This section of the Deep Learning SDK Documentation is about Refitting an Engine: https://docs.nvidia.com/deeplearning/sdk/tensorrt-developer-guide/index.html#refitting-engine-c.
My source code:
IRefitter* refitterPre = createInferRefitter(*engine, gLogger);
However when building my code it compiles fine, but the linker gives this error:
“error LNK2001: unresolved external symbol createInferRefitter_INTERNAL”
Some investigation shows me:
- NvInfer.h contains:
inline IRefitter* createInferRefitter(ICudaEngine& engine, ILogger& logger) - However, the function “createInferRefitter_INTERNAL” is missing from both:
nvinfer.dll (checked using DependencyWalker) and
nvinfer.lib (checked using dumpbin /all nvinfer.lib | find /i “createInferRefitter” from Visual Studio command prompt)
for TensorRT version 5.1.2.2 and 5.0.4.3 on Windows.
Why this discrepancy between the documentation+NvInfer.h and the nvinfer.dll/nvinfer.lib?
Is refitting still supported in TensorRT 5.1.2.2 RC (or 5.0.4.3) on Windows?
Thanks in advance!
Arjan