Caffe2 - C++ API
A deep learning, cross platform ML framework
Public Types | Public Member Functions
caffe2::BlobSerializerBase Class Referenceabstract

BlobSerializerBase is an abstract class that serializes a blob to a string. More...

#include <blob_serializer_base.h>

Inheritance diagram for caffe2::BlobSerializerBase:
caffe2::db::DBReaderSerializer caffe2::QTensorSerializer< Context > caffe2::StringSerializer caffe2::TensorSerializer< Context >

Public Types

using SerializationAcceptor = std::function< void(const std::string &blobName, const std::string &data)>
 

Public Member Functions

virtual void Serialize (const Blob &blob, const std::string &name, SerializationAcceptor acceptor)=0
 The virtual function that returns a serialized string for the input blob. More...
 
virtual void SerializeWithChunkSize (const Blob &blob, const std::string &name, SerializationAcceptor acceptor, int chunk_size)
 

Detailed Description

BlobSerializerBase is an abstract class that serializes a blob to a string.

This class exists purely for the purpose of registering type-specific serialization code. If you need to serialize a specific type, you should write your own Serializer class, and then register it using REGISTER_BLOB_SERIALIZER. For a detailed example, see TensorSerializer for details.

Definition at line 23 of file blob_serializer_base.h.

Member Function Documentation

◆ Serialize()

virtual void caffe2::BlobSerializerBase::Serialize ( const Blob blob,
const std::string &  name,
SerializationAcceptor  acceptor 
)
pure virtual

The virtual function that returns a serialized string for the input blob.

Parameters
blobthe input blob to be serialized.
namethe blob name to be used in the serialization implementation. It is up to the implementation whether this name field is going to be used or not.
acceptora lambda which accepts key value pairs to save them to storage. serailizer can use it to save blob in several chunks acceptor should be thread-safe

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