Option object to configure Paddle Inference backend. More...

#include <option.h>

Collaboration diagram for fastdeploy::PaddleBackendOption:

Public Member Functions
void	DisableTrtOps (const std::vector< std::string > &ops)
	Disable type of operators run on TensorRT.

void	DeletePass (const std::string &pass_name)
	Delete pass by name.

Public Attributes
bool	enable_log_info = false
	Print log information while initialize Paddle Inference backend.

bool	enable_mkldnn = true
	Enable MKLDNN while inference on CPU.

bool	enable_trt = false
	Use Paddle Inference + TensorRT to inference model on GPU.

bool	enable_memory_optimize = true
	Whether enable memory optimize, default true.

bool	switch_ir_debug = false
	Whether enable ir debug, default false.

bool	collect_trt_shape = false
	Collect shape for model while enable_trt is true.

int	mkldnn_cache_size = -1
	Cache input shape for mkldnn while the input data will change dynamiclly.

int	gpu_mem_init_size = 100
	initialize memory size(MB) for GPU

bool	enable_fixed_size_opt = false
	The option to enable fixed size optimization for transformer model.

Detailed Description

Option object to configure Paddle Inference backend.

The documentation for this struct was generated from the following file:

Public Member Functions