MCPcopy Create free account
hub / github.com/cyrusbehr/tensorrt-cpp-api / Options

Class Options

src/engine.h:64–80  ·  view source on GitHub ↗

Options for the network

Source from the content-addressed store, hash-verified

62
63// Options for the network
64struct Options {
65 // Precision to use for GPU inference.
66 Precision precision = Precision::FP16;
67 // If INT8 precision is selected, must provide path to calibration dataset
68 // directory.
69 std::string calibrationDataDirectoryPath;
70 // The batch size to be used when computing calibration data for INT8
71 // inference. Should be set to as large a batch number as your GPU will
72 // support.
73 int32_t calibrationBatchSize = 128;
74 // The batch size which should be optimized for.
75 int32_t optBatchSize = 1;
76 // Maximum allowable batch size
77 int32_t maxBatchSize = 16;
78 // GPU device index
79 int deviceIndex = 0;
80};
81
82// Class used for int8 calibration
83class Int8EntropyCalibrator2 : public nvinfer1::IInt8EntropyCalibrator2 {

Callers

nothing calls this directly

Calls

no outgoing calls

Tested by

no test coverage detected