Module monk.gluon.finetune.level_12_losses_main

Classes

class prototype_losses (verbose=1)

Main class for all parameters in expert mode

Args

verbose : int

Set verbosity levels 0 - Print Nothing 1 - Print desired details

Expand source code

class prototype_losses(prototype_optimizers):
    '''
    Main class for all parameters in expert mode

    Args:
        verbose (int): Set verbosity levels
                        0 - Print Nothing
                        1 - Print desired details
    '''

    def __init__(self, verbose=1):
        super().__init__(verbose=verbose);


    ###############################################################################################################################################
    def loss_l1(self, weight=None, batch_axis=0):
        '''
        Select L1 Loss

        Args:
            weight (float): global scalar for weight loss
            batch_axis (int): Axis representing number of elements in the batch - N

        Returns:
            None
        '''
        self.system_dict = l1(self.system_dict, weight=weight, batch_axis=batch_axis);

        self.custom_print("Loss");
        self.custom_print("    Name:          {}".format(self.system_dict["hyper-parameters"]["loss"]["name"]));
        self.custom_print("    Params:        {}".format(self.system_dict["hyper-parameters"]["loss"]["params"]));
        self.custom_print("");
    ###############################################################################################################################################



    ###############################################################################################################################################
    def loss_l2(self, weight=1.0, batch_axis=0):
        '''
        Select L2 Loss

        Args:
            weight (float): global scalar for weight loss
            batch_axis (int): Axis representing number of elements in the batch - N

        Returns:
            None
        '''
        self.system_dict = l2(self.system_dict, weight=weight, batch_axis=batch_axis);

        self.custom_print("Loss");
        self.custom_print("    Name:          {}".format(self.system_dict["hyper-parameters"]["loss"]["name"]));
        self.custom_print("    Params:        {}".format(self.system_dict["hyper-parameters"]["loss"]["params"]));
        self.custom_print("");
    ###############################################################################################################################################



    ###############################################################################################################################################
    def loss_softmax_crossentropy(self, weight=None, batch_axis=0, axis_to_sum_over=-1, 
                                    label_as_categories=True, label_smoothing=False):
        '''
        Select soaftmax crossentropy Loss - Auto softmax before applying loss 

        Args:
            weight (float): global scalar for weight loss
            batch_axis (int): Axis representing number of elements in the batch - N
            axis_to_sum_over (int): Set as -1
            label_as_categories (bool): Fixed as True
            label_smoothing (bool): If True, label smoothning is applied.

        Returns:
            None
        '''
        self.system_dict = softmax_crossentropy(self.system_dict, weight=weight, batch_axis=batch_axis,
                                                axis_to_sum_over=axis_to_sum_over, label_as_categories=label_as_categories, 
                                                label_smoothing=label_smoothing);

        self.custom_print("Loss");
        self.custom_print("    Name:          {}".format(self.system_dict["hyper-parameters"]["loss"]["name"]));
        self.custom_print("    Params:        {}".format(self.system_dict["hyper-parameters"]["loss"]["params"]));
        self.custom_print("");
    ###############################################################################################################################################


    ###############################################################################################################################################
    def loss_crossentropy(self, weight=None, batch_axis=0, axis_to_sum_over=-1, 
                                    label_as_categories=True, label_smoothing=False):
        '''
        Select crossentropy Loss - Need to manually apply softmax

        Args:
            weight (float): global scalar for weight loss
            batch_axis (int): Axis representing number of elements in the batch - N
            axis_to_sum_over (int): Set as -1
            label_as_categories (bool): Fixed as True
            label_smoothing (bool): If True, label smoothning is applied.

        Returns:
            None
        '''
        self.system_dict = crossentropy(self.system_dict, weight=weight, batch_axis=batch_axis,
                                                axis_to_sum_over=axis_to_sum_over, label_as_categories=label_as_categories, 
                                                label_smoothing=label_smoothing);

        self.custom_print("Loss");
        self.custom_print("    Name:          {}".format(self.system_dict["hyper-parameters"]["loss"]["name"]));
        self.custom_print("    Params:        {}".format(self.system_dict["hyper-parameters"]["loss"]["params"]));
        self.custom_print("");
    ###############################################################################################################################################



    ###############################################################################################################################################
    def loss_sigmoid_binary_crossentropy(self, weight=None, batch_axis=0):
        '''
        Select sigmoid binary crossentropy Loss - Auto sigmoid before applying loss 

        Args:
            weight (float): global scalar for weight loss
            batch_axis (int): Axis representing number of elements in the batch - N

        Returns:
            None
        '''
        self.system_dict = sigmoid_binary_crossentropy(self.system_dict, weight=weight, batch_axis=batch_axis);

        self.custom_print("Loss");
        self.custom_print("    Name:          {}".format(self.system_dict["hyper-parameters"]["loss"]["name"]));
        self.custom_print("    Params:        {}".format(self.system_dict["hyper-parameters"]["loss"]["params"]));
        self.custom_print("");
    ###############################################################################################################################################



    ###############################################################################################################################################
    def loss_binary_crossentropy(self, weight=None, batch_axis=0):
        '''
        Select binary crossentropy Loss - Need to manually apply sigmoid

        Args:
            weight (float): global scalar for weight loss
            batch_axis (int): Axis representing number of elements in the batch - N

        Returns:
            None
        '''
        self.system_dict = binary_crossentropy(self.system_dict, weight=weight, batch_axis=batch_axis);

        self.custom_print("Loss");
        self.custom_print("    Name:          {}".format(self.system_dict["hyper-parameters"]["loss"]["name"]));
        self.custom_print("    Params:        {}".format(self.system_dict["hyper-parameters"]["loss"]["params"]));
        self.custom_print("");
    ###############################################################################################################################################


    ###############################################################################################################################################
    def loss_kldiv(self, log_pre_applied=False, weight=None, batch_axis=0, axis_to_sum_over=-1):
        '''
        Select lkdiv Loss

        Args:
            weight (float): global scalar for weight loss
            batch_axis (int): Axis representing number of elements in the batch - N
            axis_to_sum_over (int): Set as -1
            log_pre_applied (bool): If set as False, then logarithmic function is auto applied over target variables

        Returns:
            None
        '''
        self.system_dict = kldiv(self.system_dict, weight=weight, batch_axis=batch_axis,
                                axis_to_sum_over=axis_to_sum_over, log_pre_applied=log_pre_applied);

        self.custom_print("Loss");
        self.custom_print("    Name:          {}".format(self.system_dict["hyper-parameters"]["loss"]["name"]));
        self.custom_print("    Params:        {}".format(self.system_dict["hyper-parameters"]["loss"]["params"]));
        self.custom_print("");
    ###############################################################################################################################################


    ###############################################################################################################################################
    def loss_poisson_nll(self, log_pre_applied=False, weight=None, batch_axis=0):
        '''
        Select poisson_nll Loss

        Args:
            weight (float): global scalar for weight loss
            batch_axis (int): Axis representing number of elements in the batch - N
            log_pre_applied (bool): If set as False, then logarithmic function is auto applied over target variables

        Returns:
            None
        '''
        self.system_dict = poisson_nll(self.system_dict, log_pre_applied=log_pre_applied,
                                        weight=weight, batch_axis=batch_axis);

        self.custom_print("Loss");
        self.custom_print("    Name:          {}".format(self.system_dict["hyper-parameters"]["loss"]["name"]));
        self.custom_print("    Params:        {}".format(self.system_dict["hyper-parameters"]["loss"]["params"]));
        self.custom_print("");
    ###############################################################################################################################################


    ###############################################################################################################################################
    def loss_huber(self, weight=None, batch_axis=0, threshold_for_mean_estimator=1):
        '''
        Select huber Loss

        Args:
            weight (float): global scalar for weight loss
            batch_axis (int): Axis representing number of elements in the batch - N
            threshold_for_mean_estimator (int): Threshold for trimmed mean estimator.

        Returns:
            None
        '''
        self.system_dict = huber(self.system_dict, threshold_for_mean_estimator=threshold_for_mean_estimator,
                                weight=weight, batch_axis=batch_axis);

        self.custom_print("Loss");
        self.custom_print("    Name:          {}".format(self.system_dict["hyper-parameters"]["loss"]["name"]));
        self.custom_print("    Params:        {}".format(self.system_dict["hyper-parameters"]["loss"]["params"]));
        self.custom_print("");
    ###############################################################################################################################################


    ###############################################################################################################################################
    def loss_hinge(self, weight=None, batch_axis=0, margin=1):
        '''
        Select hinge Loss

        Args:
            weight (float): global scalar for weight loss
            batch_axis (int): Axis representing number of elements in the batch - N
            margin (float): MArgin value.

        Returns:
            None
        '''
        self.system_dict = hinge(self.system_dict, margin=margin,
                                weight=weight, batch_axis=batch_axis);

        self.custom_print("Loss");
        self.custom_print("    Name:          {}".format(self.system_dict["hyper-parameters"]["loss"]["name"]));
        self.custom_print("    Params:        {}".format(self.system_dict["hyper-parameters"]["loss"]["params"]));
        self.custom_print("");
    ###############################################################################################################################################



    ###############################################################################################################################################
    def loss_squared_hinge(self, weight=None, batch_axis=0, margin=1):
        '''
        Select squared hinge Loss

        Args:
            weight (float): global scalar for weight loss
            batch_axis (int): Axis representing number of elements in the batch - N
            margin (float): MArgin value.

        Returns:
            None
        '''
        self.system_dict = squared_hinge(self.system_dict, margin=margin,
                                weight=weight, batch_axis=batch_axis);

        self.custom_print("Loss");
        self.custom_print("    Name:          {}".format(self.system_dict["hyper-parameters"]["loss"]["name"]));
        self.custom_print("    Params:        {}".format(self.system_dict["hyper-parameters"]["loss"]["params"]));
        self.custom_print("");

Ancestors

• gluon.finetune.level_11_optimizers_main.prototype_optimizers
• gluon.finetune.level_10_schedulers_main.prototype_schedulers
• gluon.finetune.level_9_transforms_main.prototype_transforms
• gluon.finetune.level_8_layers_main.prototype_layers
• gluon.finetune.level_7_aux_main.prototype_aux
• gluon.finetune.level_6_params_main.prototype_params
• gluon.finetune.level_5_state_base.finetune_state
• gluon.finetune.level_4_evaluation_base.finetune_evaluation
• gluon.finetune.level_3_training_base.finetune_training
• gluon.finetune.level_2_model_base.finetune_model
• gluon.finetune.level_1_dataset_base.finetune_dataset
• system.base_class.system

Methods

def loss_binary_crossentropy(self, weight=None, batch_axis=0)

Select binary crossentropy Loss - Need to manually apply sigmoid

Args

weight : float

global scalar for weight loss

batch_axis : int

Axis representing number of elements in the batch - N

Returns

None

 

Expand source code

def loss_binary_crossentropy(self, weight=None, batch_axis=0):
    '''
    Select binary crossentropy Loss - Need to manually apply sigmoid

    Args:
        weight (float): global scalar for weight loss
        batch_axis (int): Axis representing number of elements in the batch - N

    Returns:
        None
    '''
    self.system_dict = binary_crossentropy(self.system_dict, weight=weight, batch_axis=batch_axis);

    self.custom_print("Loss");
    self.custom_print("    Name:          {}".format(self.system_dict["hyper-parameters"]["loss"]["name"]));
    self.custom_print("    Params:        {}".format(self.system_dict["hyper-parameters"]["loss"]["params"]));
    self.custom_print("");

def loss_crossentropy(self, weight=None, batch_axis=0, axis_to_sum_over=-1, label_as_categories=True, label_smoothing=False)

Select crossentropy Loss - Need to manually apply softmax

Args

weight : float

global scalar for weight loss

batch_axis : int

Axis representing number of elements in the batch - N

axis_to_sum_over : int

Set as -1

label_as_categories : bool

Fixed as True

label_smoothing : bool

If True, label smoothning is applied.

Returns

None

 

Expand source code

def loss_crossentropy(self, weight=None, batch_axis=0, axis_to_sum_over=-1, 
                                label_as_categories=True, label_smoothing=False):
    '''
    Select crossentropy Loss - Need to manually apply softmax

    Args:
        weight (float): global scalar for weight loss
        batch_axis (int): Axis representing number of elements in the batch - N
        axis_to_sum_over (int): Set as -1
        label_as_categories (bool): Fixed as True
        label_smoothing (bool): If True, label smoothning is applied.

    Returns:
        None
    '''
    self.system_dict = crossentropy(self.system_dict, weight=weight, batch_axis=batch_axis,
                                            axis_to_sum_over=axis_to_sum_over, label_as_categories=label_as_categories, 
                                            label_smoothing=label_smoothing);

    self.custom_print("Loss");
    self.custom_print("    Name:          {}".format(self.system_dict["hyper-parameters"]["loss"]["name"]));
    self.custom_print("    Params:        {}".format(self.system_dict["hyper-parameters"]["loss"]["params"]));
    self.custom_print("");

def loss_hinge(self, weight=None, batch_axis=0, margin=1)

Select hinge Loss

Args

weight : float

global scalar for weight loss

batch_axis : int

Axis representing number of elements in the batch - N

margin : float

MArgin value.

Returns

None

 

Expand source code

def loss_hinge(self, weight=None, batch_axis=0, margin=1):
    '''
    Select hinge Loss

    Args:
        weight (float): global scalar for weight loss
        batch_axis (int): Axis representing number of elements in the batch - N
        margin (float): MArgin value.

    Returns:
        None
    '''
    self.system_dict = hinge(self.system_dict, margin=margin,
                            weight=weight, batch_axis=batch_axis);

    self.custom_print("Loss");
    self.custom_print("    Name:          {}".format(self.system_dict["hyper-parameters"]["loss"]["name"]));
    self.custom_print("    Params:        {}".format(self.system_dict["hyper-parameters"]["loss"]["params"]));
    self.custom_print("");

def loss_huber(self, weight=None, batch_axis=0, threshold_for_mean_estimator=1)

Select huber Loss

Args

weight : float

global scalar for weight loss

batch_axis : int

Axis representing number of elements in the batch - N

threshold_for_mean_estimator : int

Threshold for trimmed mean estimator.

Returns

None

 

Expand source code

def loss_huber(self, weight=None, batch_axis=0, threshold_for_mean_estimator=1):
    '''
    Select huber Loss

    Args:
        weight (float): global scalar for weight loss
        batch_axis (int): Axis representing number of elements in the batch - N
        threshold_for_mean_estimator (int): Threshold for trimmed mean estimator.

    Returns:
        None
    '''
    self.system_dict = huber(self.system_dict, threshold_for_mean_estimator=threshold_for_mean_estimator,
                            weight=weight, batch_axis=batch_axis);

    self.custom_print("Loss");
    self.custom_print("    Name:          {}".format(self.system_dict["hyper-parameters"]["loss"]["name"]));
    self.custom_print("    Params:        {}".format(self.system_dict["hyper-parameters"]["loss"]["params"]));
    self.custom_print("");

def loss_kldiv(self, log_pre_applied=False, weight=None, batch_axis=0, axis_to_sum_over=-1)

Select lkdiv Loss

Args

weight : float

global scalar for weight loss

batch_axis : int

Axis representing number of elements in the batch - N

axis_to_sum_over : int

Set as -1

log_pre_applied : bool

If set as False, then logarithmic function is auto applied over target variables

Returns

None

 

Expand source code

def loss_kldiv(self, log_pre_applied=False, weight=None, batch_axis=0, axis_to_sum_over=-1):
    '''
    Select lkdiv Loss

    Args:
        weight (float): global scalar for weight loss
        batch_axis (int): Axis representing number of elements in the batch - N
        axis_to_sum_over (int): Set as -1
        log_pre_applied (bool): If set as False, then logarithmic function is auto applied over target variables

    Returns:
        None
    '''
    self.system_dict = kldiv(self.system_dict, weight=weight, batch_axis=batch_axis,
                            axis_to_sum_over=axis_to_sum_over, log_pre_applied=log_pre_applied);

    self.custom_print("Loss");
    self.custom_print("    Name:          {}".format(self.system_dict["hyper-parameters"]["loss"]["name"]));
    self.custom_print("    Params:        {}".format(self.system_dict["hyper-parameters"]["loss"]["params"]));
    self.custom_print("");

def loss_l1(self, weight=None, batch_axis=0)

Select L1 Loss

Args

weight : float

global scalar for weight loss

batch_axis : int

Axis representing number of elements in the batch - N

Returns

None

 

Expand source code

def loss_l1(self, weight=None, batch_axis=0):
    '''
    Select L1 Loss

    Args:
        weight (float): global scalar for weight loss
        batch_axis (int): Axis representing number of elements in the batch - N

    Returns:
        None
    '''
    self.system_dict = l1(self.system_dict, weight=weight, batch_axis=batch_axis);

    self.custom_print("Loss");
    self.custom_print("    Name:          {}".format(self.system_dict["hyper-parameters"]["loss"]["name"]));
    self.custom_print("    Params:        {}".format(self.system_dict["hyper-parameters"]["loss"]["params"]));
    self.custom_print("");

def loss_l2(self, weight=1.0, batch_axis=0)

Select L2 Loss

Args

weight : float

global scalar for weight loss

batch_axis : int

Axis representing number of elements in the batch - N

Returns

None

 

Expand source code

def loss_l2(self, weight=1.0, batch_axis=0):
    '''
    Select L2 Loss

    Args:
        weight (float): global scalar for weight loss
        batch_axis (int): Axis representing number of elements in the batch - N

    Returns:
        None
    '''
    self.system_dict = l2(self.system_dict, weight=weight, batch_axis=batch_axis);

    self.custom_print("Loss");
    self.custom_print("    Name:          {}".format(self.system_dict["hyper-parameters"]["loss"]["name"]));
    self.custom_print("    Params:        {}".format(self.system_dict["hyper-parameters"]["loss"]["params"]));
    self.custom_print("");

def loss_poisson_nll(self, log_pre_applied=False, weight=None, batch_axis=0)

Select poisson_nll Loss

Args

weight : float

global scalar for weight loss

batch_axis : int

Axis representing number of elements in the batch - N

log_pre_applied : bool

If set as False, then logarithmic function is auto applied over target variables

Returns

None

 

Expand source code

def loss_poisson_nll(self, log_pre_applied=False, weight=None, batch_axis=0):
    '''
    Select poisson_nll Loss

    Args:
        weight (float): global scalar for weight loss
        batch_axis (int): Axis representing number of elements in the batch - N
        log_pre_applied (bool): If set as False, then logarithmic function is auto applied over target variables

    Returns:
        None
    '''
    self.system_dict = poisson_nll(self.system_dict, log_pre_applied=log_pre_applied,
                                    weight=weight, batch_axis=batch_axis);

    self.custom_print("Loss");
    self.custom_print("    Name:          {}".format(self.system_dict["hyper-parameters"]["loss"]["name"]));
    self.custom_print("    Params:        {}".format(self.system_dict["hyper-parameters"]["loss"]["params"]));
    self.custom_print("");

def loss_sigmoid_binary_crossentropy(self, weight=None, batch_axis=0)

Select sigmoid binary crossentropy Loss - Auto sigmoid before applying loss

Args

weight : float

global scalar for weight loss

batch_axis : int

Axis representing number of elements in the batch - N

Returns

None

 

Expand source code

def loss_sigmoid_binary_crossentropy(self, weight=None, batch_axis=0):
    '''
    Select sigmoid binary crossentropy Loss - Auto sigmoid before applying loss 

    Args:
        weight (float): global scalar for weight loss
        batch_axis (int): Axis representing number of elements in the batch - N

    Returns:
        None
    '''
    self.system_dict = sigmoid_binary_crossentropy(self.system_dict, weight=weight, batch_axis=batch_axis);

    self.custom_print("Loss");
    self.custom_print("    Name:          {}".format(self.system_dict["hyper-parameters"]["loss"]["name"]));
    self.custom_print("    Params:        {}".format(self.system_dict["hyper-parameters"]["loss"]["params"]));
    self.custom_print("");

def loss_softmax_crossentropy(self, weight=None, batch_axis=0, axis_to_sum_over=-1, label_as_categories=True, label_smoothing=False)

Select soaftmax crossentropy Loss - Auto softmax before applying loss

Args

weight : float

global scalar for weight loss

batch_axis : int

Axis representing number of elements in the batch - N

axis_to_sum_over : int

Set as -1

label_as_categories : bool

Fixed as True

label_smoothing : bool

If True, label smoothning is applied.

Returns

None

 

Expand source code

def loss_softmax_crossentropy(self, weight=None, batch_axis=0, axis_to_sum_over=-1, 
                                label_as_categories=True, label_smoothing=False):
    '''
    Select soaftmax crossentropy Loss - Auto softmax before applying loss 

    Args:
        weight (float): global scalar for weight loss
        batch_axis (int): Axis representing number of elements in the batch - N
        axis_to_sum_over (int): Set as -1
        label_as_categories (bool): Fixed as True
        label_smoothing (bool): If True, label smoothning is applied.

    Returns:
        None
    '''
    self.system_dict = softmax_crossentropy(self.system_dict, weight=weight, batch_axis=batch_axis,
                                            axis_to_sum_over=axis_to_sum_over, label_as_categories=label_as_categories, 
                                            label_smoothing=label_smoothing);

    self.custom_print("Loss");
    self.custom_print("    Name:          {}".format(self.system_dict["hyper-parameters"]["loss"]["name"]));
    self.custom_print("    Params:        {}".format(self.system_dict["hyper-parameters"]["loss"]["params"]));
    self.custom_print("");

def loss_squared_hinge(self, weight=None, batch_axis=0, margin=1)

Select squared hinge Loss

Args

weight : float

global scalar for weight loss

batch_axis : int

Axis representing number of elements in the batch - N

margin : float

MArgin value.

Returns

None

 

Expand source code

def loss_squared_hinge(self, weight=None, batch_axis=0, margin=1):
    '''
    Select squared hinge Loss

    Args:
        weight (float): global scalar for weight loss
        batch_axis (int): Axis representing number of elements in the batch - N
        margin (float): MArgin value.

    Returns:
        None
    '''
    self.system_dict = squared_hinge(self.system_dict, margin=margin,
                            weight=weight, batch_axis=batch_axis);

    self.custom_print("Loss");
    self.custom_print("    Name:          {}".format(self.system_dict["hyper-parameters"]["loss"]["name"]));
    self.custom_print("    Params:        {}".format(self.system_dict["hyper-parameters"]["loss"]["params"]));
    self.custom_print("");