internal Error (Assertion status == kSTATUS_SUCCESS failed. )
Segmentation fault (core dumped)
请问一下这是为什么呢?
The text was updated successfully, but these errors were encountered: