Issue: TensorRT 5 C++ API Documentations missing

[xiaoxiangcao]

Provide details on the platforms you are using:
Not relevant

Describe the problem

The documentations for C++ APIs seems to be for TensorRT 4 still. For example, missing the new pure virtual functions like clone, destroy, etc for IPluginExt. Also, NvInferPlugin.h seems to be not of version 5. The link given in TensorRT 5 documentations for new plugins like NMS_TRT is missing. Although, some part of the documentations seems to be of v5. For instance, the macro NV_TENSORRT_MAJOR is defined as 5 in the documentation. So there might be some discrepancies among different pages.

Files

N/A

[NVES]

Hello,

can you share which version of TRT5 (RC or GA) you are using? Are you referring to the online documentation or another source?

thanks
NVES

[NVES]

for example,

nvinferplugin.h as defined here https://docs.nvidia.com/deeplearning/sdk/tensorrt-api/c_api/_nv_infer_8h_source.html

has

56 #define NV_TENSORRT_MAJOR 5 
   57 #define NV_TENSORRT_MINOR 0 
   58 #define NV_TENSORRT_PATCH 2 
   59 #define NV_TENSORRT_BUILD 1 
   60 
   61 #define NV_TENSORRT_SONAME_MAJOR 5 
   62 #define NV_TENSORRT_SONAME_MINOR 0 
   63 #define NV_TENSORRT_SONAME_PATCH 2 
   64

[xiaoxiangcao]

Thank you for your reply.

An example would be createNMSPlugin:

if you would go to https://docs.nvidia.com/deeplearning/sdk/tensorrt-api/c_api/_nv_infer_plugin_8h_source.html, which is the source code webpage, and search for the function name, you shall see the API at line 416.

However, if you would go to https://docs.nvidia.com/deeplearning/sdk/tensorrt-api/c_api/_nv_infer_plugin_8h.html, which is the generated documentation webpage, and search for the same function name, you can’t find it.

The problem is that the source code page is for v5 and the API webpage is for v4 (I’m guessing).

The same would be for IPluginExt API page. Try search the function clone. Again, you can find in source code page, but not in API page. And again, clone is a new API in TensorRT 5, not 4.

Also, I’m currently working on Jetson TX2 and Xavier at the same time. The jetpacks for them come with different versions. It’ll be nice if I can have API webpage of different versions for me the refer to, since I’m maintaining source code deployed on both TensorRT v4 and v5. Although I’m quite familiar with some of the API differences, I just think it would be nice to have the options.

Thx.

[NVES]

Hello,

I can see createNMSPlugin defined in both https://docs.nvidia.com/deeplearning/sdk/tensorrt-api/c_api/_nv_infer_plugin_8h_source.html and https://docs.nvidia.com/deeplearning/sdk/tensorrt-api/c_api/_nv_infer_plugin_8h.html

Please see attached.

I see clone defined here (line 2830) https://docs.nvidia.com/deeplearning/sdk/tensorrt-api/c_api/_nv_infer_8h_source.html and https://docs.nvidia.com/deeplearning/sdk/tensorrt-api/c_api/classnvinfer1_1_1_i_plugin_v2.html#a759ed653e86b4fc160767fb879554d41

regards,
NVIDIA Enterprise Support

[xiaoxiangcao]

Hi, I still can’t see createNMSPlugin in the webpage you provided (https://docs.nvidia.com/deeplearning/sdk/tensorrt-api/c_api/_nv_infer_plugin_8h.html). It might have sth to do with cache. Thank you for your time.