mindspore::session

LiteSession

LiteSession defines sessions in MindSpore Lite for compiling Model and forwarding inference.

Constructors & Destructors

LiteSession

LiteSession()

Constructor of MindSpore Lite LiteSession using default value for parameters.

~LiteSession

~LiteSession()

Destructor of MindSpore Lite LiteSession.

Public Member Functions

BindThread

virtual void BindThread(bool if_bind)

Attempts to bind threads in the thread pool to the specified CPU core or unbind threads from the core.

Parameters
- if_bind: Define whether to bind or unbind threads.

CompileGraph

virtual int CompileGraph(lite::Model *model)

Compile MindSpore Lite model.

CompileGraph should be called before RunGraph.

Parameters
- model: Define the model to be compiled.
Returns

STATUS as an error code of compiling graph, STATUS is defined in errorcode.h.

GetInputs

virtual std::vector <tensor::MSTensor *> GetInputs() const

Get input MindSpore Lite MSTensors of model.

Returns

The vector of MindSpore Lite MSTensor.

GetInputsByTensorName

mindspore::tensor::MSTensor *GetInputsByTensorName(const std::string &name) const

Get input MindSpore Lite MSTensors of model by tensor name.

Parameters
- name: Define tensor name.
Returns

MindSpore Lite MSTensor.

RunGraph

virtual int RunGraph(const KernelCallBack &before = nullptr, const KernelCallBack &after = nullptr)

Run session with callback.

RunGraph should be called after CompileGraph.

Parameters
- before: A KernelCallBack function. Define a callback function to be called before running each node.
- after: A KernelCallBack function. Define a callback function to be called after running each node.
Returns

STATUS as an error code of running graph, STATUS is defined in errorcode.h.

GetOutputsByNodeName

virtual std::vector <tensor::MSTensor *> GetOutputsByNodeName(const std::string &node_name) const

Get output MindSpore Lite MSTensors of model by node name.

Parameters
- node_name: Define node name.
Returns

The vector of MindSpore Lite MSTensor.

GetOutputs

virtual std::unordered_map <std::string, mindspore::tensor::MSTensor *> GetOutputs() const

Get the MSTensors output of the MindSpore Lite model mapped by tensor name.

Returns

The map of output tensor name and MindSpore Lite MSTensor.

GetOutputTensorNames

virtual std::vector <std::string> GetOutputTensorNames() const

Get name of output tensors of model compiled by this session.

Returns

A string variable, contains the output tensors’ names in order.

GetOutputByTensorName

virtual mindspore::tensor::MSTensor *GetOutputByTensorName(const std::string &tensor_name) const

Get output MindSpore Lite MSTensors of model by tensor name.

Parameters
- tensor_name: Define tensor name.
Returns

Pointer of MindSpore Lite MSTensor.

Resize

virtual int Resize(const std::vector <tensor::MSTensor *> &inputs, const std::vector<std::vector<int>> &dims)

Resize inputs shape.

Parameters
- inputs: Model inputs.
- dims: defines the new inputs shape. Its order should be consistent with inputs.
Returns

STATUS as an error code of resize inputs, STATUS is defined in errorcode.h.

Static Public Member Functions

CreateSession

static LiteSession *CreateSession(const lite::Context *context)

Static method to create a LiteSession pointer.

Parameters
- context: Define the context of session to be created.
Returns

Pointer that points to MindSpore Lite MSTensor.

static LiteSession *CreateSession(const char *model_buf, size_t size, const lite::Context *context);

Static method to create a LiteSession pointer. The returned LiteSession pointer has already read model_buf and completed graph compilation.

Parameters
- model_buf: Define the buffer read from a model file.
- size: Define the byte number of model buffer.
- context: Define the context of session to be created.
Returns

Pointer that points to MindSpore Lite LiteSession.

TrainSession

#include <train_session.h>

Inherited from LiteSession, TrainSession defines the class that allows training the MindSpore model.

Constructors & Destructors

~TrainSession

virtual ~TrainSession() = default;

Static method to create a TrainSession object.

Public Member Functions

CreateSession

static TrainSession *CreateSession(const char *model_buf, size_t size, lite::Context *context, bool train_mode = false);

Static method to create a TrainSession object.

Parameters
- model_buf: A buffer that was read from a MS model file.
- size: Length of the buffer.
- context: Defines the context of the session to be created.
- train_mode: Training mode to initialize Session with.
Returns

Pointer that points to MindSpore Lite TrainSession.

static TrainSession *CreateSession(const std::string &filename, lite::Context *context, bool train_mode = false);

Static method to create a TrainSession object.

Parameters
- filename: Filename to read flatbuffer from.
- context: Defines the context of the session to be created.
- train_mode: Training mode to initialize Session with.
Returns

Pointer that points to MindSpore Lite TrainSession.

CreateTransferSession

static TrainSession *CreateTransferSession(const char *model_buf_backbone, size_t size_backbone, const char *model_buf_head, size_t size_head, lite::Context *context, bool train_mode = false);

Static method that creates the object pointer that points to the transfer learning training session.

Parameters
- model_buf_backbone: Pointer that points to the backbone network.
- size_backbone: The size of backbone network temporary saving buffer.
- model_buf_head: Pointer that points to the head network.
- size_head: The size of head network temporary saving buffer.
- context: Pointer that points to the target session.
- train_mode: Training mode to initialize Session.
Returns

Pointer that points to MindSpore Lite TrainSession.

static TrainSession *CreateTransferSession(const std::string &filename_backbone, const std::string &filename_head, lite::Context *context, bool train_mode = false);

Static method that creates the object pointer that points to the transfer learning training session.

Parameters
- filename_backbone: Filename of the backbone network to read flatbuffer.
- filename_head: Filename of the head network to read flatbuffer.
- context: Pointer that points to the target session.
- train_mode: Training mode to initialize Session.
Returns

Pointer that points to MindSpore Lite TrainSession.

ExportToBuf

virtual void *ExportToBuf(char *buf, size_t *len) const = 0;

Export the trained model into a buffer.

Parameters
- buf: The buffer to be exported into. If equal to nullptr, buf will be allocated.
- len: Size of the pre-allocated buffer, and the returned size of the exported buffer.
Returns

Pointer that points to MindSpore Lite TrainSession.

SaveToFile

virtual int SaveToFile(const std::string &filename) const = 0;

Save the trained model into a flatbuffer file.

Parameters
- filename: Filename of the file to save buffer.
Returns

0 represents success or -1 in case of error.

Train

virtual int Train() = 0;

Set model to train mode.

Returns

STATUS as an error code of compiling graph, STATUS is defined in errorcode.h

IsTrain

bool IsTrain() { return train_mode_ == true; }

Checks whether the current model is under the train mode.

Returns

Boolean indication if model is in train mode.

Eval

virtual int Eval() = 0;

Set model to eval mode.

Returns

STATUS as an error code of compiling graph, STATUS is defined in errorcode.h.

IsEval

bool IsEval() { return train_mode_ == false; }

Check mode of model.

Returns

Boolean indication if model is in eval mode.

SetLearningRate

virtual int SetLearningRate(float learning_rate) = 0;

Set the learning rate for the current model.

Returns

0 represents success or -1 in case of error.

GetLearningRate

virtual float GetLearningRate() = 0;

Get the learning rate of the current model.

Returns

The learning rate of the current model, default is 0.0.

SetupVirtualBatch

virtual int SetupVirtualBatch(int virtual_batch_multiplier, float lr = -1.0f, float momentum = -1.0f) = 0;

Customize the virtual batch size, in order to reduce memory consumption.

Parameters
- virtual_batch_multiplier: virtual batch number.
- lr: learning rate.
- momentum: momentum.
Returns

0 represents success or -1 in case of error.

GetPredictions

virtual std::vector<tensor::MSTensor *> GetPredictions() const = 0;

Get the predicting result of the trained model.

Returns

Return the pointer vector of prediction results.

SetLossName

void SetLossName(std::string loss_name) { loss_name_ = loss_name; }

Set the loss name.

Parameters
- loss_name: The name of loss kernels.

TrainLoop

#include <ltrain_loop.h>

Inherited from Session and used for reducing the RAM consumption during model training, user can set hyper-parameters and customized data preprocessing function.

Constructors & Destructors

~TrainLoop

virtual ~TrainLoop() = default;

Destructor function.

Public Member Functions

CreateTrainLoop

static TrainLoop *CreateTrainLoop(session::TrainSession *train_session, lite::Context *context, int batch_size = -1);

A static method of creating TrainLoop pointer.

Parameters
- train_session: Pointer that points to the CreateSession or CreateTransferSession object.
- context: Pointer that points to a context.
- batch_size: Batch size number.
Returns

Pointer that points to the TrainLoop object .

Reset

virtual int Reset() = 0;

Reset the epoch to 0.

Returns

0 means resetting successfully while -1 means failed.

train_session

virtual session::TrainSession *train_session() = 0;

Get the object of the current TrainSession.

Returns

Pointer that points to the object of TrainSession.

Init

virtual int Init(std::vector<mindspore::session::Metrics *> metrics) = 0;

Initialize the model evaluation matrix.

Parameters
- metrics: Pointer vector of the model evaluating matrix.
Returns

0 means initializing successfully while -1 means failed.

GetMetrics

virtual std::vector<mindspore::session::Metrics *> GetMetrics() = 0;

Get the model evaluation matrix.

Returns

Pointer vector of the model evaluation matrix.

SetKernelCallBack

virtual int SetKernelCallBack(const KernelCallBack &before, const KernelCallBack &after) = 0;

Set the callback function during training.

Parameters
- before: Callback pointer before execution.
- after: Callback pointer after execution.
Returns

0 means setting successfully while -1 means failed.

Train

virtual int Train(int epochs, mindspore::dataset::Dataset *dataset, std::vector<TrainLoopCallBack *> cbs, LoadDataFunc load_func = nullptr)= 0;

Execute training.

Parameters
- epochs: Training epoch number.
- dataset: Pointer that points to the MindData object.
- cbs: Object pointer vector.
- load_func: Class template function object.
Returns

0 means training successfully while -1 means failed.

Eval

virtual int Eval(mindspore::dataset::Dataset *dataset, std::vector<TrainLoopCallBack *> cbs, LoadDataFunc load_func = nullptr, int max_steps = INT_MAX) = 0;

Execute evaluating.

Parameters
- dataset: Pointer that points to the DataSet object.
- cbs: Object pointer vector.
- load_func: Class template function object.
- max_steps: Eval epoch number.
Returns

0 means evaluating successfully while -1 means failed.

TrainLoopCallback

#include <ltrain_loop_callback.h>

Execute the callback functions during the model training.

Constructors & Destructors

~TrainLoopCallback

virtual ~TrainLoopCallback() = default;

Destructor function.

Public Member Functions

Begin

virtual void Begin(const TrainLoopCallBackData &cb_data) {}

The method is called once before the network is executed.

Parameters
- cb_data: cb_data info about current execution.

End

virtual void End(const TrainLoopCallBackData &cb_data) {}

The method is called once after the network executed.

Parameters
- cb_data: cb_data info about current execution.

EpochBegin

virtual void EpochBegin(const TrainLoopCallBackData &cb_data) {}

The method is called at the beginning of each epoch.

Parameters
- cb_data: cb_data info about current execution.

EpochEnd

virtual int EpochEnd(const TrainLoopCallBackData &cb_data) { return RET_CONTINUE; }

The method is called at the end of each epoch.

Parameters
- cb_data: cb_data info about current execution.
Returns STATUS as an error code of compiling graph, STATUS is defined in errorcode.h.

StepBegin

virtual void StepBegin(const TrainLoopCallBackData &cb_data) {}

The method is called at the beginning of each step.

Parameters
- cb_data: cb_data info about current execution.

StepEnd

virtual void StepEnd(const TrainLoopCallBackData &cb_data) {}

The method is called after each step has finished.

Parameters
- cb_data: cb_data info about current execution.

Metrics

#include <metrics.h>

Evaluation metrics of the training model.

Constructors & Destructors

~Metrics

virtual ~Metrics() = default;

Destructor function.

Public Member Functions

Clear

virtual void Clear() {}

Reset the member variables total_accuracy_ and total_steps_ to 0.

Eval

virtual float Eval() {}

Evaluation the model.

Update

virtual void Update(std::vector<tensor::MSTensor *> inputs, std::vector<tensor::MSTensor *> outputs) = 0;

Update the member variables total_accuracy_ and total_steps_.