Collaboration diagram for BatchedXvectorComputerOptions:

[legend]

Public Member Functions
void	Register (OptionsItf *po)

Public Attributes
int32	chunk_size { 150 }

int32	batch_size { 32 }

bool	pad_input { true }

NnetComputeOptions	compute_config

NnetOptimizeOptions	optimize_config

CachingOptimizingCompilerOptions	compiler_config

Detailed Description

Definition at line 34 of file nnet3-xvector-compute-batched.cc.

Member Function Documentation

◆ Register()

void Register ( OptionsItf * po )

inline

Definition at line 43 of file nnet3-xvector-compute-batched.cc.

References BatchedXvectorComputerOptions::batch_size, BatchedXvectorComputerOptions::chunk_size, BatchedXvectorComputerOptions::pad_input, OptionsItf::Register(), NnetComputeOptions::Register(), NnetOptimizeOptions::Register(), and CachingOptimizingCompilerOptions::Register().

Referenced by main().

                                 {
     po->Register("chunk-size", &chunk_size,
                  "Size of chunk, in input frames.  Includes the nnet "
                  "context, so the number of chunks will be more than "
                  "total-input-frames / chunk-size.");
     po->Register("batch-size", &batch_size,
                  "Size of the batches of chunks that we compute at once. ");
     po->Register("pad-input", &pad_input,
                  "If true, for utterances shorter than `chunk-size` frames "
                  "we will pad with repeats of the last frame.");
     compute_config.Register(po);
     optimize_config.Register(po);
     compiler_config.Register(po);
   }