Skip to content

nezha

mindnlp.transformers.models.nezha.nezha

nezha model

mindnlp.transformers.models.nezha.nezha.NezhaAttention

Bases: Module

Nezha Attention

Source code in mindnlp/transformers/models/nezha/nezha.py 
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
class NezhaAttention(nn.Module):
    """Nezha Attention"""
    def __init__(self, config):
        """
        Initializes an instance of the NezhaAttention class.

        Args:
            self (object): The instance of the class itself.
            config (object): The configuration object that contains parameters for attention mechanism setup.

        Returns:
            None.

        Raises:
            None.
        """
        super().__init__()
        self.self = NezhaSelfAttention(config)
        self.output = NezhaSelfOutput(config)
        self.pruned_heads = set()

    def prune_heads(self, heads):
        """Prune heads"""
        if len(heads) == 0:
            return
        heads, index = find_pruneable_heads_and_indices(
            heads, self.self.num_attention_heads, self.self.attention_head_size, self.pruned_heads
        )

        #Prune linear layers
        self.self.query = prune_linear_layer(self.self.query, index)
        self.self.key = prune_linear_layer(self.self.key, index)
        self.self.value = prune_linear_layer(self.self.value, index)
        self.output.dense = prune_linear_layer(self.output.dense, index, dim=1)

        # Update hyper params and store pruned heads
        self.self.num_attention_heads = self.self.num_attention_heads - len(heads)
        self.self.all_head_size = self.self.attention_head_size * self.self.num_attention_heads
        self.pruned_heads = self.pruned_heads.union(heads)

    def forward(self, hidden_states, attention_mask = None,
                  head_mask = None, encoder_hidden_states = None,
                  encoder_attention_mask = None, past_key_value = None,
                  output_attentions = False):
        """
        Constructs the attention mechanism for the Nezha model.

        Args:
            self (NezhaAttention): The instance of NezhaAttention class.
            hidden_states (torch.Tensor): The input hidden states of the model.
                Shape: (batch_size, sequence_length, hidden_size).
            attention_mask (torch.Tensor, optional): The attention mask tensor.
                If provided, it should be a 2D tensor of shape (batch_size, sequence_length), where 1 indicates a token
                that should be attended to and 0 indicates a token that should not be attended to. Defaults to None.
            head_mask (torch.Tensor, optional): The head mask tensor.
                If provided, it should be a 1D tensor of shape (num_heads,), where 1 indicates a head that should be
                masked and 0 indicates a head that should not be masked. Defaults to None.
            encoder_hidden_states (torch.Tensor, optional): The hidden states of the encoder.
                Shape: (batch_size, sequence_length, hidden_size). Defaults to None.
            encoder_attention_mask (torch.Tensor, optional): The attention mask tensor for the encoder.
                If provided, it should be a 2D tensor of shape (batch_size, sequence_length), where 1 indicates a token
                that should be attended to and 0 indicates a token that should not be attended to. Defaults to None.
            past_key_value (tuple, optional): The cached key-value pairs of the previous time steps. Defaults to None.
            output_attentions (bool, optional): Whether to return attention weights. Defaults to False.

        Returns:
            tuple:
                A tuple containing:

                - attention_output (torch.Tensor): The output of the attention mechanism.
                Shape: (batch_size, sequence_length, hidden_size).
                - self_outputs (tuple): A tuple containing the outputs of the self-attention mechanism.
                - output_attentions (torch.Tensor, optional): The attention weights tensor.
                Shape: (num_attention_heads, batch_size, sequence_length, sequence_length).
                Only returned if `output_attentions` is set to True.

        Raises:
            None

        """
        self_outputs = self.self(
            hidden_states,
            attention_mask,
            head_mask,
            encoder_hidden_states,
            encoder_attention_mask,
            past_key_value,
            output_attentions,
        )
        attention_output = self.output(self_outputs[0], hidden_states)
        outputs = (attention_output,) + self_outputs[1:]  # add attentions if we output them
        return outputs

mindnlp.transformers.models.nezha.nezha.NezhaAttention.__init__(config)

Initializes an instance of the NezhaAttention class.

PARAMETER DESCRIPTION
self

The instance of the class itself.

TYPE: object

config

The configuration object that contains parameters for attention mechanism setup.

TYPE: object

RETURNS DESCRIPTION

None.
Source code in mindnlp/transformers/models/nezha/nezha.py 
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
def __init__(self, config):
    """
    Initializes an instance of the NezhaAttention class.

    Args:
        self (object): The instance of the class itself.
        config (object): The configuration object that contains parameters for attention mechanism setup.

    Returns:
        None.

    Raises:
        None.
    """
    super().__init__()
    self.self = NezhaSelfAttention(config)
    self.output = NezhaSelfOutput(config)
    self.pruned_heads = set()

mindnlp.transformers.models.nezha.nezha.NezhaAttention.forward(hidden_states, attention_mask=None, head_mask=None, encoder_hidden_states=None, encoder_attention_mask=None, past_key_value=None, output_attentions=False)

Constructs the attention mechanism for the Nezha model.

PARAMETER DESCRIPTION
self

The instance of NezhaAttention class.

TYPE: NezhaAttention

hidden_states

The input hidden states of the model. Shape: (batch_size, sequence_length, hidden_size).

TYPE: Tensor

attention_mask

The attention mask tensor. If provided, it should be a 2D tensor of shape (batch_size, sequence_length), where 1 indicates a token that should be attended to and 0 indicates a token that should not be attended to. Defaults to None.

TYPE: Tensor DEFAULT: None

head_mask

The head mask tensor. If provided, it should be a 1D tensor of shape (num_heads,), where 1 indicates a head that should be masked and 0 indicates a head that should not be masked. Defaults to None.

TYPE: Tensor DEFAULT: None

encoder_hidden_states

The hidden states of the encoder. Shape: (batch_size, sequence_length, hidden_size). Defaults to None.

TYPE: Tensor DEFAULT: None

encoder_attention_mask

The attention mask tensor for the encoder. If provided, it should be a 2D tensor of shape (batch_size, sequence_length), where 1 indicates a token that should be attended to and 0 indicates a token that should not be attended to. Defaults to None.

TYPE: Tensor DEFAULT: None

past_key_value

The cached key-value pairs of the previous time steps. Defaults to None.

TYPE: tuple DEFAULT: None

output_attentions

Whether to return attention weights. Defaults to False.

TYPE: bool DEFAULT: False

RETURNS DESCRIPTION
tuple

A tuple containing:

  • attention_output (torch.Tensor): The output of the attention mechanism. Shape: (batch_size, sequence_length, hidden_size).
  • self_outputs (tuple): A tuple containing the outputs of the self-attention mechanism.
  • output_attentions (torch.Tensor, optional): The attention weights tensor. Shape: (num_attention_heads, batch_size, sequence_length, sequence_length). Only returned if output_attentions is set to True.
Source code in mindnlp/transformers/models/nezha/nezha.py 
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
def forward(self, hidden_states, attention_mask = None,
              head_mask = None, encoder_hidden_states = None,
              encoder_attention_mask = None, past_key_value = None,
              output_attentions = False):
    """
    Constructs the attention mechanism for the Nezha model.

    Args:
        self (NezhaAttention): The instance of NezhaAttention class.
        hidden_states (torch.Tensor): The input hidden states of the model.
            Shape: (batch_size, sequence_length, hidden_size).
        attention_mask (torch.Tensor, optional): The attention mask tensor.
            If provided, it should be a 2D tensor of shape (batch_size, sequence_length), where 1 indicates a token
            that should be attended to and 0 indicates a token that should not be attended to. Defaults to None.
        head_mask (torch.Tensor, optional): The head mask tensor.
            If provided, it should be a 1D tensor of shape (num_heads,), where 1 indicates a head that should be
            masked and 0 indicates a head that should not be masked. Defaults to None.
        encoder_hidden_states (torch.Tensor, optional): The hidden states of the encoder.
            Shape: (batch_size, sequence_length, hidden_size). Defaults to None.
        encoder_attention_mask (torch.Tensor, optional): The attention mask tensor for the encoder.
            If provided, it should be a 2D tensor of shape (batch_size, sequence_length), where 1 indicates a token
            that should be attended to and 0 indicates a token that should not be attended to. Defaults to None.
        past_key_value (tuple, optional): The cached key-value pairs of the previous time steps. Defaults to None.
        output_attentions (bool, optional): Whether to return attention weights. Defaults to False.

    Returns:
        tuple:
            A tuple containing:

            - attention_output (torch.Tensor): The output of the attention mechanism.
            Shape: (batch_size, sequence_length, hidden_size).
            - self_outputs (tuple): A tuple containing the outputs of the self-attention mechanism.
            - output_attentions (torch.Tensor, optional): The attention weights tensor.
            Shape: (num_attention_heads, batch_size, sequence_length, sequence_length).
            Only returned if `output_attentions` is set to True.

    Raises:
        None

    """
    self_outputs = self.self(
        hidden_states,
        attention_mask,
        head_mask,
        encoder_hidden_states,
        encoder_attention_mask,
        past_key_value,
        output_attentions,
    )
    attention_output = self.output(self_outputs[0], hidden_states)
    outputs = (attention_output,) + self_outputs[1:]  # add attentions if we output them
    return outputs

mindnlp.transformers.models.nezha.nezha.NezhaAttention.prune_heads(heads)

Prune heads

Source code in mindnlp/transformers/models/nezha/nezha.py 
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
def prune_heads(self, heads):
    """Prune heads"""
    if len(heads) == 0:
        return
    heads, index = find_pruneable_heads_and_indices(
        heads, self.self.num_attention_heads, self.self.attention_head_size, self.pruned_heads
    )

    #Prune linear layers
    self.self.query = prune_linear_layer(self.self.query, index)
    self.self.key = prune_linear_layer(self.self.key, index)
    self.self.value = prune_linear_layer(self.self.value, index)
    self.output.dense = prune_linear_layer(self.output.dense, index, dim=1)

    # Update hyper params and store pruned heads
    self.self.num_attention_heads = self.self.num_attention_heads - len(heads)
    self.self.all_head_size = self.self.attention_head_size * self.self.num_attention_heads
    self.pruned_heads = self.pruned_heads.union(heads)

mindnlp.transformers.models.nezha.nezha.NezhaEmbeddings

Bases: Module

Construct the embeddings from word and token_type embeddings.

Source code in mindnlp/transformers/models/nezha/nezha.py 
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
class NezhaEmbeddings(nn.Module):
    """Construct the embeddings from word and token_type embeddings."""
    def __init__(self, config):
        """
        Initialize the NezhaEmbeddings class.

        Args:
            self: Instance of the NezhaEmbeddings class.
            config (object):
                Configuration object containing parameters for initializing embeddings.

                - vocab_size (int): Size of the vocabulary.
                - hidden_size (int): Size of the hidden layer.
                - pad_token_id (int): Index of the padding token in the vocabulary.
                - type_vocab_size (int): Size of the type vocabulary.
                - layer_norm_eps (float): Epsilon value for layer normalization.
                - hidden_dropout_prob (float): Probability of dropout.
                - max_position_embeddings (int): Maximum number of position embeddings.

        Returns:
            None.

        Raises:
            ValueError: If any of the configuration parameters are invalid.
            AttributeError: If there are issues with attribute assignments.
            RuntimeError: If there are runtime errors during initialization.
        """
        super().__init__()
        self.word_embeddings = nn.Embedding(config.vocab_size, config.hidden_size,
                                            padding_idx=config.pad_token_id)
        self.token_type_embeddings = nn.Embedding(config.type_vocab_size, config.hidden_size)
        self.LayerNorm = nn.LayerNorm((config.hidden_size,), eps=config.layer_norm_eps)
        self.dropout = nn.Dropout(p=config.hidden_dropout_prob)
        self.token_type_ids = ops.zeros((1, config.max_position_embeddings), dtype=mindspore.int64)

    def forward(self, input_ids = None, token_type_ids = None, inputs_embeds = None):
        """
        This method forwards Nezha embeddings based on the input_ids, token_type_ids, and inputs_embeds.

        Args:
            self: The instance of the class.
            input_ids (Tensor, optional): The input tensor representing the tokenized input sequence.
                Default is None.
            token_type_ids (Tensor, optional): The input tensor representing the type of each token in the input sequence.
                Default is None.
            inputs_embeds (Tensor, optional): The input tensor containing precomputed embeddings for the input sequence.
                Default is None.

        Returns:
            None.

        Raises:
            ValueError: If input_ids and inputs_embeds have incompatible shapes.
            ValueError: If token_type_ids and input_shape have incompatible shapes.
            ValueError: If token_type_ids and buffered_token_type_ids_expanded have incompatible shapes.
            TypeError: If input_ids, token_type_ids, or inputs_embeds are not of type Tensor.
            TypeError: If token_type_ids or input_ids are not of type Tensor.
            TypeError: If token_type_ids or buffered_token_type_ids_expanded are not of type Tensor.
            TypeError: If input_shape is not of type tuple.
            TypeError: If seq_length is not of type int.
            RuntimeError: If self.token_type_embeddings or self.LayerNorm encounters a runtime error.
        """
        if input_ids is not None:
            input_shape = input_ids.shape
        else:
            input_shape = inputs_embeds.shape[:-1]

        seq_length = input_shape[1]

        if inputs_embeds is None:
            inputs_embeds = self.word_embeddings(input_ids)

        if token_type_ids is None:
            token_type_ids = ops.zeros_like(input_ids)

        if token_type_ids is None:
            if hasattr(self, "token_type_ids"):
                buffered_token_type_ids = self.token_type_ids[:, :seq_length]
                buffered_token_type_ids_expanded = ops.broadcast_to(buffered_token_type_ids,
                                                                    (input_shape[0], seq_length))
                token_type_ids = buffered_token_type_ids_expanded
            else:
                token_type_ids = ops.zeros(input_shape, dtype=mindspore.int64)

        token_type_embeddings = self.token_type_embeddings(token_type_ids)
        embeddings = inputs_embeds + token_type_embeddings
        embeddings = self.LayerNorm(embeddings)
        embeddings = self.dropout(embeddings)
        return embeddings

mindnlp.transformers.models.nezha.nezha.NezhaEmbeddings.__init__(config)

Initialize the NezhaEmbeddings class.

PARAMETER DESCRIPTION
self

Instance of the NezhaEmbeddings class.
config

Configuration object containing parameters for initializing embeddings.

  • vocab_size (int): Size of the vocabulary.
  • hidden_size (int): Size of the hidden layer.
  • pad_token_id (int): Index of the padding token in the vocabulary.
  • type_vocab_size (int): Size of the type vocabulary.
  • layer_norm_eps (float): Epsilon value for layer normalization.
  • hidden_dropout_prob (float): Probability of dropout.
  • max_position_embeddings (int): Maximum number of position embeddings.

TYPE: object

RETURNS DESCRIPTION

None.
RAISES DESCRIPTION
ValueError

If any of the configuration parameters are invalid.
AttributeError

If there are issues with attribute assignments.
RuntimeError

If there are runtime errors during initialization.
Source code in mindnlp/transformers/models/nezha/nezha.py 
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
def __init__(self, config):
    """
    Initialize the NezhaEmbeddings class.

    Args:
        self: Instance of the NezhaEmbeddings class.
        config (object):
            Configuration object containing parameters for initializing embeddings.

            - vocab_size (int): Size of the vocabulary.
            - hidden_size (int): Size of the hidden layer.
            - pad_token_id (int): Index of the padding token in the vocabulary.
            - type_vocab_size (int): Size of the type vocabulary.
            - layer_norm_eps (float): Epsilon value for layer normalization.
            - hidden_dropout_prob (float): Probability of dropout.
            - max_position_embeddings (int): Maximum number of position embeddings.

    Returns:
        None.

    Raises:
        ValueError: If any of the configuration parameters are invalid.
        AttributeError: If there are issues with attribute assignments.
        RuntimeError: If there are runtime errors during initialization.
    """
    super().__init__()
    self.word_embeddings = nn.Embedding(config.vocab_size, config.hidden_size,
                                        padding_idx=config.pad_token_id)
    self.token_type_embeddings = nn.Embedding(config.type_vocab_size, config.hidden_size)
    self.LayerNorm = nn.LayerNorm((config.hidden_size,), eps=config.layer_norm_eps)
    self.dropout = nn.Dropout(p=config.hidden_dropout_prob)
    self.token_type_ids = ops.zeros((1, config.max_position_embeddings), dtype=mindspore.int64)

mindnlp.transformers.models.nezha.nezha.NezhaEmbeddings.forward(input_ids=None, token_type_ids=None, inputs_embeds=None)

This method forwards Nezha embeddings based on the input_ids, token_type_ids, and inputs_embeds.

PARAMETER DESCRIPTION
self

The instance of the class.
input_ids

The input tensor representing the tokenized input sequence. Default is None.

TYPE: Tensor DEFAULT: None

token_type_ids

The input tensor representing the type of each token in the input sequence. Default is None.

TYPE: Tensor DEFAULT: None

inputs_embeds

The input tensor containing precomputed embeddings for the input sequence. Default is None.

TYPE: Tensor DEFAULT: None

RETURNS DESCRIPTION

None.
RAISES DESCRIPTION
ValueError

If input_ids and inputs_embeds have incompatible shapes.
ValueError

If token_type_ids and input_shape have incompatible shapes.
ValueError

If token_type_ids and buffered_token_type_ids_expanded have incompatible shapes.
TypeError

If input_ids, token_type_ids, or inputs_embeds are not of type Tensor.
TypeError

If token_type_ids or input_ids are not of type Tensor.
TypeError

If token_type_ids or buffered_token_type_ids_expanded are not of type Tensor.
TypeError

If input_shape is not of type tuple.
TypeError

If seq_length is not of type int.
RuntimeError

If self.token_type_embeddings or self.LayerNorm encounters a runtime error.
Source code in mindnlp/transformers/models/nezha/nezha.py 
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
def forward(self, input_ids = None, token_type_ids = None, inputs_embeds = None):
    """
    This method forwards Nezha embeddings based on the input_ids, token_type_ids, and inputs_embeds.

    Args:
        self: The instance of the class.
        input_ids (Tensor, optional): The input tensor representing the tokenized input sequence.
            Default is None.
        token_type_ids (Tensor, optional): The input tensor representing the type of each token in the input sequence.
            Default is None.
        inputs_embeds (Tensor, optional): The input tensor containing precomputed embeddings for the input sequence.
            Default is None.

    Returns:
        None.

    Raises:
        ValueError: If input_ids and inputs_embeds have incompatible shapes.
        ValueError: If token_type_ids and input_shape have incompatible shapes.
        ValueError: If token_type_ids and buffered_token_type_ids_expanded have incompatible shapes.
        TypeError: If input_ids, token_type_ids, or inputs_embeds are not of type Tensor.
        TypeError: If token_type_ids or input_ids are not of type Tensor.
        TypeError: If token_type_ids or buffered_token_type_ids_expanded are not of type Tensor.
        TypeError: If input_shape is not of type tuple.
        TypeError: If seq_length is not of type int.
        RuntimeError: If self.token_type_embeddings or self.LayerNorm encounters a runtime error.
    """
    if input_ids is not None:
        input_shape = input_ids.shape
    else:
        input_shape = inputs_embeds.shape[:-1]

    seq_length = input_shape[1]

    if inputs_embeds is None:
        inputs_embeds = self.word_embeddings(input_ids)

    if token_type_ids is None:
        token_type_ids = ops.zeros_like(input_ids)

    if token_type_ids is None:
        if hasattr(self, "token_type_ids"):
            buffered_token_type_ids = self.token_type_ids[:, :seq_length]
            buffered_token_type_ids_expanded = ops.broadcast_to(buffered_token_type_ids,
                                                                (input_shape[0], seq_length))
            token_type_ids = buffered_token_type_ids_expanded
        else:
            token_type_ids = ops.zeros(input_shape, dtype=mindspore.int64)

    token_type_embeddings = self.token_type_embeddings(token_type_ids)
    embeddings = inputs_embeds + token_type_embeddings
    embeddings = self.LayerNorm(embeddings)
    embeddings = self.dropout(embeddings)
    return embeddings

mindnlp.transformers.models.nezha.nezha.NezhaEncoder

Bases: Module

Nezha Encoder

Source code in mindnlp/transformers/models/nezha/nezha.py 
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
877
class NezhaEncoder(nn.Module):
    """Nezha Encoder"""
    def __init__(self, config):
        """
        Initializes a NezhaEncoder instance with the provided configuration.

        Args:
            self (NezhaEncoder): The NezhaEncoder instance itself.
            config (dict): A dictionary containing the configuration parameters for the NezhaEncoder.
                This dictionary should include the following keys:

                - num_hidden_layers (int): The number of hidden layers in the NezhaEncoder configuration.

        Returns:
            None.

        Raises:
            TypeError: If the input parameters are not of the expected types.
            ValueError: If the configuration dictionary is missing required keys or if the values are invalid.
        """
        super().__init__()
        self.config = config
        self.layer = nn.ModuleList([NezhaLayer(config) for _ in range(config.num_hidden_layers)])
        self.gradient_checkpointing = False

    def forward(self, hidden_states, attention_mask = None,
                  head_mask = None, encoder_hidden_states = None,
                  encoder_attention_mask = None, past_key_values = None,
                  use_cache = None, output_attentions = False,
                  output_hidden_states = False):
        """
        Constructs the NezhaEncoder.

        Args:
            self: The NezhaEncoder object.
            hidden_states (Tensor): The input hidden states of the encoder.
            attention_mask (Tensor, optional): An attention mask tensor. Defaults to None.
            head_mask (List[Tensor], optional): A list of attention mask tensors for each layer. Defaults to None.
            encoder_hidden_states (Tensor, optional): The hidden states of the encoder. Defaults to None.
            encoder_attention_mask (Tensor, optional): An attention mask tensor for the encoder. Defaults to None.
            past_key_values (Tuple[Tensor], optional): A tuple of key-value tensors from previous decoder outputs.
                Defaults to None.
            use_cache (bool, optional): Whether to use cache. Defaults to None.
            output_attentions (bool, optional): Whether to output attention tensors. Defaults to False.
            output_hidden_states (bool, optional): Whether to output hidden states of each layer. Defaults to False.

        Returns:
            Tuple[Tensor, Tuple[Tensor], Optional[Tuple[Tensor]], Optional[Tuple[Tensor]], Optional[Tuple[Tensor]]]:

                - hidden_states (Tensor): The output hidden states of the encoder.
                - next_decoder_cache (Tuple[Tensor]): The cache for the next decoder.
                - all_hidden_states (Optional[Tuple[Tensor]]): The hidden states of each layer.
                None if `output_hidden_states` is False.
                - all_self_attentions (Optional[Tuple[Tensor]]): The attention tensors of each layer.
                None if `output_attentions` is False.
                - all_cross_attentions (Optional[Tuple[Tensor]]): The cross-attention tensors of each layer.
                None if `output_attentions` is False or `self.config.add_cross_attention` is False.

        Raises:
            None.
        """
        all_hidden_states = () if output_hidden_states else None
        all_self_attentions = () if output_attentions else None
        all_cross_attentions = () if output_attentions and self.config.add_cross_attention else None

        if self.gradient_checkpointing and self.training:
            if use_cache:
                logger.warning_once(
                    "`use_cache=True` is incompatible with gradient checkpointing. Setting `use_cache=False`..."
                )
                use_cache = False

        next_decoder_cache = () if use_cache else None
        for i, layer_module in enumerate(self.layer):
            if output_hidden_states:
                all_hidden_states = all_hidden_states + (hidden_states,)

            layer_head_mask = head_mask[i] if head_mask is not None else None
            past_key_value = past_key_values[i] if past_key_values is not None else None

            # TODO
            # if self.gradient_checkpointing and self.training:

            #     def create_custom_forward(module):
            #         def custom_forward(*inputs):
            #             return module(*inputs, past_key_value, output_attentions)

            #         return custom_forward
            # layer_outputs = torch.utils.checkpoint.checkpoint(
            #         create_custom_forward(layer_module),
            #         hidden_states,
            #         attention_mask,
            #         layer_head_mask,
            #         encoder_hidden_states,
            #         encoder_attention_mask,
            #     )
            # else:
            layer_outputs = layer_module(
                hidden_states,
                attention_mask,
                layer_head_mask,
                encoder_hidden_states,
                encoder_attention_mask,
                past_key_value,
                output_attentions,
            )

            hidden_states = layer_outputs[0]
            if use_cache:
                next_decoder_cache += (layer_outputs[-1],)
            if output_attentions:
                all_self_attentions = all_self_attentions + (layer_outputs[1],)
                if self.config.add_cross_attention:
                    all_cross_attentions = all_cross_attentions + (layer_outputs[2],)

        if output_hidden_states:
            all_hidden_states = all_hidden_states + (hidden_states,)

        return tuple(
            v
            for v in [
                hidden_states,
                next_decoder_cache,
                all_hidden_states,
                all_self_attentions,
                all_cross_attentions,
            ]
            if v is not None
        )

mindnlp.transformers.models.nezha.nezha.NezhaEncoder.__init__(config)

Initializes a NezhaEncoder instance with the provided configuration.

PARAMETER DESCRIPTION
self

The NezhaEncoder instance itself.

TYPE: NezhaEncoder

config

A dictionary containing the configuration parameters for the NezhaEncoder. This dictionary should include the following keys:

  • num_hidden_layers (int): The number of hidden layers in the NezhaEncoder configuration.

TYPE: dict

RETURNS DESCRIPTION

None.
RAISES DESCRIPTION
TypeError

If the input parameters are not of the expected types.
ValueError

If the configuration dictionary is missing required keys or if the values are invalid.
Source code in mindnlp/transformers/models/nezha/nezha.py 
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
def __init__(self, config):
    """
    Initializes a NezhaEncoder instance with the provided configuration.

    Args:
        self (NezhaEncoder): The NezhaEncoder instance itself.
        config (dict): A dictionary containing the configuration parameters for the NezhaEncoder.
            This dictionary should include the following keys:

            - num_hidden_layers (int): The number of hidden layers in the NezhaEncoder configuration.

    Returns:
        None.

    Raises:
        TypeError: If the input parameters are not of the expected types.
        ValueError: If the configuration dictionary is missing required keys or if the values are invalid.
    """
    super().__init__()
    self.config = config
    self.layer = nn.ModuleList([NezhaLayer(config) for _ in range(config.num_hidden_layers)])
    self.gradient_checkpointing = False

mindnlp.transformers.models.nezha.nezha.NezhaEncoder.forward(hidden_states, attention_mask=None, head_mask=None, encoder_hidden_states=None, encoder_attention_mask=None, past_key_values=None, use_cache=None, output_attentions=False, output_hidden_states=False)

Constructs the NezhaEncoder.

PARAMETER DESCRIPTION
self

The NezhaEncoder object.
hidden_states

The input hidden states of the encoder.

TYPE: Tensor

attention_mask

An attention mask tensor. Defaults to None.

TYPE: Tensor DEFAULT: None

head_mask

A list of attention mask tensors for each layer. Defaults to None.

TYPE: List[Tensor] DEFAULT: None

encoder_hidden_states

The hidden states of the encoder. Defaults to None.

TYPE: Tensor DEFAULT: None

encoder_attention_mask

An attention mask tensor for the encoder. Defaults to None.

TYPE: Tensor DEFAULT: None

past_key_values

A tuple of key-value tensors from previous decoder outputs. Defaults to None.

TYPE: Tuple[Tensor] DEFAULT: None

use_cache

Whether to use cache. Defaults to None.

TYPE: bool DEFAULT: None

output_attentions

Whether to output attention tensors. Defaults to False.

TYPE: bool DEFAULT: False

output_hidden_states

Whether to output hidden states of each layer. Defaults to False.

TYPE: bool DEFAULT: False

RETURNS DESCRIPTION

Tuple[Tensor, Tuple[Tensor], Optional[Tuple[Tensor]], Optional[Tuple[Tensor]], Optional[Tuple[Tensor]]]:

  • hidden_states (Tensor): The output hidden states of the encoder.
  • next_decoder_cache (Tuple[Tensor]): The cache for the next decoder.
  • all_hidden_states (Optional[Tuple[Tensor]]): The hidden states of each layer. None if output_hidden_states is False.
  • all_self_attentions (Optional[Tuple[Tensor]]): The attention tensors of each layer. None if output_attentions is False.
  • all_cross_attentions (Optional[Tuple[Tensor]]): The cross-attention tensors of each layer. None if output_attentions is False or self.config.add_cross_attention is False.
Source code in mindnlp/transformers/models/nezha/nezha.py 
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
877
def forward(self, hidden_states, attention_mask = None,
              head_mask = None, encoder_hidden_states = None,
              encoder_attention_mask = None, past_key_values = None,
              use_cache = None, output_attentions = False,
              output_hidden_states = False):
    """
    Constructs the NezhaEncoder.

    Args:
        self: The NezhaEncoder object.
        hidden_states (Tensor): The input hidden states of the encoder.
        attention_mask (Tensor, optional): An attention mask tensor. Defaults to None.
        head_mask (List[Tensor], optional): A list of attention mask tensors for each layer. Defaults to None.
        encoder_hidden_states (Tensor, optional): The hidden states of the encoder. Defaults to None.
        encoder_attention_mask (Tensor, optional): An attention mask tensor for the encoder. Defaults to None.
        past_key_values (Tuple[Tensor], optional): A tuple of key-value tensors from previous decoder outputs.
            Defaults to None.
        use_cache (bool, optional): Whether to use cache. Defaults to None.
        output_attentions (bool, optional): Whether to output attention tensors. Defaults to False.
        output_hidden_states (bool, optional): Whether to output hidden states of each layer. Defaults to False.

    Returns:
        Tuple[Tensor, Tuple[Tensor], Optional[Tuple[Tensor]], Optional[Tuple[Tensor]], Optional[Tuple[Tensor]]]:

            - hidden_states (Tensor): The output hidden states of the encoder.
            - next_decoder_cache (Tuple[Tensor]): The cache for the next decoder.
            - all_hidden_states (Optional[Tuple[Tensor]]): The hidden states of each layer.
            None if `output_hidden_states` is False.
            - all_self_attentions (Optional[Tuple[Tensor]]): The attention tensors of each layer.
            None if `output_attentions` is False.
            - all_cross_attentions (Optional[Tuple[Tensor]]): The cross-attention tensors of each layer.
            None if `output_attentions` is False or `self.config.add_cross_attention` is False.

    Raises:
        None.
    """
    all_hidden_states = () if output_hidden_states else None
    all_self_attentions = () if output_attentions else None
    all_cross_attentions = () if output_attentions and self.config.add_cross_attention else None

    if self.gradient_checkpointing and self.training:
        if use_cache:
            logger.warning_once(
                "`use_cache=True` is incompatible with gradient checkpointing. Setting `use_cache=False`..."
            )
            use_cache = False

    next_decoder_cache = () if use_cache else None
    for i, layer_module in enumerate(self.layer):
        if output_hidden_states:
            all_hidden_states = all_hidden_states + (hidden_states,)

        layer_head_mask = head_mask[i] if head_mask is not None else None
        past_key_value = past_key_values[i] if past_key_values is not None else None

        # TODO
        # if self.gradient_checkpointing and self.training:

        #     def create_custom_forward(module):
        #         def custom_forward(*inputs):
        #             return module(*inputs, past_key_value, output_attentions)

        #         return custom_forward
        # layer_outputs = torch.utils.checkpoint.checkpoint(
        #         create_custom_forward(layer_module),
        #         hidden_states,
        #         attention_mask,
        #         layer_head_mask,
        #         encoder_hidden_states,
        #         encoder_attention_mask,
        #     )
        # else:
        layer_outputs = layer_module(
            hidden_states,
            attention_mask,
            layer_head_mask,
            encoder_hidden_states,
            encoder_attention_mask,
            past_key_value,
            output_attentions,
        )

        hidden_states = layer_outputs[0]
        if use_cache:
            next_decoder_cache += (layer_outputs[-1],)
        if output_attentions:
            all_self_attentions = all_self_attentions + (layer_outputs[1],)
            if self.config.add_cross_attention:
                all_cross_attentions = all_cross_attentions + (layer_outputs[2],)

    if output_hidden_states:
        all_hidden_states = all_hidden_states + (hidden_states,)

    return tuple(
        v
        for v in [
            hidden_states,
            next_decoder_cache,
            all_hidden_states,
            all_self_attentions,
            all_cross_attentions,
        ]
        if v is not None
    )

mindnlp.transformers.models.nezha.nezha.NezhaForMaskedLM

Bases: NezhaPreTrainedModel

NezhaForMaskedLM

Source code in mindnlp/transformers/models/nezha/nezha.py 
1530
1531
1532
1533
1534
1535
1536
1537
1538
1539
1540
1541
1542
1543
1544
1545
1546
1547
1548
1549
1550
1551
1552
1553
1554
1555
1556
1557
1558
1559
1560
1561
1562
1563
1564
1565
1566
1567
1568
1569
1570
1571
1572
1573
1574
1575
1576
1577
1578
1579
1580
1581
1582
1583
1584
1585
1586
1587
1588
1589
1590
1591
1592
1593
1594
1595
1596
1597
1598
1599
1600
1601
1602
1603
1604
1605
1606
1607
1608
1609
1610
1611
1612
1613
1614
1615
1616
1617
1618
1619
1620
1621
1622
1623
1624
1625
1626
1627
1628
1629
1630
1631
1632
1633
1634
1635
1636
1637
1638
1639
1640
1641
1642
1643
1644
1645
1646
class NezhaForMaskedLM(NezhaPreTrainedModel):
    """NezhaForMaskedLM"""
    _keys_to_ignore_on_load_unexpected = [r"pooler"]
    _keys_to_ignore_on_load_missing = [r"cls.predictions.decoder", r"positions_encoding"]

    def __init__(self, config):
        """
        Initializes a new NezhaForMaskedLM instance.

        Args:
            self (object): The NezhaForMaskedLM instance itself.
            config (object): An instance of the configuration class containing the model configuration settings.
                It is used to customize the behavior of the NezhaForMaskedLM model.
                Must have the property 'is_decoder' to determine if the model is a decoder.
                If 'is_decoder' is True, a warning will be logged regarding the bidirectional
                self-attention configuration.

        Returns:
            None.

        Raises:
            None.
        """
        super().__init__(config)

        if config.is_decoder:
            logger.warning(
                "If you want to use `NezhaForMaskedLM` make sure `config.is_decoder=False` for "
                "bi-directional self-attention."
            )

        self.nezha = NezhaModel(config, add_pooling_layer=False)
        self.cls = NezhaOnlyMLMHead(config)

    def get_output_embeddings(self):
        """get output embeddings"""
        return self.cls.predictions.decoder

    def set_output_embeddings(self, new_embeddings):
        """set output embeddings"""
        self.cls.predictions.decoder = new_embeddings

    def forward(self, input_ids = None, attention_mask = None,
                  token_type_ids = None, head_mask = None, inputs_embeds = None,
                  encoder_hidden_states = None, encoder_attention_mask = None,
                  labels = None, output_attentions = None, output_hidden_states = None):
        """
        Constructs the Nezha model for masked language modeling (MLM).

        Args:
            self (NezhaForMaskedLM): The instance of the NezhaForMaskedLM class.
            input_ids (torch.Tensor, optional): The input tensor containing the tokenized input sequence IDs.
                Default: None.
            attention_mask (torch.Tensor, optional): The attention mask tensor to indicate which tokens
                should be attended to. Default: None.
            token_type_ids (torch.Tensor, optional): The token type IDs tensor to distinguish different parts
                of the input. Default: None.
            head_mask (torch.Tensor, optional): The head mask tensor to mask specific attention heads. Default: None.
            inputs_embeds (torch.Tensor, optional): The embedded inputs tensor. Default: None.
            encoder_hidden_states (torch.Tensor, optional): The hidden states of the encoder. Default: None.
            encoder_attention_mask (torch.Tensor, optional): The attention mask for the encoder. Default: None.
            labels (torch.Tensor, optional): The tensor containing the labels for the masked language modeling task.
                Default: None.
            output_attentions (bool, optional): Whether to output attentions. Default: None.
            output_hidden_states (bool, optional): Whether to output hidden states. Default: None.

        Returns:
            tuple:
                A tuple containing the masked language modeling loss (if labels are provided)
                and the output of the model.

                - masked_lm_loss (torch.Tensor): The masked language modeling loss.
                - prediction_scores (torch.Tensor): The predicted scores for each token.
                - outputs[2:] (tuple): Additional outputs from the model. (output_attentions, output_hidden_states, ...)

        Raises:
            None.
        """
        outputs = self.nezha(
            input_ids,
            attention_mask=attention_mask,
            token_type_ids=token_type_ids,
            head_mask=head_mask,
            inputs_embeds=inputs_embeds,
            encoder_hidden_states=encoder_hidden_states,
            encoder_attention_mask=encoder_attention_mask,
            output_attentions=output_attentions,
            output_hidden_states=output_hidden_states,
        )

        sequence_output = outputs[0]
        prediction_scores = self.cls(sequence_output)

        masked_lm_loss = None
        if labels is not None:
            loss_fct = nn.CrossEntropyLoss()  # -100 index = padding token
            masked_lm_loss = loss_fct(prediction_scores.view(-1, self.config.vocab_size), labels.view(-1))

        output = (prediction_scores,) + outputs[2:]
        return ((masked_lm_loss,) + output) if masked_lm_loss is not None else output

    def prepare_inputs_for_generation(self, input_ids, attention_mask=None):
        """prepare inputs for generation"""
        input_shape = input_ids.shape
        effective_batch_size = input_shape[0]

        #  add a dummy token
        if self.config.pad_token_id is None:
            raise ValueError("The PAD token should be defined for generation")

        attention_mask = ops.cat([attention_mask, attention_mask.new_zeros((attention_mask.shape[0], 1))], axis=-1)
        dummy_token = ops.full(
            (effective_batch_size, 1), self.config.pad_token_id, dtype=mindspore.int64
        )
        input_ids = ops.cat([input_ids, dummy_token], axis=1)

        return {"input_ids": input_ids, "attention_mask": attention_mask}

mindnlp.transformers.models.nezha.nezha.NezhaForMaskedLM.__init__(config)

Initializes a new NezhaForMaskedLM instance.

PARAMETER DESCRIPTION
self

The NezhaForMaskedLM instance itself.

TYPE: object

config

An instance of the configuration class containing the model configuration settings. It is used to customize the behavior of the NezhaForMaskedLM model. Must have the property 'is_decoder' to determine if the model is a decoder. If 'is_decoder' is True, a warning will be logged regarding the bidirectional self-attention configuration.

TYPE: object

RETURNS DESCRIPTION

None.
Source code in mindnlp/transformers/models/nezha/nezha.py 
1535
1536
1537
1538
1539
1540
1541
1542
1543
1544
1545
1546
1547
1548
1549
1550
1551
1552
1553
1554
1555
1556
1557
1558
1559
1560
1561
1562
def __init__(self, config):
    """
    Initializes a new NezhaForMaskedLM instance.

    Args:
        self (object): The NezhaForMaskedLM instance itself.
        config (object): An instance of the configuration class containing the model configuration settings.
            It is used to customize the behavior of the NezhaForMaskedLM model.
            Must have the property 'is_decoder' to determine if the model is a decoder.
            If 'is_decoder' is True, a warning will be logged regarding the bidirectional
            self-attention configuration.

    Returns:
        None.

    Raises:
        None.
    """
    super().__init__(config)

    if config.is_decoder:
        logger.warning(
            "If you want to use `NezhaForMaskedLM` make sure `config.is_decoder=False` for "
            "bi-directional self-attention."
        )

    self.nezha = NezhaModel(config, add_pooling_layer=False)
    self.cls = NezhaOnlyMLMHead(config)

mindnlp.transformers.models.nezha.nezha.NezhaForMaskedLM.forward(input_ids=None, attention_mask=None, token_type_ids=None, head_mask=None, inputs_embeds=None, encoder_hidden_states=None, encoder_attention_mask=None, labels=None, output_attentions=None, output_hidden_states=None)

Constructs the Nezha model for masked language modeling (MLM).

PARAMETER DESCRIPTION
self

The instance of the NezhaForMaskedLM class.

TYPE: NezhaForMaskedLM

input_ids

The input tensor containing the tokenized input sequence IDs. Default: None.

TYPE: Tensor DEFAULT: None

attention_mask

The attention mask tensor to indicate which tokens should be attended to. Default: None.

TYPE: Tensor DEFAULT: None

token_type_ids

The token type IDs tensor to distinguish different parts of the input. Default: None.

TYPE: Tensor DEFAULT: None

head_mask

The head mask tensor to mask specific attention heads. Default: None.

TYPE: Tensor DEFAULT: None

inputs_embeds

The embedded inputs tensor. Default: None.

TYPE: Tensor DEFAULT: None

encoder_hidden_states

The hidden states of the encoder. Default: None.

TYPE: Tensor DEFAULT: None

encoder_attention_mask

The attention mask for the encoder. Default: None.

TYPE: Tensor DEFAULT: None

labels

The tensor containing the labels for the masked language modeling task. Default: None.

TYPE: Tensor DEFAULT: None

output_attentions

Whether to output attentions. Default: None.

TYPE: bool DEFAULT: None

output_hidden_states

Whether to output hidden states. Default: None.

TYPE: bool DEFAULT: None

RETURNS DESCRIPTION
tuple

A tuple containing the masked language modeling loss (if labels are provided) and the output of the model.

  • masked_lm_loss (torch.Tensor): The masked language modeling loss.
  • prediction_scores (torch.Tensor): The predicted scores for each token.
  • outputs[2:] (tuple): Additional outputs from the model. (output_attentions, output_hidden_states, ...)
Source code in mindnlp/transformers/models/nezha/nezha.py 
1572
1573
1574
1575
1576
1577
1578
1579
1580
1581
1582
1583
1584
1585
1586
1587
1588
1589
1590
1591
1592
1593
1594
1595
1596
1597
1598
1599
1600
1601
1602
1603
1604
1605
1606
1607
1608
1609
1610
1611
1612
1613
1614
1615
1616
1617
1618
1619
1620
1621
1622
1623
1624
1625
1626
1627
1628
1629
def forward(self, input_ids = None, attention_mask = None,
              token_type_ids = None, head_mask = None, inputs_embeds = None,
              encoder_hidden_states = None, encoder_attention_mask = None,
              labels = None, output_attentions = None, output_hidden_states = None):
    """
    Constructs the Nezha model for masked language modeling (MLM).

    Args:
        self (NezhaForMaskedLM): The instance of the NezhaForMaskedLM class.
        input_ids (torch.Tensor, optional): The input tensor containing the tokenized input sequence IDs.
            Default: None.
        attention_mask (torch.Tensor, optional): The attention mask tensor to indicate which tokens
            should be attended to. Default: None.
        token_type_ids (torch.Tensor, optional): The token type IDs tensor to distinguish different parts
            of the input. Default: None.
        head_mask (torch.Tensor, optional): The head mask tensor to mask specific attention heads. Default: None.
        inputs_embeds (torch.Tensor, optional): The embedded inputs tensor. Default: None.
        encoder_hidden_states (torch.Tensor, optional): The hidden states of the encoder. Default: None.
        encoder_attention_mask (torch.Tensor, optional): The attention mask for the encoder. Default: None.
        labels (torch.Tensor, optional): The tensor containing the labels for the masked language modeling task.
            Default: None.
        output_attentions (bool, optional): Whether to output attentions. Default: None.
        output_hidden_states (bool, optional): Whether to output hidden states. Default: None.

    Returns:
        tuple:
            A tuple containing the masked language modeling loss (if labels are provided)
            and the output of the model.

            - masked_lm_loss (torch.Tensor): The masked language modeling loss.
            - prediction_scores (torch.Tensor): The predicted scores for each token.
            - outputs[2:] (tuple): Additional outputs from the model. (output_attentions, output_hidden_states, ...)

    Raises:
        None.
    """
    outputs = self.nezha(
        input_ids,
        attention_mask=attention_mask,
        token_type_ids=token_type_ids,
        head_mask=head_mask,
        inputs_embeds=inputs_embeds,
        encoder_hidden_states=encoder_hidden_states,
        encoder_attention_mask=encoder_attention_mask,
        output_attentions=output_attentions,
        output_hidden_states=output_hidden_states,
    )

    sequence_output = outputs[0]
    prediction_scores = self.cls(sequence_output)

    masked_lm_loss = None
    if labels is not None:
        loss_fct = nn.CrossEntropyLoss()  # -100 index = padding token
        masked_lm_loss = loss_fct(prediction_scores.view(-1, self.config.vocab_size), labels.view(-1))

    output = (prediction_scores,) + outputs[2:]
    return ((masked_lm_loss,) + output) if masked_lm_loss is not None else output

mindnlp.transformers.models.nezha.nezha.NezhaForMaskedLM.get_output_embeddings()

get output embeddings

Source code in mindnlp/transformers/models/nezha/nezha.py 
1564
1565
1566
def get_output_embeddings(self):
    """get output embeddings"""
    return self.cls.predictions.decoder

mindnlp.transformers.models.nezha.nezha.NezhaForMaskedLM.prepare_inputs_for_generation(input_ids, attention_mask=None)

prepare inputs for generation

Source code in mindnlp/transformers/models/nezha/nezha.py 
1631
1632
1633
1634
1635
1636
1637
1638
1639
1640
1641
1642
1643
1644
1645
1646
def prepare_inputs_for_generation(self, input_ids, attention_mask=None):
    """prepare inputs for generation"""
    input_shape = input_ids.shape
    effective_batch_size = input_shape[0]

    #  add a dummy token
    if self.config.pad_token_id is None:
        raise ValueError("The PAD token should be defined for generation")

    attention_mask = ops.cat([attention_mask, attention_mask.new_zeros((attention_mask.shape[0], 1))], axis=-1)
    dummy_token = ops.full(
        (effective_batch_size, 1), self.config.pad_token_id, dtype=mindspore.int64
    )
    input_ids = ops.cat([input_ids, dummy_token], axis=1)

    return {"input_ids": input_ids, "attention_mask": attention_mask}

mindnlp.transformers.models.nezha.nezha.NezhaForMaskedLM.set_output_embeddings(new_embeddings)

set output embeddings

Source code in mindnlp/transformers/models/nezha/nezha.py 
1568
1569
1570
def set_output_embeddings(self, new_embeddings):
    """set output embeddings"""
    self.cls.predictions.decoder = new_embeddings

mindnlp.transformers.models.nezha.nezha.NezhaForMultipleChoice

Bases: NezhaPreTrainedModel

NezhaForMultipleChoice

Source code in mindnlp/transformers/models/nezha/nezha.py 
1843
1844
1845
1846
1847
1848
1849
1850
1851
1852
1853
1854
1855
1856
1857
1858
1859
1860
1861
1862
1863
1864
1865
1866
1867
1868
1869
1870
1871
1872
1873
1874
1875
1876
1877
1878
1879
1880
1881
1882
1883
1884
1885
1886
1887
1888
1889
1890
1891
1892
1893
1894
1895
1896
1897
1898
1899
1900
1901
1902
1903
1904
1905
1906
1907
1908
1909
class NezhaForMultipleChoice(NezhaPreTrainedModel):
    """NezhaForMultipleChoice"""
    def __init__(self, config):
        """
        Initialize the NezhaForMultipleChoice model with the given configuration.

        Args:
            self (NezhaForMultipleChoice): The NezhaForMultipleChoice instance.
            config (NezhaConfig): The configuration object containing various hyperparameters for the model.

        Returns:
            None.

        Raises:
            None.
        """
        super().__init__(config)

        self.nezha = NezhaModel(config)
        classifier_dropout = (
            config.classifier_dropout if config.classifier_dropout is not None else config.hidden_dropout_prob
        )
        self.dropout = nn.Dropout(p=classifier_dropout)
        self.classifier = nn.Linear(config.hidden_size, 1)

    def forward(self, input_ids = None, attention_mask = None,
                  token_type_ids = None, head_mask = None, inputs_embeds = None,
                  labels = None, output_attentions = None, output_hidden_states = None):
        r"""
        Args:
            labels (`mindspore.Tensor[int64]` of shape `(batch_size,)`, *optional*):
                Labels for computing the multiple choice classification loss. Indices should be in `[0, ...,
                num_choices-1]` where `num_choices` is the size of the second dimension of the input tensors. (See
                `input_ids` above)
        """
        num_choices = input_ids.shape[1] if input_ids is not None else inputs_embeds.shape[1]
        input_ids = input_ids.view(-1, input_ids.shape[-1]) if input_ids is not None else None
        attention_mask = attention_mask.view(-1, attention_mask.shape[-1]) if attention_mask is not None else None
        token_type_ids = token_type_ids.view(-1, token_type_ids.shape[-1]) if token_type_ids is not None else None
        inputs_embeds = (
            inputs_embeds.view(-1, inputs_embeds.shape[-2], inputs_embeds.shape[-2])
            if inputs_embeds is not None
            else None
        )

        outputs = self.nezha(
            input_ids,
            attention_mask=attention_mask,
            token_type_ids=token_type_ids,
            head_mask=head_mask,
            inputs_embeds=inputs_embeds,
            output_attentions=output_attentions,
            output_hidden_states=output_hidden_states
        )

        pooled_output = outputs[1]
        pooled_output = self.dropout(pooled_output)
        logits = self.classifier(pooled_output)
        reshaped_logits = logits.view(-1, num_choices)

        loss = None
        if labels is not None:
            loss_fct = nn.CrossEntropyLoss()
            loss = loss_fct(reshaped_logits, labels)

        output = (reshaped_logits,) + outputs[2:]
        return ((loss,) + output) if loss is not None else output

mindnlp.transformers.models.nezha.nezha.NezhaForMultipleChoice.__init__(config)

Initialize the NezhaForMultipleChoice model with the given configuration.

PARAMETER DESCRIPTION
self

The NezhaForMultipleChoice instance.

TYPE: NezhaForMultipleChoice

config

The configuration object containing various hyperparameters for the model.

TYPE: NezhaConfig

RETURNS DESCRIPTION

None.
Source code in mindnlp/transformers/models/nezha/nezha.py 
1845
1846
1847
1848
1849
1850
1851
1852
1853
1854
1855
1856
1857
1858
1859
1860
1861
1862
1863
1864
1865
1866
def __init__(self, config):
    """
    Initialize the NezhaForMultipleChoice model with the given configuration.

    Args:
        self (NezhaForMultipleChoice): The NezhaForMultipleChoice instance.
        config (NezhaConfig): The configuration object containing various hyperparameters for the model.

    Returns:
        None.

    Raises:
        None.
    """
    super().__init__(config)

    self.nezha = NezhaModel(config)
    classifier_dropout = (
        config.classifier_dropout if config.classifier_dropout is not None else config.hidden_dropout_prob
    )
    self.dropout = nn.Dropout(p=classifier_dropout)
    self.classifier = nn.Linear(config.hidden_size, 1)

mindnlp.transformers.models.nezha.nezha.NezhaForMultipleChoice.forward(input_ids=None, attention_mask=None, token_type_ids=None, head_mask=None, inputs_embeds=None, labels=None, output_attentions=None, output_hidden_states=None)

PARAMETER DESCRIPTION
labels

Labels for computing the multiple choice classification loss. Indices should be in [0, ..., num_choices-1] where num_choices is the size of the second dimension of the input tensors. (See input_ids above)

TYPE: `mindspore.Tensor[int64]` of shape `(batch_size,)`, *optional* DEFAULT: None

Source code in mindnlp/transformers/models/nezha/nezha.py 
1868
1869
1870
1871
1872
1873
1874
1875
1876
1877
1878
1879
1880
1881
1882
1883
1884
1885
1886
1887
1888
1889
1890
1891
1892
1893
1894
1895
1896
1897
1898
1899
1900
1901
1902
1903
1904
1905
1906
1907
1908
1909
def forward(self, input_ids = None, attention_mask = None,
              token_type_ids = None, head_mask = None, inputs_embeds = None,
              labels = None, output_attentions = None, output_hidden_states = None):
    r"""
    Args:
        labels (`mindspore.Tensor[int64]` of shape `(batch_size,)`, *optional*):
            Labels for computing the multiple choice classification loss. Indices should be in `[0, ...,
            num_choices-1]` where `num_choices` is the size of the second dimension of the input tensors. (See
            `input_ids` above)
    """
    num_choices = input_ids.shape[1] if input_ids is not None else inputs_embeds.shape[1]
    input_ids = input_ids.view(-1, input_ids.shape[-1]) if input_ids is not None else None
    attention_mask = attention_mask.view(-1, attention_mask.shape[-1]) if attention_mask is not None else None
    token_type_ids = token_type_ids.view(-1, token_type_ids.shape[-1]) if token_type_ids is not None else None
    inputs_embeds = (
        inputs_embeds.view(-1, inputs_embeds.shape[-2], inputs_embeds.shape[-2])
        if inputs_embeds is not None
        else None
    )

    outputs = self.nezha(
        input_ids,
        attention_mask=attention_mask,
        token_type_ids=token_type_ids,
        head_mask=head_mask,
        inputs_embeds=inputs_embeds,
        output_attentions=output_attentions,
        output_hidden_states=output_hidden_states
    )

    pooled_output = outputs[1]
    pooled_output = self.dropout(pooled_output)
    logits = self.classifier(pooled_output)
    reshaped_logits = logits.view(-1, num_choices)

    loss = None
    if labels is not None:
        loss_fct = nn.CrossEntropyLoss()
        loss = loss_fct(reshaped_logits, labels)

    output = (reshaped_logits,) + outputs[2:]
    return ((loss,) + output) if loss is not None else output

mindnlp.transformers.models.nezha.nezha.NezhaForNextSentencePrediction

Bases: NezhaPreTrainedModel

NezhaForNextSentencePrediction

Source code in mindnlp/transformers/models/nezha/nezha.py 
1649
1650
1651
1652
1653
1654
1655
1656
1657
1658
1659
1660
1661
1662
1663
1664
1665
1666
1667
1668
1669
1670
1671
1672
1673
1674
1675
1676
1677
1678
1679
1680
1681
1682
1683
1684
1685
1686
1687
1688
1689
1690
1691
1692
1693
1694
1695
1696
1697
1698
1699
1700
1701
1702
1703
1704
1705
1706
1707
1708
1709
1710
1711
1712
1713
1714
1715
1716
1717
1718
1719
1720
1721
1722
1723
1724
1725
1726
1727
1728
1729
1730
1731
1732
1733
1734
1735
1736
1737
1738
1739
1740
class NezhaForNextSentencePrediction(NezhaPreTrainedModel):
    """NezhaForNextSentencePrediction"""
    def __init__(self, config):
        """
        Initializes an instance of the NezhaForNextSentencePrediction class.

        Args:
            self: The instance of the class.
            config: An instance of the configuration class containing the model configuration settings.

        Returns:
            None

        Raises:
            None
        """
        super().__init__(config)

        self.nezha = NezhaModel(config)
        self.cls = NezhaOnlyNSPHead(config)

    def forward(self, input_ids = None, attention_mask = None,
        token_type_ids = None, head_mask = None, inputs_embeds = None,
        labels = None, output_attentions = None, output_hidden_states = None, **kwargs):
        """
        Constructs the Nezha model for next sentence prediction.

        Args:
            self (NezhaForNextSentencePrediction): The instance of the NezhaForNextSentencePrediction class.
            input_ids (torch.Tensor, optional): The input tensor of shape (batch_size, sequence_length) containing
                the input sequence indices. Defaults to None.
            attention_mask (torch.Tensor, optional): The attention mask tensor of shape (batch_size, sequence_length)
                containing the attention mask values. Defaults to None.
            token_type_ids (torch.Tensor, optional): The token type tensor of shape (batch_size, sequence_length)
                containing the token type indices. Defaults to None.
            head_mask (torch.Tensor, optional): The head mask tensor of shape
                (batch_size, num_heads, sequence_length, sequence_length) containing the head mask values.
                Defaults to None.
            inputs_embeds (torch.Tensor, optional): The embedded inputs tensor of shape
                (batch_size, sequence_length, embedding_size) containing the embedded input sequence. Defaults to None.
            labels (torch.Tensor, optional): The labels tensor of shape (batch_size) containing the next sentence labels.
                Defaults to None.
            output_attentions (bool, optional): Whether to output attention weights. Defaults to None.
            output_hidden_states (bool, optional): Whether to output hidden states. Defaults to None.

        Returns:
            tuple:
                A tuple containing the next sentence loss (if labels are provided) and the outputs of the Nezha model.

                - next_sentence_loss (torch.Tensor, optional): The loss tensor of shape (batch_size) representing
                the next sentence prediction loss. Defaults to None.
                - seq_relationship_scores (torch.Tensor): The tensor of shape (batch_size, 2) containing the next
                sentence prediction scores.
                - hidden_states (tuple, optional): A tuple of hidden states (torch.Tensor) of shape
                (batch_size, sequence_length, hidden_size) from all layers. Defaults to None.
                - attentions (tuple, optional): A tuple of attention weights (torch.Tensor) of shape
                (batch_size, num_heads, sequence_length, sequence_length) from all layers. Defaults to None.

        Raises:
            TypeError: If any of the input arguments are not of the expected type.
            ValueError: If the input tensors do not have the correct shape.

        """
        if "next_sentence_label" in kwargs:
            #warnings.warn(
            #     "The `next_sentence_label` argument is deprecated and will be removed in a future version, use"
            #     " `labels` instead.",
            #     FutureWarning,
            # )
            labels = kwargs.pop("next_sentence_label")

        outputs = self.nezha(
            input_ids,
            attention_mask=attention_mask,
            token_type_ids=token_type_ids,
            head_mask=head_mask,
            inputs_embeds=inputs_embeds,
            output_attentions=output_attentions,
            output_hidden_states=output_hidden_states
        )

        pooled_output = outputs[1]

        seq_relationship_scores = self.cls(pooled_output)

        next_sentence_loss = None
        if labels is not None:
            loss_fct = nn.CrossEntropyLoss()
            next_sentence_loss = loss_fct(seq_relationship_scores.view(-1, 2), labels.view(-1))

        output = (seq_relationship_scores,) + outputs[2:]
        return ((next_sentence_loss,) + output) if next_sentence_loss is not None else output

mindnlp.transformers.models.nezha.nezha.NezhaForNextSentencePrediction.__init__(config)

Initializes an instance of the NezhaForNextSentencePrediction class.

PARAMETER DESCRIPTION
self

The instance of the class.
config

An instance of the configuration class containing the model configuration settings.
RETURNS DESCRIPTION

None
Source code in mindnlp/transformers/models/nezha/nezha.py 
1651
1652
1653
1654
1655
1656
1657
1658
1659
1660
1661
1662
1663
1664
1665
1666
1667
1668
def __init__(self, config):
    """
    Initializes an instance of the NezhaForNextSentencePrediction class.

    Args:
        self: The instance of the class.
        config: An instance of the configuration class containing the model configuration settings.

    Returns:
        None

    Raises:
        None
    """
    super().__init__(config)

    self.nezha = NezhaModel(config)
    self.cls = NezhaOnlyNSPHead(config)

mindnlp.transformers.models.nezha.nezha.NezhaForNextSentencePrediction.forward(input_ids=None, attention_mask=None, token_type_ids=None, head_mask=None, inputs_embeds=None, labels=None, output_attentions=None, output_hidden_states=None, **kwargs)

Constructs the Nezha model for next sentence prediction.

PARAMETER DESCRIPTION
self

The instance of the NezhaForNextSentencePrediction class.

TYPE: NezhaForNextSentencePrediction

input_ids

The input tensor of shape (batch_size, sequence_length) containing the input sequence indices. Defaults to None.

TYPE: Tensor DEFAULT: None

attention_mask

The attention mask tensor of shape (batch_size, sequence_length) containing the attention mask values. Defaults to None.

TYPE: Tensor DEFAULT: None

token_type_ids

The token type tensor of shape (batch_size, sequence_length) containing the token type indices. Defaults to None.

TYPE: Tensor DEFAULT: None

head_mask

The head mask tensor of shape (batch_size, num_heads, sequence_length, sequence_length) containing the head mask values. Defaults to None.

TYPE: Tensor DEFAULT: None

inputs_embeds

The embedded inputs tensor of shape (batch_size, sequence_length, embedding_size) containing the embedded input sequence. Defaults to None.

TYPE: Tensor DEFAULT: None

labels

The labels tensor of shape (batch_size) containing the next sentence labels. Defaults to None.

TYPE: Tensor DEFAULT: None

output_attentions

Whether to output attention weights. Defaults to None.

TYPE: bool DEFAULT: None

output_hidden_states

Whether to output hidden states. Defaults to None.

TYPE: bool DEFAULT: None

RETURNS DESCRIPTION
tuple

A tuple containing the next sentence loss (if labels are provided) and the outputs of the Nezha model.

  • next_sentence_loss (torch.Tensor, optional): The loss tensor of shape (batch_size) representing the next sentence prediction loss. Defaults to None.
  • seq_relationship_scores (torch.Tensor): The tensor of shape (batch_size, 2) containing the next sentence prediction scores.
  • hidden_states (tuple, optional): A tuple of hidden states (torch.Tensor) of shape (batch_size, sequence_length, hidden_size) from all layers. Defaults to None.
  • attentions (tuple, optional): A tuple of attention weights (torch.Tensor) of shape (batch_size, num_heads, sequence_length, sequence_length) from all layers. Defaults to None.
RAISES DESCRIPTION
TypeError

If any of the input arguments are not of the expected type.
ValueError

If the input tensors do not have the correct shape.
Source code in mindnlp/transformers/models/nezha/nezha.py 
1670
1671
1672
1673
1674
1675
1676
1677
1678
1679
1680
1681
1682
1683
1684
1685
1686
1687
1688
1689
1690
1691
1692
1693
1694
1695
1696
1697
1698
1699
1700
1701
1702
1703
1704
1705
1706
1707
1708
1709
1710
1711
1712
1713
1714
1715
1716
1717
1718
1719
1720
1721
1722
1723
1724
1725
1726
1727
1728
1729
1730
1731
1732
1733
1734
1735
1736
1737
1738
1739
1740
def forward(self, input_ids = None, attention_mask = None,
    token_type_ids = None, head_mask = None, inputs_embeds = None,
    labels = None, output_attentions = None, output_hidden_states = None, **kwargs):
    """
    Constructs the Nezha model for next sentence prediction.

    Args:
        self (NezhaForNextSentencePrediction): The instance of the NezhaForNextSentencePrediction class.
        input_ids (torch.Tensor, optional): The input tensor of shape (batch_size, sequence_length) containing
            the input sequence indices. Defaults to None.
        attention_mask (torch.Tensor, optional): The attention mask tensor of shape (batch_size, sequence_length)
            containing the attention mask values. Defaults to None.
        token_type_ids (torch.Tensor, optional): The token type tensor of shape (batch_size, sequence_length)
            containing the token type indices. Defaults to None.
        head_mask (torch.Tensor, optional): The head mask tensor of shape
            (batch_size, num_heads, sequence_length, sequence_length) containing the head mask values.
            Defaults to None.
        inputs_embeds (torch.Tensor, optional): The embedded inputs tensor of shape
            (batch_size, sequence_length, embedding_size) containing the embedded input sequence. Defaults to None.
        labels (torch.Tensor, optional): The labels tensor of shape (batch_size) containing the next sentence labels.
            Defaults to None.
        output_attentions (bool, optional): Whether to output attention weights. Defaults to None.
        output_hidden_states (bool, optional): Whether to output hidden states. Defaults to None.

    Returns:
        tuple:
            A tuple containing the next sentence loss (if labels are provided) and the outputs of the Nezha model.

            - next_sentence_loss (torch.Tensor, optional): The loss tensor of shape (batch_size) representing
            the next sentence prediction loss. Defaults to None.
            - seq_relationship_scores (torch.Tensor): The tensor of shape (batch_size, 2) containing the next
            sentence prediction scores.
            - hidden_states (tuple, optional): A tuple of hidden states (torch.Tensor) of shape
            (batch_size, sequence_length, hidden_size) from all layers. Defaults to None.
            - attentions (tuple, optional): A tuple of attention weights (torch.Tensor) of shape
            (batch_size, num_heads, sequence_length, sequence_length) from all layers. Defaults to None.

    Raises:
        TypeError: If any of the input arguments are not of the expected type.
        ValueError: If the input tensors do not have the correct shape.

    """
    if "next_sentence_label" in kwargs:
        #warnings.warn(
        #     "The `next_sentence_label` argument is deprecated and will be removed in a future version, use"
        #     " `labels` instead.",
        #     FutureWarning,
        # )
        labels = kwargs.pop("next_sentence_label")

    outputs = self.nezha(
        input_ids,
        attention_mask=attention_mask,
        token_type_ids=token_type_ids,
        head_mask=head_mask,
        inputs_embeds=inputs_embeds,
        output_attentions=output_attentions,
        output_hidden_states=output_hidden_states
    )

    pooled_output = outputs[1]

    seq_relationship_scores = self.cls(pooled_output)

    next_sentence_loss = None
    if labels is not None:
        loss_fct = nn.CrossEntropyLoss()
        next_sentence_loss = loss_fct(seq_relationship_scores.view(-1, 2), labels.view(-1))

    output = (seq_relationship_scores,) + outputs[2:]
    return ((next_sentence_loss,) + output) if next_sentence_loss is not None else output

mindnlp.transformers.models.nezha.nezha.NezhaForPreTraining

Bases: NezhaPreTrainedModel

NezhaForPreTraining

Source code in mindnlp/transformers/models/nezha/nezha.py 
1443
1444
1445
1446
1447
1448
1449
1450
1451
1452
1453
1454
1455
1456
1457
1458
1459
1460
1461
1462
1463
1464
1465
1466
1467
1468
1469
1470
1471
1472
1473
1474
1475
1476
1477
1478
1479
1480
1481
1482
1483
1484
1485
1486
1487
1488
1489
1490
1491
1492
1493
1494
1495
1496
1497
1498
1499
1500
1501
1502
1503
1504
1505
1506
1507
1508
1509
1510
1511
1512
1513
1514
1515
1516
1517
1518
1519
1520
1521
1522
1523
1524
1525
1526
1527
class NezhaForPreTraining(NezhaPreTrainedModel):
    """NezhaForPreTraining"""
    _keys_to_ignore_on_load_missing = ["cls.predictions.decoder"]
    def __init__(self, config):
        """
        Initializes an instance of the 'NezhaForPreTraining' class.

        Args:
            self: The instance of the class.
            config:
                A configuration object containing various settings for the model.

                - Type: Config object
                - Purpose: Specifies the model configuration.
                - Restrictions: None

        Returns:
            None.

        Raises:
            None.
        """
        super().__init__(config)

        self.nezha = NezhaModel(config)
        self.cls = NezhaPreTrainingHeads(config)

        # Initialize weights and apply final processing
        # self.post_init()

    def get_output_embeddings(self):
        """get output embeddings"""
        return self.cls.predictions.decoder

    def set_output_embeddings(self, new_embeddings):
        """set output embeddings"""
        self.cls.predictions.decoder = new_embeddings

    def forward(self, input_ids = None, attention_mask = None,
                  token_type_ids = None, head_mask = None, inputs_embeds = None,
                  labels = None, next_sentence_label = None,
                  output_attentions = None, output_hidden_states = None):
        """
        Constructs the Nezha model for pre-training.

        Args:
            self (NezhaForPreTraining): The instance of the NezhaForPreTraining class.
            input_ids (torch.Tensor, optional): The input sequence tensor. Default: None.
            attention_mask (torch.Tensor, optional): The attention mask tensor. Default: None.
            token_type_ids (torch.Tensor, optional): The token type ids tensor. Default: None.
            head_mask (torch.Tensor, optional): The head mask tensor. Default: None.
            inputs_embeds (torch.Tensor, optional): The input embeddings tensor. Default: None.
            labels (torch.Tensor, optional): The labels tensor. Default: None.
            next_sentence_label (torch.Tensor, optional): The next sentence label tensor. Default: None.
            output_attentions (bool, optional): Whether to output attentions. Default: None.
            output_hidden_states (bool, optional): Whether to output hidden states. Default: None.

        Returns:
            tuple or torch.Tensor: A tuple of output tensors or a single tensor representing the total loss.

        Raises:
            None
        """
        outputs = self.nezha(
            input_ids,
            attention_mask=attention_mask,
            token_type_ids=token_type_ids,
            head_mask=head_mask,
            inputs_embeds=inputs_embeds,
            output_attentions=output_attentions,
            output_hidden_states=output_hidden_states,
        )

        sequence_output, pooled_output = outputs[:2]
        prediction_scores, seq_relationship_score = self.cls(sequence_output, pooled_output)

        total_loss = None
        if labels is not None and next_sentence_label is not None:
            loss_fct = nn.CrossEntropyLoss()
            masked_lm_loss = loss_fct(prediction_scores.view(-1, self.config.vocab_size), labels.view(-1))
            next_sentence_loss = loss_fct(seq_relationship_score.view(-1, 2), next_sentence_label.view(-1))
            total_loss = masked_lm_loss + next_sentence_loss

        output = (prediction_scores, seq_relationship_score) + outputs[2:]
        return ((total_loss,) + output) if total_loss is not None else output

mindnlp.transformers.models.nezha.nezha.NezhaForPreTraining.__init__(config)

Initializes an instance of the 'NezhaForPreTraining' class.

PARAMETER DESCRIPTION
self

The instance of the class.
config

A configuration object containing various settings for the model.

  • Type: Config object
  • Purpose: Specifies the model configuration.
  • Restrictions: None
RETURNS DESCRIPTION

None.
Source code in mindnlp/transformers/models/nezha/nezha.py 
1446
1447
1448
1449
1450
1451
1452
1453
1454
1455
1456
1457
1458
1459
1460
1461
1462
1463
1464
1465
1466
1467
1468
def __init__(self, config):
    """
    Initializes an instance of the 'NezhaForPreTraining' class.

    Args:
        self: The instance of the class.
        config:
            A configuration object containing various settings for the model.

            - Type: Config object
            - Purpose: Specifies the model configuration.
            - Restrictions: None

    Returns:
        None.

    Raises:
        None.
    """
    super().__init__(config)

    self.nezha = NezhaModel(config)
    self.cls = NezhaPreTrainingHeads(config)

mindnlp.transformers.models.nezha.nezha.NezhaForPreTraining.forward(input_ids=None, attention_mask=None, token_type_ids=None, head_mask=None, inputs_embeds=None, labels=None, next_sentence_label=None, output_attentions=None, output_hidden_states=None)

Constructs the Nezha model for pre-training.

PARAMETER DESCRIPTION
self

The instance of the NezhaForPreTraining class.

TYPE: NezhaForPreTraining

input_ids

The input sequence tensor. Default: None.

TYPE: Tensor DEFAULT: None

attention_mask

The attention mask tensor. Default: None.

TYPE: Tensor DEFAULT: None

token_type_ids

The token type ids tensor. Default: None.

TYPE: Tensor DEFAULT: None

head_mask

The head mask tensor. Default: None.

TYPE: Tensor DEFAULT: None

inputs_embeds

The input embeddings tensor. Default: None.

TYPE: Tensor DEFAULT: None

labels

The labels tensor. Default: None.

TYPE: Tensor DEFAULT: None

next_sentence_label

The next sentence label tensor. Default: None.

TYPE: Tensor DEFAULT: None

output_attentions

Whether to output attentions. Default: None.

TYPE: bool DEFAULT: None

output_hidden_states

Whether to output hidden states. Default: None.

TYPE: bool DEFAULT: None

RETURNS DESCRIPTION

tuple or torch.Tensor: A tuple of output tensors or a single tensor representing the total loss.
Source code in mindnlp/transformers/models/nezha/nezha.py 
1481
1482
1483
1484
1485
1486
1487
1488
1489
1490
1491
1492
1493
1494
1495
1496
1497
1498
1499
1500
1501
1502
1503
1504
1505
1506
1507
1508
1509
1510
1511
1512
1513
1514
1515
1516
1517
1518
1519
1520
1521
1522
1523
1524
1525
1526
1527
def forward(self, input_ids = None, attention_mask = None,
              token_type_ids = None, head_mask = None, inputs_embeds = None,
              labels = None, next_sentence_label = None,
              output_attentions = None, output_hidden_states = None):
    """
    Constructs the Nezha model for pre-training.

    Args:
        self (NezhaForPreTraining): The instance of the NezhaForPreTraining class.
        input_ids (torch.Tensor, optional): The input sequence tensor. Default: None.
        attention_mask (torch.Tensor, optional): The attention mask tensor. Default: None.
        token_type_ids (torch.Tensor, optional): The token type ids tensor. Default: None.
        head_mask (torch.Tensor, optional): The head mask tensor. Default: None.
        inputs_embeds (torch.Tensor, optional): The input embeddings tensor. Default: None.
        labels (torch.Tensor, optional): The labels tensor. Default: None.
        next_sentence_label (torch.Tensor, optional): The next sentence label tensor. Default: None.
        output_attentions (bool, optional): Whether to output attentions. Default: None.
        output_hidden_states (bool, optional): Whether to output hidden states. Default: None.

    Returns:
        tuple or torch.Tensor: A tuple of output tensors or a single tensor representing the total loss.

    Raises:
        None
    """
    outputs = self.nezha(
        input_ids,
        attention_mask=attention_mask,
        token_type_ids=token_type_ids,
        head_mask=head_mask,
        inputs_embeds=inputs_embeds,
        output_attentions=output_attentions,
        output_hidden_states=output_hidden_states,
    )

    sequence_output, pooled_output = outputs[:2]
    prediction_scores, seq_relationship_score = self.cls(sequence_output, pooled_output)

    total_loss = None
    if labels is not None and next_sentence_label is not None:
        loss_fct = nn.CrossEntropyLoss()
        masked_lm_loss = loss_fct(prediction_scores.view(-1, self.config.vocab_size), labels.view(-1))
        next_sentence_loss = loss_fct(seq_relationship_score.view(-1, 2), next_sentence_label.view(-1))
        total_loss = masked_lm_loss + next_sentence_loss

    output = (prediction_scores, seq_relationship_score) + outputs[2:]
    return ((total_loss,) + output) if total_loss is not None else output

mindnlp.transformers.models.nezha.nezha.NezhaForPreTraining.get_output_embeddings()

get output embeddings

Source code in mindnlp/transformers/models/nezha/nezha.py 
1473
1474
1475
def get_output_embeddings(self):
    """get output embeddings"""
    return self.cls.predictions.decoder

mindnlp.transformers.models.nezha.nezha.NezhaForPreTraining.set_output_embeddings(new_embeddings)

set output embeddings

Source code in mindnlp/transformers/models/nezha/nezha.py 
1477
1478
1479
def set_output_embeddings(self, new_embeddings):
    """set output embeddings"""
    self.cls.predictions.decoder = new_embeddings

mindnlp.transformers.models.nezha.nezha.NezhaForQuestionAnswering

Bases: NezhaPreTrainedModel

NezhaForQuestionAnswering

Source code in mindnlp/transformers/models/nezha/nezha.py 
1977
1978
1979
1980
1981
1982
1983
1984
1985
1986
1987
1988
1989
1990
1991
1992
1993
1994
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
2006
2007
2008
2009
2010
2011
2012
2013
2014
2015
2016
2017
2018
2019
2020
2021
2022
2023
2024
2025
2026
2027
2028
2029
2030
2031
2032
2033
2034
2035
2036
2037
2038
2039
2040
2041
2042
2043
2044
2045
2046
2047
2048
2049
2050
2051
2052
2053
class NezhaForQuestionAnswering(NezhaPreTrainedModel):
    """NezhaForQuestionAnswering"""
    _keys_to_ignore_on_load_unexpected = [r"pooler"]

    def __init__(self, config):
        """
        This method initializes an instance of the NezhaForQuestionAnswering class.

        Args:
            self (NezhaForQuestionAnswering): The instance of the NezhaForQuestionAnswering class.
            config: An instance of the NezhaConfig class containing the model configuration.

        Returns:
            None.

        Raises:
            TypeError: If the config parameter is not of type NezhaConfig.
            ValueError: If the config.num_labels is not defined or is not a positive integer.
            RuntimeError: If an error occurs during the initialization process.
        """
        super().__init__(config)
        self.num_labels = config.num_labels

        self.nezha = NezhaModel(config, add_pooling_layer=False)
        self.qa_outputs = nn.Linear(config.hidden_size, config.num_labels)

    def forward(self, input_ids = None, attention_mask = None,
                  token_type_ids = None, head_mask = None, inputs_embeds = None,
                  start_positions = None, end_positions = None, output_attentions = None,
                  output_hidden_states = None):
        r"""
        Args:
            start_positions (`mindspore.Tensor[int64]` of shape `(batch_size,)`, *optional*):
                Labels for position (index) of the start of the labelled span for computing the token classification loss.
                Positions are clamped to the length of the sequence (`sequence_length`). Position outside of the sequence
                are not taken into account for computing the loss.
            end_positions (`mindspore.Tensor[int64]` of shape `(batch_size,)`, *optional*):
                Labels for position (index) of the end of the labelled span for computing the token classification loss.
                Positions are clamped to the length of the sequence (`sequence_length`). Position outside of the sequence
                are not taken into account for computing the loss.
        """
        outputs = self.nezha(
            input_ids,
            attention_mask=attention_mask,
            token_type_ids=token_type_ids,
            head_mask=head_mask,
            inputs_embeds=inputs_embeds,
            output_attentions=output_attentions,
            output_hidden_states=output_hidden_states
        )

        sequence_output = outputs[0]

        logits = self.qa_outputs(sequence_output)
        start_logits, end_logits = logits.split(1, axis=-1)
        start_logits = start_logits.squeeze(-1)
        end_logits = end_logits.squeeze(-1)

        total_loss = None
        if start_positions is not None and end_positions is not None:
            # If we are on multi-GPU, split add a dimension
            if len(start_positions.shape) > 1:
                start_positions = start_positions.squeeze(-1)
            if len(end_positions.shape) > 1:
                end_positions = end_positions.squeeze(-1)
            # sometimes the start/end positions are outside our model inputs, we ignore these terms
            ignored_index = start_logits.shape[1]
            start_positions = start_positions.clamp(0, ignored_index)
            end_positions = end_positions.clamp(0, ignored_index)

            loss_fct = nn.CrossEntropyLoss(ignore_index=ignored_index)
            start_loss = loss_fct(start_logits, start_positions)
            end_loss = loss_fct(end_logits, end_positions)
            total_loss = (start_loss + end_loss) / 2

        output = (start_logits, end_logits) + outputs[2:]
        return ((total_loss,) + output) if total_loss is not None else output

mindnlp.transformers.models.nezha.nezha.NezhaForQuestionAnswering.__init__(config)

This method initializes an instance of the NezhaForQuestionAnswering class.

PARAMETER DESCRIPTION
self

The instance of the NezhaForQuestionAnswering class.

TYPE: NezhaForQuestionAnswering

config

An instance of the NezhaConfig class containing the model configuration.
RETURNS DESCRIPTION

None.
RAISES DESCRIPTION
TypeError

If the config parameter is not of type NezhaConfig.
ValueError

If the config.num_labels is not defined or is not a positive integer.
RuntimeError

If an error occurs during the initialization process.
Source code in mindnlp/transformers/models/nezha/nezha.py 
1981
1982
1983
1984
1985
1986
1987
1988
1989
1990
1991
1992
1993
1994
1995
1996
1997
1998
1999
2000
2001
def __init__(self, config):
    """
    This method initializes an instance of the NezhaForQuestionAnswering class.

    Args:
        self (NezhaForQuestionAnswering): The instance of the NezhaForQuestionAnswering class.
        config: An instance of the NezhaConfig class containing the model configuration.

    Returns:
        None.

    Raises:
        TypeError: If the config parameter is not of type NezhaConfig.
        ValueError: If the config.num_labels is not defined or is not a positive integer.
        RuntimeError: If an error occurs during the initialization process.
    """
    super().__init__(config)
    self.num_labels = config.num_labels

    self.nezha = NezhaModel(config, add_pooling_layer=False)
    self.qa_outputs = nn.Linear(config.hidden_size, config.num_labels)

mindnlp.transformers.models.nezha.nezha.NezhaForQuestionAnswering.forward(input_ids=None, attention_mask=None, token_type_ids=None, head_mask=None, inputs_embeds=None, start_positions=None, end_positions=None, output_attentions=None, output_hidden_states=None)

PARAMETER DESCRIPTION
start_positions

Labels for position (index) of the start of the labelled span for computing the token classification loss. Positions are clamped to the length of the sequence (sequence_length). Position outside of the sequence are not taken into account for computing the loss.

TYPE: `mindspore.Tensor[int64]` of shape `(batch_size,)`, *optional* DEFAULT: None

end_positions

Labels for position (index) of the end of the labelled span for computing the token classification loss. Positions are clamped to the length of the sequence (sequence_length). Position outside of the sequence are not taken into account for computing the loss.

TYPE: `mindspore.Tensor[int64]` of shape `(batch_size,)`, *optional* DEFAULT: None

Source code in mindnlp/transformers/models/nezha/nezha.py 
2003
2004
2005
2006
2007
2008
2009
2010
2011
2012
2013
2014
2015
2016
2017
2018
2019
2020
2021
2022
2023
2024
2025
2026
2027
2028
2029
2030
2031
2032
2033
2034
2035
2036
2037
2038
2039
2040
2041
2042
2043
2044
2045
2046
2047
2048
2049
2050
2051
2052
2053
def forward(self, input_ids = None, attention_mask = None,
              token_type_ids = None, head_mask = None, inputs_embeds = None,
              start_positions = None, end_positions = None, output_attentions = None,
              output_hidden_states = None):
    r"""
    Args:
        start_positions (`mindspore.Tensor[int64]` of shape `(batch_size,)`, *optional*):
            Labels for position (index) of the start of the labelled span for computing the token classification loss.
            Positions are clamped to the length of the sequence (`sequence_length`). Position outside of the sequence
            are not taken into account for computing the loss.
        end_positions (`mindspore.Tensor[int64]` of shape `(batch_size,)`, *optional*):
            Labels for position (index) of the end of the labelled span for computing the token classification loss.
            Positions are clamped to the length of the sequence (`sequence_length`). Position outside of the sequence
            are not taken into account for computing the loss.
    """
    outputs = self.nezha(
        input_ids,
        attention_mask=attention_mask,
        token_type_ids=token_type_ids,
        head_mask=head_mask,
        inputs_embeds=inputs_embeds,
        output_attentions=output_attentions,
        output_hidden_states=output_hidden_states
    )

    sequence_output = outputs[0]

    logits = self.qa_outputs(sequence_output)
    start_logits, end_logits = logits.split(1, axis=-1)
    start_logits = start_logits.squeeze(-1)
    end_logits = end_logits.squeeze(-1)

    total_loss = None
    if start_positions is not None and end_positions is not None:
        # If we are on multi-GPU, split add a dimension
        if len(start_positions.shape) > 1:
            start_positions = start_positions.squeeze(-1)
        if len(end_positions.shape) > 1:
            end_positions = end_positions.squeeze(-1)
        # sometimes the start/end positions are outside our model inputs, we ignore these terms
        ignored_index = start_logits.shape[1]
        start_positions = start_positions.clamp(0, ignored_index)
        end_positions = end_positions.clamp(0, ignored_index)

        loss_fct = nn.CrossEntropyLoss(ignore_index=ignored_index)
        start_loss = loss_fct(start_logits, start_positions)
        end_loss = loss_fct(end_logits, end_positions)
        total_loss = (start_loss + end_loss) / 2

    output = (start_logits, end_logits) + outputs[2:]
    return ((total_loss,) + output) if total_loss is not None else output

mindnlp.transformers.models.nezha.nezha.NezhaForSequenceClassification

Bases: NezhaPreTrainedModel

NezhaForSequenceClassification

Source code in mindnlp/transformers/models/nezha/nezha.py 
1743
1744
1745
1746
1747
1748
1749
1750
1751
1752
1753
1754
1755
1756
1757
1758
1759
1760
1761
1762
1763
1764
1765
1766
1767
1768
1769
1770
1771
1772
1773
1774
1775
1776
1777
1778
1779
1780
1781
1782
1783
1784
1785
1786
1787
1788
1789
1790
1791
1792
1793
1794
1795
1796
1797
1798
1799
1800
1801
1802
1803
1804
1805
1806
1807
1808
1809
1810
1811
1812
1813
1814
1815
1816
1817
1818
1819
1820
1821
1822
1823
1824
1825
1826
1827
1828
1829
1830
1831
1832
1833
1834
1835
1836
1837
1838
1839
1840
class NezhaForSequenceClassification(NezhaPreTrainedModel):
    """NezhaForSequenceClassification"""
    def __init__(self, config):
        """
        Initializes a new instance of the NezhaForSequenceClassification class.

        Args:
            self (NezhaForSequenceClassification): An instance of the NezhaForSequenceClassification class.
            config (NezhaConfig): The configuration class instance specifying the model's hyperparameters.

        Returns:
            None

        Raises:
            None
        """
        super().__init__(config)
        self.num_labels = config.num_labels
        self.config = config

        self.nezha = NezhaModel(config)
        classifier_dropout = (
            config.classifier_dropout if config.classifier_dropout is not None else config.hidden_dropout_prob
        )
        self.dropout = nn.Dropout(p=classifier_dropout)
        self.classifier = nn.Linear(config.hidden_size, config.num_labels)

    def forward(self, input_ids = None, attention_mask = None,
                  token_type_ids = None, head_mask = None, inputs_embeds = None,
                  labels = None, output_attentions = None, output_hidden_states = None):
        '''
        This method forwards a Nezha model for sequence classification.

        Args:
            self (object): The instance of the NezhaForSequenceClassification class.
            input_ids (list, optional): A list of tokenized input sequence IDs. Defaults to None.
            attention_mask (list, optional): A list of attention masks indicating which tokens should be attended to.
                Defaults to None.
            token_type_ids (list, optional): A list of token type IDs to indicate which parts of the input belong to
                the first sequence and which belong to the second sequence. Defaults to None.
            head_mask (list, optional): A list of masks for attention heads. Defaults to None.
            inputs_embeds (list, optional): A list of input embeddings. Defaults to None.
            labels (list, optional): A list of target labels for the input sequence. Defaults to None.
            output_attentions (bool, optional): A boolean flag indicating whether to return the attentions tensor.
                Defaults to None.
            output_hidden_states (bool, optional): A boolean flag indicating whether to return the hidden states tensor.
                Defaults to None.

        Returns:
            tuple: A tuple containing the loss value and the output logits.
                If no loss is calculated, only the logits are returned.

        Raises:
            ValueError: If the problem type is not recognized.
            RuntimeError: If the number of labels is not compatible with the specified problem type.
            TypeError: If the labels data type is not supported for the specified problem type.
            AssertionError: If the loss function encounters an unexpected condition.
        '''
        outputs = self.nezha(
            input_ids,
            attention_mask=attention_mask,
            token_type_ids=token_type_ids,
            head_mask=head_mask,
            inputs_embeds=inputs_embeds,
            output_attentions=output_attentions,
            output_hidden_states=output_hidden_states
        )

        pooled_output = outputs[1]

        pooled_output = self.dropout(pooled_output)
        logits = self.classifier(pooled_output)

        loss = None
        if labels is not None:
            if self.config.problem_type is None:
                if self.num_labels == 1:
                    self.config.problem_type = "regression"
                elif self.num_labels > 1 and labels.dtype in (mindspore.int64, mindspore.int32):
                    self.config.problem_type = "single_label_classification"
                else:
                    self.config.problem_type = "multi_label_classification"

            if self.config.problem_type == "regression":
                loss_fct = nn.MSELoss()
                if self.num_labels == 1:
                    loss = loss_fct(logits.squeeze(), labels.squeeze())
                else:
                    loss = loss_fct(logits, labels)
            elif self.config.problem_type == "single_label_classification":
                loss_fct = nn.CrossEntropyLoss()
                loss = loss_fct(logits.view(-1, self.num_labels), labels.view(-1))
            elif self.config.problem_type == "multi_label_classification":
                loss_fct = nn.BCEWithLogitsLoss()
                loss = loss_fct(logits, labels)

        output = (logits,) + outputs[2:]
        return ((loss,) + output) if loss is not None else output

mindnlp.transformers.models.nezha.nezha.NezhaForSequenceClassification.__init__(config)

Initializes a new instance of the NezhaForSequenceClassification class.

PARAMETER DESCRIPTION
self

An instance of the NezhaForSequenceClassification class.

TYPE: NezhaForSequenceClassification

config

The configuration class instance specifying the model's hyperparameters.

TYPE: NezhaConfig

RETURNS DESCRIPTION

None
Source code in mindnlp/transformers/models/nezha/nezha.py 
1745
1746
1747
1748
1749
1750
1751
1752
1753
1754
1755
1756
1757
1758
1759
1760
1761
1762
1763
1764
1765
1766
1767
1768
def __init__(self, config):
    """
    Initializes a new instance of the NezhaForSequenceClassification class.

    Args:
        self (NezhaForSequenceClassification): An instance of the NezhaForSequenceClassification class.
        config (NezhaConfig): The configuration class instance specifying the model's hyperparameters.

    Returns:
        None

    Raises:
        None
    """
    super().__init__(config)
    self.num_labels = config.num_labels
    self.config = config

    self.nezha = NezhaModel(config)
    classifier_dropout = (
        config.classifier_dropout if config.classifier_dropout is not None else config.hidden_dropout_prob
    )
    self.dropout = nn.Dropout(p=classifier_dropout)
    self.classifier = nn.Linear(config.hidden_size, config.num_labels)

mindnlp.transformers.models.nezha.nezha.NezhaForSequenceClassification.forward(input_ids=None, attention_mask=None, token_type_ids=None, head_mask=None, inputs_embeds=None, labels=None, output_attentions=None, output_hidden_states=None)

This method forwards a Nezha model for sequence classification.

PARAMETER DESCRIPTION
self

The instance of the NezhaForSequenceClassification class.

TYPE: object

input_ids

A list of tokenized input sequence IDs. Defaults to None.

TYPE: list DEFAULT: None

attention_mask

A list of attention masks indicating which tokens should be attended to. Defaults to None.

TYPE: list DEFAULT: None

token_type_ids

A list of token type IDs to indicate which parts of the input belong to the first sequence and which belong to the second sequence. Defaults to None.

TYPE: list DEFAULT: None

head_mask

A list of masks for attention heads. Defaults to None.

TYPE: list DEFAULT: None

inputs_embeds

A list of input embeddings. Defaults to None.

TYPE: list DEFAULT: None

labels

A list of target labels for the input sequence. Defaults to None.

TYPE: list DEFAULT: None

output_attentions

A boolean flag indicating whether to return the attentions tensor. Defaults to None.

TYPE: bool DEFAULT: None

output_hidden_states

A boolean flag indicating whether to return the hidden states tensor. Defaults to None.

TYPE: bool DEFAULT: None

RETURNS DESCRIPTION
tuple

A tuple containing the loss value and the output logits. If no loss is calculated, only the logits are returned.
RAISES DESCRIPTION
ValueError

If the problem type is not recognized.
RuntimeError

If the number of labels is not compatible with the specified problem type.
TypeError

If the labels data type is not supported for the specified problem type.
AssertionError

If the loss function encounters an unexpected condition.
Source code in mindnlp/transformers/models/nezha/nezha.py 
1770
1771
1772
1773
1774
1775
1776
1777
1778
1779
1780
1781
1782
1783
1784
1785
1786
1787
1788
1789
1790
1791
1792
1793
1794
1795
1796
1797
1798
1799
1800
1801
1802
1803
1804
1805
1806
1807
1808
1809
1810
1811
1812
1813
1814
1815
1816
1817
1818
1819
1820
1821
1822
1823
1824
1825
1826
1827
1828
1829
1830
1831
1832
1833
1834
1835
1836
1837
1838
1839
1840
def forward(self, input_ids = None, attention_mask = None,
              token_type_ids = None, head_mask = None, inputs_embeds = None,
              labels = None, output_attentions = None, output_hidden_states = None):
    '''
    This method forwards a Nezha model for sequence classification.

    Args:
        self (object): The instance of the NezhaForSequenceClassification class.
        input_ids (list, optional): A list of tokenized input sequence IDs. Defaults to None.
        attention_mask (list, optional): A list of attention masks indicating which tokens should be attended to.
            Defaults to None.
        token_type_ids (list, optional): A list of token type IDs to indicate which parts of the input belong to
            the first sequence and which belong to the second sequence. Defaults to None.
        head_mask (list, optional): A list of masks for attention heads. Defaults to None.
        inputs_embeds (list, optional): A list of input embeddings. Defaults to None.
        labels (list, optional): A list of target labels for the input sequence. Defaults to None.
        output_attentions (bool, optional): A boolean flag indicating whether to return the attentions tensor.
            Defaults to None.
        output_hidden_states (bool, optional): A boolean flag indicating whether to return the hidden states tensor.
            Defaults to None.

    Returns:
        tuple: A tuple containing the loss value and the output logits.
            If no loss is calculated, only the logits are returned.

    Raises:
        ValueError: If the problem type is not recognized.
        RuntimeError: If the number of labels is not compatible with the specified problem type.
        TypeError: If the labels data type is not supported for the specified problem type.
        AssertionError: If the loss function encounters an unexpected condition.
    '''
    outputs = self.nezha(
        input_ids,
        attention_mask=attention_mask,
        token_type_ids=token_type_ids,
        head_mask=head_mask,
        inputs_embeds=inputs_embeds,
        output_attentions=output_attentions,
        output_hidden_states=output_hidden_states
    )

    pooled_output = outputs[1]

    pooled_output = self.dropout(pooled_output)
    logits = self.classifier(pooled_output)

    loss = None
    if labels is not None:
        if self.config.problem_type is None:
            if self.num_labels == 1:
                self.config.problem_type = "regression"
            elif self.num_labels > 1 and labels.dtype in (mindspore.int64, mindspore.int32):
                self.config.problem_type = "single_label_classification"
            else:
                self.config.problem_type = "multi_label_classification"

        if self.config.problem_type == "regression":
            loss_fct = nn.MSELoss()
            if self.num_labels == 1:
                loss = loss_fct(logits.squeeze(), labels.squeeze())
            else:
                loss = loss_fct(logits, labels)
        elif self.config.problem_type == "single_label_classification":
            loss_fct = nn.CrossEntropyLoss()
            loss = loss_fct(logits.view(-1, self.num_labels), labels.view(-1))
        elif self.config.problem_type == "multi_label_classification":
            loss_fct = nn.BCEWithLogitsLoss()
            loss = loss_fct(logits, labels)

    output = (logits,) + outputs[2:]
    return ((loss,) + output) if loss is not None else output

mindnlp.transformers.models.nezha.nezha.NezhaForTokenClassification

Bases: NezhaPreTrainedModel

NezhaForTokenClassification

Source code in mindnlp/transformers/models/nezha/nezha.py 
1912
1913
1914
1915
1916
1917
1918
1919
1920
1921
1922
1923
1924
1925
1926
1927
1928
1929
1930
1931
1932
1933
1934
1935
1936
1937
1938
1939
1940
1941
1942
1943
1944
1945
1946
1947
1948
1949
1950
1951
1952
1953
1954
1955
1956
1957
1958
1959
1960
1961
1962
1963
1964
1965
1966
1967
1968
1969
1970
1971
1972
1973
1974
class NezhaForTokenClassification(NezhaPreTrainedModel):
    """NezhaForTokenClassification"""
    _keys_to_ignore_on_load_unexpected = [r"pooler"]

    def __init__(self, config):
        """
        Initializes a new instance of the NezhaForTokenClassification class.

        Args:
            self: The object itself.
            config:
                An instance of the NezhaConfig class containing the model configuration settings.

                - Type: NezhaConfig
                - Purpose: Specifies the configuration for the Nezha model.
                - Restrictions: None

        Returns:
            None

        Raises:
            None
        """
        super().__init__(config)
        self.num_labels = config.num_labels

        self.nezha = NezhaModel(config, add_pooling_layer=False)
        classifier_dropout = (
            config.classifier_dropout if config.classifier_dropout is not None else config.hidden_dropout_prob
        )
        self.dropout = nn.Dropout(p=classifier_dropout)
        self.classifier = nn.Linear(config.hidden_size, config.num_labels)

    def forward(self, input_ids = None, attention_mask = None,
                  token_type_ids = None, head_mask = None, inputs_embeds = None,
                  labels = None, output_attentions = None, output_hidden_states = None):
        r"""
        Args:
            labels (`mindspore.Tensor[int64]` of shape `(batch_size, sequence_length)`, *optional*):
                Labels for computing the token classification loss. Indices should be in `[0, ..., config.num_labels - 1]`.
        """
        outputs = self.nezha(
            input_ids,
            attention_mask=attention_mask,
            token_type_ids=token_type_ids,
            head_mask=head_mask,
            inputs_embeds=inputs_embeds,
            output_attentions=output_attentions,
            output_hidden_states=output_hidden_states
        )

        sequence_output = outputs[0]

        sequence_output = self.dropout(sequence_output)
        logits = self.classifier(sequence_output)

        loss = None
        if labels is not None:
            loss_fct = nn.CrossEntropyLoss()
            loss = loss_fct(logits.view(-1, self.num_labels), labels.view(-1))

        output = (logits,) + outputs[2:]
        return ((loss,) + output) if loss is not None else output

mindnlp.transformers.models.nezha.nezha.NezhaForTokenClassification.__init__(config)

Initializes a new instance of the NezhaForTokenClassification class.

PARAMETER DESCRIPTION
self

The object itself.
config

An instance of the NezhaConfig class containing the model configuration settings.

  • Type: NezhaConfig
  • Purpose: Specifies the configuration for the Nezha model.
  • Restrictions: None
RETURNS DESCRIPTION

None
Source code in mindnlp/transformers/models/nezha/nezha.py 
1916
1917
1918
1919
1920
1921
1922
1923
1924
1925
1926
1927
1928
1929
1930
1931
1932
1933
1934
1935
1936
1937
1938
1939
1940
1941
1942
1943
def __init__(self, config):
    """
    Initializes a new instance of the NezhaForTokenClassification class.

    Args:
        self: The object itself.
        config:
            An instance of the NezhaConfig class containing the model configuration settings.

            - Type: NezhaConfig
            - Purpose: Specifies the configuration for the Nezha model.
            - Restrictions: None

    Returns:
        None

    Raises:
        None
    """
    super().__init__(config)
    self.num_labels = config.num_labels

    self.nezha = NezhaModel(config, add_pooling_layer=False)
    classifier_dropout = (
        config.classifier_dropout if config.classifier_dropout is not None else config.hidden_dropout_prob
    )
    self.dropout = nn.Dropout(p=classifier_dropout)
    self.classifier = nn.Linear(config.hidden_size, config.num_labels)

mindnlp.transformers.models.nezha.nezha.NezhaForTokenClassification.forward(input_ids=None, attention_mask=None, token_type_ids=None, head_mask=None, inputs_embeds=None, labels=None, output_attentions=None, output_hidden_states=None)

PARAMETER DESCRIPTION
labels

Labels for computing the token classification loss. Indices should be in [0, ..., config.num_labels - 1].

TYPE: `mindspore.Tensor[int64]` of shape `(batch_size, sequence_length)`, *optional* DEFAULT: None

Source code in mindnlp/transformers/models/nezha/nezha.py 
1945
1946
1947
1948
1949
1950
1951
1952
1953
1954
1955
1956
1957
1958
1959
1960
1961
1962
1963
1964
1965
1966
1967
1968
1969
1970
1971
1972
1973
1974
def forward(self, input_ids = None, attention_mask = None,
              token_type_ids = None, head_mask = None, inputs_embeds = None,
              labels = None, output_attentions = None, output_hidden_states = None):
    r"""
    Args:
        labels (`mindspore.Tensor[int64]` of shape `(batch_size, sequence_length)`, *optional*):
            Labels for computing the token classification loss. Indices should be in `[0, ..., config.num_labels - 1]`.
    """
    outputs = self.nezha(
        input_ids,
        attention_mask=attention_mask,
        token_type_ids=token_type_ids,
        head_mask=head_mask,
        inputs_embeds=inputs_embeds,
        output_attentions=output_attentions,
        output_hidden_states=output_hidden_states
    )

    sequence_output = outputs[0]

    sequence_output = self.dropout(sequence_output)
    logits = self.classifier(sequence_output)

    loss = None
    if labels is not None:
        loss_fct = nn.CrossEntropyLoss()
        loss = loss_fct(logits.view(-1, self.num_labels), labels.view(-1))

    output = (logits,) + outputs[2:]
    return ((loss,) + output) if loss is not None else output

mindnlp.transformers.models.nezha.nezha.NezhaIntermediate

Bases: Module

Nezha Intermediate

Source code in mindnlp/transformers/models/nezha/nezha.py 
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
class NezhaIntermediate(nn.Module):
    """Nezha Intermediate"""
    def __init__(self, config):
        """
        Initializes a NezhaIntermediate object with the provided configuration.

        Args:
            self: The object instance.
            config:
                An object containing configuration settings.

                - Type: Any valid object.
                - Purpose: The configuration settings for the NezhaIntermediate object.
                - Restrictions: None.

        Returns:
            None.

        Raises:
            TypeError: If the 'config' parameter is not provided.
            ValueError: If the 'config.hidden_size' or 'config.intermediate_size' are invalid.
            KeyError: If the 'config.hidden_act' value is not found in the ACT2FN dictionary.
        """
        super().__init__()
        self.dense = nn.Linear(config.hidden_size, config.intermediate_size)
        if isinstance(config.hidden_act, str):
            self.intermediate_act_fn = ACT2FN[config.hidden_act]
        else:
            self.intermediate_act_fn = config.hidden_act

    def forward(self, hidden_states):
        """
        This method forwards the intermediate hidden states for the NezhaIntermediate class.

        Args:
            self (NezhaIntermediate): The instance of the NezhaIntermediate class.
            hidden_states (tensor): The input hidden states to be processed.

        Returns:
            tensor: The processed intermediate hidden states.

        Raises:
            None.
        """
        hidden_states = self.dense(hidden_states)
        hidden_states = self.intermediate_act_fn(hidden_states)
        return hidden_states

mindnlp.transformers.models.nezha.nezha.NezhaIntermediate.__init__(config)

Initializes a NezhaIntermediate object with the provided configuration.

PARAMETER DESCRIPTION
self

The object instance.
config

An object containing configuration settings.

  • Type: Any valid object.
  • Purpose: The configuration settings for the NezhaIntermediate object.
  • Restrictions: None.
RETURNS DESCRIPTION

None.
RAISES DESCRIPTION
TypeError

If the 'config' parameter is not provided.
ValueError

If the 'config.hidden_size' or 'config.intermediate_size' are invalid.
KeyError

If the 'config.hidden_act' value is not found in the ACT2FN dictionary.
Source code in mindnlp/transformers/models/nezha/nezha.py 
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
def __init__(self, config):
    """
    Initializes a NezhaIntermediate object with the provided configuration.

    Args:
        self: The object instance.
        config:
            An object containing configuration settings.

            - Type: Any valid object.
            - Purpose: The configuration settings for the NezhaIntermediate object.
            - Restrictions: None.

    Returns:
        None.

    Raises:
        TypeError: If the 'config' parameter is not provided.
        ValueError: If the 'config.hidden_size' or 'config.intermediate_size' are invalid.
        KeyError: If the 'config.hidden_act' value is not found in the ACT2FN dictionary.
    """
    super().__init__()
    self.dense = nn.Linear(config.hidden_size, config.intermediate_size)
    if isinstance(config.hidden_act, str):
        self.intermediate_act_fn = ACT2FN[config.hidden_act]
    else:
        self.intermediate_act_fn = config.hidden_act

mindnlp.transformers.models.nezha.nezha.NezhaIntermediate.forward(hidden_states)

This method forwards the intermediate hidden states for the NezhaIntermediate class.

PARAMETER DESCRIPTION
self

The instance of the NezhaIntermediate class.

TYPE: NezhaIntermediate

hidden_states

The input hidden states to be processed.

TYPE: tensor

RETURNS DESCRIPTION
tensor

The processed intermediate hidden states.
Source code in mindnlp/transformers/models/nezha/nezha.py 
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
def forward(self, hidden_states):
    """
    This method forwards the intermediate hidden states for the NezhaIntermediate class.

    Args:
        self (NezhaIntermediate): The instance of the NezhaIntermediate class.
        hidden_states (tensor): The input hidden states to be processed.

    Returns:
        tensor: The processed intermediate hidden states.

    Raises:
        None.
    """
    hidden_states = self.dense(hidden_states)
    hidden_states = self.intermediate_act_fn(hidden_states)
    return hidden_states

mindnlp.transformers.models.nezha.nezha.NezhaLMPredictionHead

Bases: Module

Nezha LMLMPredictionHead

Source code in mindnlp/transformers/models/nezha/nezha.py 
 975
 976
 977
 978
 979
 980
 981
 982
 983
 984
 985
 986
 987
 988
 989
 990
 991
 992
 993
 994
 995
 996
 997
 998
 999
1000
1001
1002
1003
1004
1005
1006
1007
1008
1009
1010
1011
1012
1013
1014
1015
1016
1017
1018
1019
1020
1021
1022
1023
1024
class NezhaLMPredictionHead(nn.Module):
    """Nezha LMLMPredictionHead"""
    def __init__(self, config):
        """
        Initializes the NezhaLMPredictionHead class.

        Args:
            self: The object instance.
            config:
                A configuration object that holds various parameters for the model.

                - Type: Any valid configuration object.
                - Purpose: Specifies the configuration settings for the NezhaLMPredictionHead.
                - Restrictions: None.

        Returns:
            None

        Raises:
            None
        """
        super().__init__()
        self.transform = NezhaPredictionHeadTransform(config)

        # The output weights are the same as the input embeddings, but there is
        # an output-only bias for each token.
        self.decoder = nn.Linear(config.hidden_size, config.vocab_size, bias=False)

        self.bias = Parameter(ops.zeros(config.vocab_size))

        # Need a link between the two variables so that the bias is correctly resized with `resize_token_embeddings`
        self.decoder.bias = self.bias

    def forward(self, hidden_states):
        """
        Constructs the prediction head for Nezha Language Model.

        Args:
            self (NezhaLMPredictionHead): The instance of NezhaLMPredictionHead class.
            hidden_states (Tensor): The hidden states to be processed for prediction.

        Returns:
            hidden_states: The forwarded prediction head for Nezha Language Model.

        Raises:
            None.
        """
        hidden_states = self.transform(hidden_states)
        hidden_states = self.decoder(hidden_states)
        return hidden_states

mindnlp.transformers.models.nezha.nezha.NezhaLMPredictionHead.__init__(config)

Initializes the NezhaLMPredictionHead class.

PARAMETER DESCRIPTION
self

The object instance.
config

A configuration object that holds various parameters for the model.

  • Type: Any valid configuration object.
  • Purpose: Specifies the configuration settings for the NezhaLMPredictionHead.
  • Restrictions: None.
RETURNS DESCRIPTION

None
Source code in mindnlp/transformers/models/nezha/nezha.py 
 977
 978
 979
 980
 981
 982
 983
 984
 985
 986
 987
 988
 989
 990
 991
 992
 993
 994
 995
 996
 997
 998
 999
1000
1001
1002
1003
1004
1005
1006
def __init__(self, config):
    """
    Initializes the NezhaLMPredictionHead class.

    Args:
        self: The object instance.
        config:
            A configuration object that holds various parameters for the model.

            - Type: Any valid configuration object.
            - Purpose: Specifies the configuration settings for the NezhaLMPredictionHead.
            - Restrictions: None.

    Returns:
        None

    Raises:
        None
    """
    super().__init__()
    self.transform = NezhaPredictionHeadTransform(config)

    # The output weights are the same as the input embeddings, but there is
    # an output-only bias for each token.
    self.decoder = nn.Linear(config.hidden_size, config.vocab_size, bias=False)

    self.bias = Parameter(ops.zeros(config.vocab_size))

    # Need a link between the two variables so that the bias is correctly resized with `resize_token_embeddings`
    self.decoder.bias = self.bias

mindnlp.transformers.models.nezha.nezha.NezhaLMPredictionHead.forward(hidden_states)

Constructs the prediction head for Nezha Language Model.

PARAMETER DESCRIPTION
self

The instance of NezhaLMPredictionHead class.

TYPE: NezhaLMPredictionHead

hidden_states

The hidden states to be processed for prediction.

TYPE: Tensor

RETURNS DESCRIPTION
hidden_states

The forwarded prediction head for Nezha Language Model.
Source code in mindnlp/transformers/models/nezha/nezha.py 
1008
1009
1010
1011
1012
1013
1014
1015
1016
1017
1018
1019
1020
1021
1022
1023
1024
def forward(self, hidden_states):
    """
    Constructs the prediction head for Nezha Language Model.

    Args:
        self (NezhaLMPredictionHead): The instance of NezhaLMPredictionHead class.
        hidden_states (Tensor): The hidden states to be processed for prediction.

    Returns:
        hidden_states: The forwarded prediction head for Nezha Language Model.

    Raises:
        None.
    """
    hidden_states = self.transform(hidden_states)
    hidden_states = self.decoder(hidden_states)
    return hidden_states

mindnlp.transformers.models.nezha.nezha.NezhaLayer

Bases: Module

Nezha Layer

Source code in mindnlp/transformers/models/nezha/nezha.py 
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
class NezhaLayer(nn.Module):
    """Nezha Layer"""
    def __init__(self, config):
        """
        Initializes a NezhaLayer object with the provided configuration.

        Args:
            self: The instance of the NezhaLayer class.
            config:
                An object containing configuration parameters for the NezhaLayer.

                - Type: Config object
                - Purpose: Specifies the configuration settings for the NezhaLayer.
                - Restrictions: Must be a valid configuration object.

        Returns:
            None.

        Raises:
            ValueError: Raised if the cross attention is added but the model is not used as a decoder model.
        """
        super().__init__()
        self.chunk_size_feed_forward = config.chunk_size_feed_forward
        self.seq_len_dim = 1
        self.attention = NezhaAttention(config)
        self.is_decoder = config.is_decoder
        self.add_cross_attention = config.add_cross_attention
        if self.add_cross_attention:
            if not self.is_decoder:
                raise ValueError(f"{self} should be used as a decoder model if cross attention is added")
            self.crossattention = NezhaAttention(config)
        self.intermediate = NezhaIntermediate(config)
        self.output = NezhaOutput(config)

    def forward(self, hidden_states, attention_mask = None,
                  head_mask = None, encoder_hidden_states = None,
                  encoder_attention_mask = None, past_key_value = None,
                  output_attentions = False):
        """
        Method: forward

        Description:
            Constructs the NezhaLayer by performing self-attention and potentially cross-attention operations based on
            the provided parameters.

        Args:
            self: The object instance.
            hidden_states (Tensor): The input hidden states for the layer.
            attention_mask (Tensor, optional): Mask to prevent attention to certain positions.
            head_mask (Tensor, optional): Mask to prevent attention to certain heads.
            encoder_hidden_states (Tensor, optional): Hidden states of the encoder if cross-attention is needed.
            encoder_attention_mask (Tensor, optional): Mask for encoder attention.
            past_key_value (Tuple, optional): Tuple containing past key and value tensors for optimization.
            output_attentions (bool): Flag to indicate whether to output attentions.

        Returns:
            None.

        Raises:
            ValueError: If `encoder_hidden_states` are provided but cross-attention layers are not instantiated
                in the model.
        """
        # decoder uni-directional self-attention cached key/values tuple is at positions 1,2
        self_attn_past_key_value = past_key_value[:2] if past_key_value is not None else None
        self_attention_outputs = self.attention(
            hidden_states,
            attention_mask,
            head_mask,
            output_attentions=output_attentions,
            past_key_value=self_attn_past_key_value,
        )
        attention_output = self_attention_outputs[0]

        # if decoder, the last output is tuple of self-attn cache
        if self.is_decoder:
            outputs = self_attention_outputs[1:-1]
            present_key_value = self_attention_outputs[-1]
        else:
            outputs = self_attention_outputs[1:]  # add self attentions if we output attention weights

        cross_attn_present_key_value = None
        if self.is_decoder and encoder_hidden_states is not None:
            if not hasattr(self, "crossattention"):
                raise ValueError(
                    f"If `encoder_hidden_states` are passed, {self} has to be instantiated with cross-attention layers"
                    " by setting `config.add_cross_attention=True`"
                )

            # cross_attn cached key/values tuple is at positions 3,4 of past_key_value tuple
            cross_attn_past_key_value = past_key_value[-2:] if past_key_value is not None else None
            cross_attention_outputs = self.crossattention(
                attention_output,
                attention_mask,
                head_mask,
                encoder_hidden_states,
                encoder_attention_mask,
                cross_attn_past_key_value,
                output_attentions,
            )
            attention_output = cross_attention_outputs[0]
            outputs = outputs + cross_attention_outputs[1:-1]  # add cross attentions if we output attention weights

            # add cross-attn cache to positions 3,4 of present_key_value tuple
            cross_attn_present_key_value = cross_attention_outputs[-1]
            present_key_value = present_key_value + cross_attn_present_key_value

        layer_output = apply_chunking_to_forward(
            self.feed_forward_chunk, self.chunk_size_feed_forward, self.seq_len_dim, attention_output
        )
        outputs = (layer_output,) + outputs

        # if decoder, return the attn key/values as the last output
        if self.is_decoder:
            outputs = outputs + (present_key_value,)

        return outputs

    def feed_forward_chunk(self, attention_output):
        """feed forward chunk"""
        intermediate_output = self.intermediate(attention_output)
        layer_output = self.output(intermediate_output, attention_output)
        return layer_output

mindnlp.transformers.models.nezha.nezha.NezhaLayer.__init__(config)

Initializes a NezhaLayer object with the provided configuration.

PARAMETER DESCRIPTION
self

The instance of the NezhaLayer class.
config

An object containing configuration parameters for the NezhaLayer.

  • Type: Config object
  • Purpose: Specifies the configuration settings for the NezhaLayer.
  • Restrictions: Must be a valid configuration object.
RETURNS DESCRIPTION

None.
RAISES DESCRIPTION
ValueError

Raised if the cross attention is added but the model is not used as a decoder model.
Source code in mindnlp/transformers/models/nezha/nezha.py 
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
def __init__(self, config):
    """
    Initializes a NezhaLayer object with the provided configuration.

    Args:
        self: The instance of the NezhaLayer class.
        config:
            An object containing configuration parameters for the NezhaLayer.

            - Type: Config object
            - Purpose: Specifies the configuration settings for the NezhaLayer.
            - Restrictions: Must be a valid configuration object.

    Returns:
        None.

    Raises:
        ValueError: Raised if the cross attention is added but the model is not used as a decoder model.
    """
    super().__init__()
    self.chunk_size_feed_forward = config.chunk_size_feed_forward
    self.seq_len_dim = 1
    self.attention = NezhaAttention(config)
    self.is_decoder = config.is_decoder
    self.add_cross_attention = config.add_cross_attention
    if self.add_cross_attention:
        if not self.is_decoder:
            raise ValueError(f"{self} should be used as a decoder model if cross attention is added")
        self.crossattention = NezhaAttention(config)
    self.intermediate = NezhaIntermediate(config)
    self.output = NezhaOutput(config)

mindnlp.transformers.models.nezha.nezha.NezhaLayer.feed_forward_chunk(attention_output)

feed forward chunk

Source code in mindnlp/transformers/models/nezha/nezha.py 
742
743
744
745
746
def feed_forward_chunk(self, attention_output):
    """feed forward chunk"""
    intermediate_output = self.intermediate(attention_output)
    layer_output = self.output(intermediate_output, attention_output)
    return layer_output

mindnlp.transformers.models.nezha.nezha.NezhaLayer.forward(hidden_states, attention_mask=None, head_mask=None, encoder_hidden_states=None, encoder_attention_mask=None, past_key_value=None, output_attentions=False)

Description

Constructs the NezhaLayer by performing self-attention and potentially cross-attention operations based on the provided parameters.

PARAMETER DESCRIPTION
self

The object instance.
hidden_states

The input hidden states for the layer.

TYPE: Tensor

attention_mask

Mask to prevent attention to certain positions.

TYPE: Tensor DEFAULT: None

head_mask

Mask to prevent attention to certain heads.

TYPE: Tensor DEFAULT: None

encoder_hidden_states

Hidden states of the encoder if cross-attention is needed.

TYPE: Tensor DEFAULT: None

encoder_attention_mask

Mask for encoder attention.

TYPE: Tensor DEFAULT: None

past_key_value

Tuple containing past key and value tensors for optimization.

TYPE: Tuple DEFAULT: None

output_attentions

Flag to indicate whether to output attentions.

TYPE: bool DEFAULT: False

RETURNS DESCRIPTION

None.
RAISES DESCRIPTION
ValueError

If encoder_hidden_states are provided but cross-attention layers are not instantiated in the model.
Source code in mindnlp/transformers/models/nezha/nezha.py 
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
def forward(self, hidden_states, attention_mask = None,
              head_mask = None, encoder_hidden_states = None,
              encoder_attention_mask = None, past_key_value = None,
              output_attentions = False):
    """
    Method: forward

    Description:
        Constructs the NezhaLayer by performing self-attention and potentially cross-attention operations based on
        the provided parameters.

    Args:
        self: The object instance.
        hidden_states (Tensor): The input hidden states for the layer.
        attention_mask (Tensor, optional): Mask to prevent attention to certain positions.
        head_mask (Tensor, optional): Mask to prevent attention to certain heads.
        encoder_hidden_states (Tensor, optional): Hidden states of the encoder if cross-attention is needed.
        encoder_attention_mask (Tensor, optional): Mask for encoder attention.
        past_key_value (Tuple, optional): Tuple containing past key and value tensors for optimization.
        output_attentions (bool): Flag to indicate whether to output attentions.

    Returns:
        None.

    Raises:
        ValueError: If `encoder_hidden_states` are provided but cross-attention layers are not instantiated
            in the model.
    """
    # decoder uni-directional self-attention cached key/values tuple is at positions 1,2
    self_attn_past_key_value = past_key_value[:2] if past_key_value is not None else None
    self_attention_outputs = self.attention(
        hidden_states,
        attention_mask,
        head_mask,
        output_attentions=output_attentions,
        past_key_value=self_attn_past_key_value,
    )
    attention_output = self_attention_outputs[0]

    # if decoder, the last output is tuple of self-attn cache
    if self.is_decoder:
        outputs = self_attention_outputs[1:-1]
        present_key_value = self_attention_outputs[-1]
    else:
        outputs = self_attention_outputs[1:]  # add self attentions if we output attention weights

    cross_attn_present_key_value = None
    if self.is_decoder and encoder_hidden_states is not None:
        if not hasattr(self, "crossattention"):
            raise ValueError(
                f"If `encoder_hidden_states` are passed, {self} has to be instantiated with cross-attention layers"
                " by setting `config.add_cross_attention=True`"
            )

        # cross_attn cached key/values tuple is at positions 3,4 of past_key_value tuple
        cross_attn_past_key_value = past_key_value[-2:] if past_key_value is not None else None
        cross_attention_outputs = self.crossattention(
            attention_output,
            attention_mask,
            head_mask,
            encoder_hidden_states,
            encoder_attention_mask,
            cross_attn_past_key_value,
            output_attentions,
        )
        attention_output = cross_attention_outputs[0]
        outputs = outputs + cross_attention_outputs[1:-1]  # add cross attentions if we output attention weights

        # add cross-attn cache to positions 3,4 of present_key_value tuple
        cross_attn_present_key_value = cross_attention_outputs[-1]
        present_key_value = present_key_value + cross_attn_present_key_value

    layer_output = apply_chunking_to_forward(
        self.feed_forward_chunk, self.chunk_size_feed_forward, self.seq_len_dim, attention_output
    )
    outputs = (layer_output,) + outputs

    # if decoder, return the attn key/values as the last output
    if self.is_decoder:
        outputs = outputs + (present_key_value,)

    return outputs

mindnlp.transformers.models.nezha.nezha.NezhaModel

Bases: NezhaPreTrainedModel

Nezha Model

Source code in mindnlp/transformers/models/nezha/nezha.py 
1274
1275
1276
1277
1278
1279
1280
1281
1282
1283
1284
1285
1286
1287
1288
1289
1290
1291
1292
1293
1294
1295
1296
1297
1298
1299
1300
1301
1302
1303
1304
1305
1306
1307
1308
1309
1310
1311
1312
1313
1314
1315
1316
1317
1318
1319
1320
1321
1322
1323
1324
1325
1326
1327
1328
1329
1330
1331
1332
1333
1334
1335
1336
1337
1338
1339
1340
1341
1342
1343
1344
1345
1346
1347
1348
1349
1350
1351
1352
1353
1354
1355
1356
1357
1358
1359
1360
1361
1362
1363
1364
1365
1366
1367
1368
1369
1370
1371
1372
1373
1374
1375
1376
1377
1378
1379
1380
1381
1382
1383
1384
1385
1386
1387
1388
1389
1390
1391
1392
1393
1394
1395
1396
1397
1398
1399
1400
1401
1402
1403
1404
1405
1406
1407
1408
1409
1410
1411
1412
1413
1414
1415
1416
1417
1418
1419
1420
1421
1422
1423
1424
1425
1426
1427
1428
1429
1430
1431
1432
1433
1434
1435
1436
1437
1438
1439
1440
class NezhaModel(NezhaPreTrainedModel):
    """Nezha Model"""
    def __init__(self, config, add_pooling_layer=True):
        """
        Initializes a new instance of the NezhaModel class.

        Args:
            self: The instance of the NezhaModel class.
            config: An instance of the configuration for the NezhaModel. It is used to configure the model's behavior.
            add_pooling_layer (bool): A boolean flag indicating whether to add a pooling layer to the model.
                Default is True.

        Returns:
            None.

        Raises:
            None.
        """
        super().__init__(config)
        self.config = config
        self.embeddings = NezhaEmbeddings(config)
        self.encoder = NezhaEncoder(config)
        self.pooler = NezhaPooler(config) if add_pooling_layer else None

    def get_input_embeddings(self):
        """
        Retrieve the input embeddings from the NezhaModel.

        Args:
            self (NezhaModel): The instance of NezhaModel.

        Returns:
            None.

        Raises:
            None.
        """
        return self.embeddings.word_embeddings

    def set_input_embeddings(self, value):
        """
        Sets the input embeddings for the NezhaModel.

        Args:
            self (NezhaModel): The instance of the NezhaModel class.
            value: The input embeddings to be set. It should be of type 'torch.Tensor' or any tensor-like object.

        Returns:
            None.

        Raises:
            None.
        """
        self.embeddings.word_embeddings = value

    def _prune_heads(self, heads_to_prune):
        """
        Prunes heads of the model. heads_to_prune: dict of {layer_num: list of heads to prune in this layer} See base
        class PreTrainedModel
        """
        for layer, heads in heads_to_prune.items():
            self.encoder.layer[layer].attention.prune_heads(heads)

    def forward(self, input_ids = None, attention_mask = None,
                  token_type_ids = None, head_mask = None,
                  inputs_embeds = None, encoder_hidden_states = None,
                  encoder_attention_mask = None, past_key_values = None,
                  use_cache = None, output_attentions = None,
                  output_hidden_states = None):
        """
        This method forwards the NezhaModel by processing input data through the model's encoder and embeddings.

        Args:
            self: The instance of the NezhaModel class.
            input_ids (Tensor, optional): The input token IDs. Default is None.
            attention_mask (Tensor, optional): The attention mask for the input. Default is None.
            token_type_ids (Tensor, optional): The token type IDs for the input. Default is None.
            head_mask (Tensor, optional): The head mask for the model's multi-head attention layers. Default is None.
            inputs_embeds (Tensor, optional): The embedded input tokens. Default is None.
            encoder_hidden_states (Tensor, optional): The hidden states from the encoder. Default is None.
            encoder_attention_mask (Tensor, optional): The attention mask for the encoder. Default is None.
            past_key_values (Tuple, optional): Cached key-value states from previous iterations. Default is None.
            use_cache (bool, optional): Flag indicating whether to use caching. Default is None.
            output_attentions (bool, optional): Flag indicating whether to output attentions. Default is None.
            output_hidden_states (bool, optional): Flag indicating whether to output hidden states. Default is None.

        Returns:
            Tuple: A tuple containing the sequence output, pooled output, and any additional encoder outputs.

        Raises:
            ValueError: Raised when both input_ids and inputs_embeds are provided simultaneously.
            ValueError: Raised when neither input_ids nor inputs_embeds are specified.
        """
        output_attentions = output_attentions if output_attentions is not None else self.config.output_attentions
        output_hidden_states = (
            output_hidden_states if output_hidden_states is not None else self.config.output_hidden_states
        )

        if self.config.is_decoder:
            use_cache = use_cache if use_cache is not None else self.config.use_cache
        else:
            use_cache = False

        if input_ids is not None and inputs_embeds is not None:
            raise ValueError("You cannot specify both input_ids and inputs_embeds at the same time")
        if input_ids is not None:
            input_shape = input_ids.shape
        elif inputs_embeds is not None:
            input_shape = inputs_embeds.shape[:-1]
        else:
            raise ValueError("You have to specify either input_ids or inputs_embeds")

        batch_size, seq_length = input_shape
        past_key_values_length = past_key_values[0][0].shape[2] if past_key_values is not None else 0

        if attention_mask is None:
            attention_mask = ops.ones(((batch_size, seq_length + past_key_values_length)))

        if token_type_ids is None:
            if hasattr(self.embeddings, "token_type_ids"):
                buffered_token_type_ids = self.embeddings.token_type_ids[:, :seq_length]
                buffered_token_type_ids_expanded = ops.broadcast_to(buffered_token_type_ids, (batch_size, seq_length))
                token_type_ids = buffered_token_type_ids_expanded
            else:
                token_type_ids = ops.zeros(input_shape, dtype=mindspore.int64)

        # We can provide a self-attention mask of dimensions [batch_size, from_seq_length, to_seq_length]
        # ourselves in which case we just need to make it broadcastable to all heads.
        extended_attention_mask = self.get_extended_attention_mask(attention_mask, input_shape)
        # If a 2D or 3D attention mask is provided for the cross-attention
        # we need to make broadcastable to [batch_size, num_heads, seq_length, seq_length]
        if self.config.is_decoder and encoder_hidden_states is not None:
            encoder_batch_size, encoder_sequence_length, _ = encoder_hidden_states.shape
            encoder_hidden_shape = (encoder_batch_size, encoder_sequence_length)
            if encoder_attention_mask is None:
                encoder_attention_mask = ops.ones(encoder_hidden_shape)
            encoder_extended_attention_mask = self.invert_attention_mask(encoder_attention_mask)
        else:
            encoder_extended_attention_mask = None

        # Prepare head mask if needed
        # 1.0 in head_mask indicate we keep the head
        # attention_probs has shape bsz x n_heads x N x N
        # input head_mask has shape [num_heads] or [num_hidden_layers x num_heads]
        # and head_mask is converted to shape [num_hidden_layers x batch x num_heads x seq_length x seq_length]
        head_mask = self.get_head_mask(head_mask, self.config.num_hidden_layers)

        embedding_output = self.embeddings(
            input_ids=input_ids,
            token_type_ids=token_type_ids,
            inputs_embeds=inputs_embeds,
        )
        encoder_outputs = self.encoder(
            embedding_output,
            attention_mask=extended_attention_mask,
            head_mask=head_mask,
            encoder_hidden_states=encoder_hidden_states,
            encoder_attention_mask=encoder_extended_attention_mask,
            past_key_values=past_key_values,
            use_cache=use_cache,
            output_attentions=output_attentions,
            output_hidden_states=output_hidden_states
        )
        sequence_output = encoder_outputs[0]
        pooled_output = self.pooler(sequence_output) if self.pooler is not None else None

        return (sequence_output, pooled_output) + encoder_outputs[1:]

mindnlp.transformers.models.nezha.nezha.NezhaModel.__init__(config, add_pooling_layer=True)

Initializes a new instance of the NezhaModel class.

PARAMETER DESCRIPTION
self

The instance of the NezhaModel class.
config

An instance of the configuration for the NezhaModel. It is used to configure the model's behavior.
add_pooling_layer

A boolean flag indicating whether to add a pooling layer to the model. Default is True.

TYPE: bool DEFAULT: True

RETURNS DESCRIPTION

None.
Source code in mindnlp/transformers/models/nezha/nezha.py 
1276
1277
1278
1279
1280
1281
1282
1283
1284
1285
1286
1287
1288
1289
1290
1291
1292
1293
1294
1295
1296
def __init__(self, config, add_pooling_layer=True):
    """
    Initializes a new instance of the NezhaModel class.

    Args:
        self: The instance of the NezhaModel class.
        config: An instance of the configuration for the NezhaModel. It is used to configure the model's behavior.
        add_pooling_layer (bool): A boolean flag indicating whether to add a pooling layer to the model.
            Default is True.

    Returns:
        None.

    Raises:
        None.
    """
    super().__init__(config)
    self.config = config
    self.embeddings = NezhaEmbeddings(config)
    self.encoder = NezhaEncoder(config)
    self.pooler = NezhaPooler(config) if add_pooling_layer else None

mindnlp.transformers.models.nezha.nezha.NezhaModel.forward(input_ids=None, attention_mask=None, token_type_ids=None, head_mask=None, inputs_embeds=None, encoder_hidden_states=None, encoder_attention_mask=None, past_key_values=None, use_cache=None, output_attentions=None, output_hidden_states=None)

This method forwards the NezhaModel by processing input data through the model's encoder and embeddings.

PARAMETER DESCRIPTION
self

The instance of the NezhaModel class.
input_ids

The input token IDs. Default is None.

TYPE: Tensor DEFAULT: None

attention_mask

The attention mask for the input. Default is None.

TYPE: Tensor DEFAULT: None

token_type_ids

The token type IDs for the input. Default is None.

TYPE: Tensor DEFAULT: None

head_mask

The head mask for the model's multi-head attention layers. Default is None.

TYPE: Tensor DEFAULT: None

inputs_embeds

The embedded input tokens. Default is None.

TYPE: Tensor DEFAULT: None

encoder_hidden_states

The hidden states from the encoder. Default is None.

TYPE: Tensor DEFAULT: None

encoder_attention_mask

The attention mask for the encoder. Default is None.

TYPE: Tensor DEFAULT: None

past_key_values

Cached key-value states from previous iterations. Default is None.

TYPE: Tuple DEFAULT: None

use_cache

Flag indicating whether to use caching. Default is None.

TYPE: bool DEFAULT: None

output_attentions

Flag indicating whether to output attentions. Default is None.

TYPE: bool DEFAULT: None

output_hidden_states

Flag indicating whether to output hidden states. Default is None.

TYPE: bool DEFAULT: None

RETURNS DESCRIPTION
Tuple

A tuple containing the sequence output, pooled output, and any additional encoder outputs.
RAISES DESCRIPTION
ValueError

Raised when both input_ids and inputs_embeds are provided simultaneously.
ValueError

Raised when neither input_ids nor inputs_embeds are specified.
Source code in mindnlp/transformers/models/nezha/nezha.py 
1337
1338
1339
1340
1341
1342
1343
1344
1345
1346
1347
1348
1349
1350
1351
1352
1353
1354
1355
1356
1357
1358
1359
1360
1361
1362
1363
1364
1365
1366
1367
1368
1369
1370
1371
1372
1373
1374
1375
1376
1377
1378
1379
1380
1381
1382
1383
1384
1385
1386
1387
1388
1389
1390
1391
1392
1393
1394
1395
1396
1397
1398
1399
1400
1401
1402
1403
1404
1405
1406
1407
1408
1409
1410
1411
1412
1413
1414
1415
1416
1417
1418
1419
1420
1421
1422
1423
1424
1425
1426
1427
1428
1429
1430
1431
1432
1433
1434
1435
1436
1437
1438
1439
1440
def forward(self, input_ids = None, attention_mask = None,
              token_type_ids = None, head_mask = None,
              inputs_embeds = None, encoder_hidden_states = None,
              encoder_attention_mask = None, past_key_values = None,
              use_cache = None, output_attentions = None,
              output_hidden_states = None):
    """
    This method forwards the NezhaModel by processing input data through the model's encoder and embeddings.

    Args:
        self: The instance of the NezhaModel class.
        input_ids (Tensor, optional): The input token IDs. Default is None.
        attention_mask (Tensor, optional): The attention mask for the input. Default is None.
        token_type_ids (Tensor, optional): The token type IDs for the input. Default is None.
        head_mask (Tensor, optional): The head mask for the model's multi-head attention layers. Default is None.
        inputs_embeds (Tensor, optional): The embedded input tokens. Default is None.
        encoder_hidden_states (Tensor, optional): The hidden states from the encoder. Default is None.
        encoder_attention_mask (Tensor, optional): The attention mask for the encoder. Default is None.
        past_key_values (Tuple, optional): Cached key-value states from previous iterations. Default is None.
        use_cache (bool, optional): Flag indicating whether to use caching. Default is None.
        output_attentions (bool, optional): Flag indicating whether to output attentions. Default is None.
        output_hidden_states (bool, optional): Flag indicating whether to output hidden states. Default is None.

    Returns:
        Tuple: A tuple containing the sequence output, pooled output, and any additional encoder outputs.

    Raises:
        ValueError: Raised when both input_ids and inputs_embeds are provided simultaneously.
        ValueError: Raised when neither input_ids nor inputs_embeds are specified.
    """
    output_attentions = output_attentions if output_attentions is not None else self.config.output_attentions
    output_hidden_states = (
        output_hidden_states if output_hidden_states is not None else self.config.output_hidden_states
    )

    if self.config.is_decoder:
        use_cache = use_cache if use_cache is not None else self.config.use_cache
    else:
        use_cache = False

    if input_ids is not None and inputs_embeds is not None:
        raise ValueError("You cannot specify both input_ids and inputs_embeds at the same time")
    if input_ids is not None:
        input_shape = input_ids.shape
    elif inputs_embeds is not None:
        input_shape = inputs_embeds.shape[:-1]
    else:
        raise ValueError("You have to specify either input_ids or inputs_embeds")

    batch_size, seq_length = input_shape
    past_key_values_length = past_key_values[0][0].shape[2] if past_key_values is not None else 0

    if attention_mask is None:
        attention_mask = ops.ones(((batch_size, seq_length + past_key_values_length)))

    if token_type_ids is None:
        if hasattr(self.embeddings, "token_type_ids"):
            buffered_token_type_ids = self.embeddings.token_type_ids[:, :seq_length]
            buffered_token_type_ids_expanded = ops.broadcast_to(buffered_token_type_ids, (batch_size, seq_length))
            token_type_ids = buffered_token_type_ids_expanded
        else:
            token_type_ids = ops.zeros(input_shape, dtype=mindspore.int64)

    # We can provide a self-attention mask of dimensions [batch_size, from_seq_length, to_seq_length]
    # ourselves in which case we just need to make it broadcastable to all heads.
    extended_attention_mask = self.get_extended_attention_mask(attention_mask, input_shape)
    # If a 2D or 3D attention mask is provided for the cross-attention
    # we need to make broadcastable to [batch_size, num_heads, seq_length, seq_length]
    if self.config.is_decoder and encoder_hidden_states is not None:
        encoder_batch_size, encoder_sequence_length, _ = encoder_hidden_states.shape
        encoder_hidden_shape = (encoder_batch_size, encoder_sequence_length)
        if encoder_attention_mask is None:
            encoder_attention_mask = ops.ones(encoder_hidden_shape)
        encoder_extended_attention_mask = self.invert_attention_mask(encoder_attention_mask)
    else:
        encoder_extended_attention_mask = None

    # Prepare head mask if needed
    # 1.0 in head_mask indicate we keep the head
    # attention_probs has shape bsz x n_heads x N x N
    # input head_mask has shape [num_heads] or [num_hidden_layers x num_heads]
    # and head_mask is converted to shape [num_hidden_layers x batch x num_heads x seq_length x seq_length]
    head_mask = self.get_head_mask(head_mask, self.config.num_hidden_layers)

    embedding_output = self.embeddings(
        input_ids=input_ids,
        token_type_ids=token_type_ids,
        inputs_embeds=inputs_embeds,
    )
    encoder_outputs = self.encoder(
        embedding_output,
        attention_mask=extended_attention_mask,
        head_mask=head_mask,
        encoder_hidden_states=encoder_hidden_states,
        encoder_attention_mask=encoder_extended_attention_mask,
        past_key_values=past_key_values,
        use_cache=use_cache,
        output_attentions=output_attentions,
        output_hidden_states=output_hidden_states
    )
    sequence_output = encoder_outputs[0]
    pooled_output = self.pooler(sequence_output) if self.pooler is not None else None

    return (sequence_output, pooled_output) + encoder_outputs[1:]

mindnlp.transformers.models.nezha.nezha.NezhaModel.get_input_embeddings()

Retrieve the input embeddings from the NezhaModel.

PARAMETER DESCRIPTION
self

The instance of NezhaModel.

TYPE: NezhaModel

RETURNS DESCRIPTION

None.
Source code in mindnlp/transformers/models/nezha/nezha.py 
1298
1299
1300
1301
1302
1303
1304
1305
1306
1307
1308
1309
1310
1311
def get_input_embeddings(self):
    """
    Retrieve the input embeddings from the NezhaModel.

    Args:
        self (NezhaModel): The instance of NezhaModel.

    Returns:
        None.

    Raises:
        None.
    """
    return self.embeddings.word_embeddings

mindnlp.transformers.models.nezha.nezha.NezhaModel.set_input_embeddings(value)

Sets the input embeddings for the NezhaModel.

PARAMETER DESCRIPTION
self

The instance of the NezhaModel class.

TYPE: NezhaModel

value

The input embeddings to be set. It should be of type 'torch.Tensor' or any tensor-like object.
RETURNS DESCRIPTION

None.
Source code in mindnlp/transformers/models/nezha/nezha.py 
1313
1314
1315
1316
1317
1318
1319
1320
1321
1322
1323
1324
1325
1326
1327
def set_input_embeddings(self, value):
    """
    Sets the input embeddings for the NezhaModel.

    Args:
        self (NezhaModel): The instance of the NezhaModel class.
        value: The input embeddings to be set. It should be of type 'torch.Tensor' or any tensor-like object.

    Returns:
        None.

    Raises:
        None.
    """
    self.embeddings.word_embeddings = value

mindnlp.transformers.models.nezha.nezha.NezhaOnlyMLMHead

Bases: Module

Nezha OnlyMLMHead

Source code in mindnlp/transformers/models/nezha/nezha.py 
1027
1028
1029
1030
1031
1032
1033
1034
1035
1036
1037
1038
1039
1040
1041
1042
1043
1044
1045
1046
1047
1048
1049
1050
1051
1052
1053
1054
1055
1056
1057
1058
1059
1060
1061
class NezhaOnlyMLMHead(nn.Module):
    """Nezha OnlyMLMHead"""
    def __init__(self, config):
        """Initializes a new instance of the NezhaOnlyMLMHead class.

        Args:
            self: The instance of the class.
            config: An object of type 'config' containing the configuration parameters.

        Returns:
            None.

        Raises:
            None.
        """
        super().__init__()
        self.predictions = NezhaLMPredictionHead(config)

    def forward(self, sequence_output):
        """
        Constructs the Masked Language Model (MLM) head for the Nezha model.

        Args:
            self (NezhaOnlyMLMHead): An instance of the NezhaOnlyMLMHead class.
            sequence_output (torch.Tensor): The output tensor of the Nezha model's encoder.
                Shape: (batch_size, sequence_length, hidden_size).

        Returns:
            None: This method modifies the internal state of the NezhaOnlyMLMHead instance.

        Raises:
            None.
        """
        prediction_scores = self.predictions(sequence_output)
        return prediction_scores

mindnlp.transformers.models.nezha.nezha.NezhaOnlyMLMHead.__init__(config)

Initializes a new instance of the NezhaOnlyMLMHead class.

PARAMETER DESCRIPTION
self

The instance of the class.
config

An object of type 'config' containing the configuration parameters.
RETURNS DESCRIPTION

None.
Source code in mindnlp/transformers/models/nezha/nezha.py 
1029
1030
1031
1032
1033
1034
1035
1036
1037
1038
1039
1040
1041
1042
1043
def __init__(self, config):
    """Initializes a new instance of the NezhaOnlyMLMHead class.

    Args:
        self: The instance of the class.
        config: An object of type 'config' containing the configuration parameters.

    Returns:
        None.

    Raises:
        None.
    """
    super().__init__()
    self.predictions = NezhaLMPredictionHead(config)

mindnlp.transformers.models.nezha.nezha.NezhaOnlyMLMHead.forward(sequence_output)

Constructs the Masked Language Model (MLM) head for the Nezha model.

PARAMETER DESCRIPTION
self

An instance of the NezhaOnlyMLMHead class.

TYPE: NezhaOnlyMLMHead

sequence_output

The output tensor of the Nezha model's encoder. Shape: (batch_size, sequence_length, hidden_size).

TYPE: Tensor

RETURNS DESCRIPTION
None

This method modifies the internal state of the NezhaOnlyMLMHead instance.
Source code in mindnlp/transformers/models/nezha/nezha.py 
1045
1046
1047
1048
1049
1050
1051
1052
1053
1054
1055
1056
1057
1058
1059
1060
1061
def forward(self, sequence_output):
    """
    Constructs the Masked Language Model (MLM) head for the Nezha model.

    Args:
        self (NezhaOnlyMLMHead): An instance of the NezhaOnlyMLMHead class.
        sequence_output (torch.Tensor): The output tensor of the Nezha model's encoder.
            Shape: (batch_size, sequence_length, hidden_size).

    Returns:
        None: This method modifies the internal state of the NezhaOnlyMLMHead instance.

    Raises:
        None.
    """
    prediction_scores = self.predictions(sequence_output)
    return prediction_scores

mindnlp.transformers.models.nezha.nezha.NezhaOnlyNSPHead

Bases: Module

Nezha OnlyNSPHead

Source code in mindnlp/transformers/models/nezha/nezha.py 
1064
1065
1066
1067
1068
1069
1070
1071
1072
1073
1074
1075
1076
1077
1078
1079
1080
1081
1082
1083
1084
1085
1086
1087
1088
1089
1090
1091
1092
1093
1094
1095
1096
1097
1098
1099
1100
1101
1102
1103
1104
1105
1106
1107
1108
class NezhaOnlyNSPHead(nn.Module):
    """Nezha OnlyNSPHead"""
    def __init__(self, config):
        """
        Initializes a new instance of the NezhaOnlyNSPHead class.

        Args:
            self: The instance of the NezhaOnlyNSPHead class.
            config: An instance of configuration class containing the hidden size parameter.
                It specifies the configuration settings for the NezhaOnlyNSPHead class.
                It is expected to have a hidden_size attribute, which represents the size of the hidden layer.

        Returns:
            None.

        Raises:
            TypeError: If the config parameter is not of the expected type.
            ValueError: If the config parameter does not contain the required hidden_size attribute.
        """
        super().__init__()
        self.seq_relationship = nn.Linear(config.hidden_size, 2)

    def forward(self, pooled_output):
        """
        Constructs the NSP (Next Sentence Prediction) head for the Nezha model.

        Args:
            self (NezhaOnlyNSPHead): An instance of the NezhaOnlyNSPHead class.
            pooled_output (torch.Tensor): The pooled output tensor of shape (batch_size, hidden_size).
                The pooled output is typically obtained by applying pooling operations (e.g., mean pooling, max pooling)
                over the sequence-level representations of the input tokens. It serves as the input to the NSP head.

        Returns:
            None.

        Raises:
            None.

        Note:
            The NSP head is responsible for predicting whether two input sentences are consecutive or not.
            It takes the pooled output tensor from the Nezha model and computes the sequence relationship score.
            The sequence relationship score is used to determine if the two input sentences are consecutive or not.
        """
        seq_relationship_score = self.seq_relationship(pooled_output)
        return seq_relationship_score

mindnlp.transformers.models.nezha.nezha.NezhaOnlyNSPHead.__init__(config)

Initializes a new instance of the NezhaOnlyNSPHead class.

PARAMETER DESCRIPTION
self

The instance of the NezhaOnlyNSPHead class.
config

An instance of configuration class containing the hidden size parameter. It specifies the configuration settings for the NezhaOnlyNSPHead class. It is expected to have a hidden_size attribute, which represents the size of the hidden layer.
RETURNS DESCRIPTION

None.
RAISES DESCRIPTION
TypeError

If the config parameter is not of the expected type.
ValueError

If the config parameter does not contain the required hidden_size attribute.
Source code in mindnlp/transformers/models/nezha/nezha.py 
1066
1067
1068
1069
1070
1071
1072
1073
1074
1075
1076
1077
1078
1079
1080
1081
1082
1083
1084
def __init__(self, config):
    """
    Initializes a new instance of the NezhaOnlyNSPHead class.

    Args:
        self: The instance of the NezhaOnlyNSPHead class.
        config: An instance of configuration class containing the hidden size parameter.
            It specifies the configuration settings for the NezhaOnlyNSPHead class.
            It is expected to have a hidden_size attribute, which represents the size of the hidden layer.

    Returns:
        None.

    Raises:
        TypeError: If the config parameter is not of the expected type.
        ValueError: If the config parameter does not contain the required hidden_size attribute.
    """
    super().__init__()
    self.seq_relationship = nn.Linear(config.hidden_size, 2)

mindnlp.transformers.models.nezha.nezha.NezhaOnlyNSPHead.forward(pooled_output)

Constructs the NSP (Next Sentence Prediction) head for the Nezha model.

PARAMETER DESCRIPTION
self

An instance of the NezhaOnlyNSPHead class.

TYPE: NezhaOnlyNSPHead

pooled_output

The pooled output tensor of shape (batch_size, hidden_size). The pooled output is typically obtained by applying pooling operations (e.g., mean pooling, max pooling) over the sequence-level representations of the input tokens. It serves as the input to the NSP head.

TYPE: Tensor

RETURNS DESCRIPTION

None.
Note

The NSP head is responsible for predicting whether two input sentences are consecutive or not. It takes the pooled output tensor from the Nezha model and computes the sequence relationship score. The sequence relationship score is used to determine if the two input sentences are consecutive or not.

Source code in mindnlp/transformers/models/nezha/nezha.py 
1086
1087
1088
1089
1090
1091
1092
1093
1094
1095
1096
1097
1098
1099
1100
1101
1102
1103
1104
1105
1106
1107
1108
def forward(self, pooled_output):
    """
    Constructs the NSP (Next Sentence Prediction) head for the Nezha model.

    Args:
        self (NezhaOnlyNSPHead): An instance of the NezhaOnlyNSPHead class.
        pooled_output (torch.Tensor): The pooled output tensor of shape (batch_size, hidden_size).
            The pooled output is typically obtained by applying pooling operations (e.g., mean pooling, max pooling)
            over the sequence-level representations of the input tokens. It serves as the input to the NSP head.

    Returns:
        None.

    Raises:
        None.

    Note:
        The NSP head is responsible for predicting whether two input sentences are consecutive or not.
        It takes the pooled output tensor from the Nezha model and computes the sequence relationship score.
        The sequence relationship score is used to determine if the two input sentences are consecutive or not.
    """
    seq_relationship_score = self.seq_relationship(pooled_output)
    return seq_relationship_score

mindnlp.transformers.models.nezha.nezha.NezhaOutput

Bases: Module

Nezha Output

Source code in mindnlp/transformers/models/nezha/nezha.py 
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
class NezhaOutput(nn.Module):
    """Nezha Output"""
    def __init__(self, config):
        """
        Initializes a new instance of the NezhaOutput class.

        Args:
            self: The object instance.
            config:
                An instance of the configuration class containing the model's configuration parameters.

                - Type: Any
                - Purpose: Specifies the configuration settings for the NezhaOutput instance.
                - Restrictions: None

        Returns:
            None

        Raises:
            None
        """
        super().__init__()
        self.dense = nn.Linear(config.intermediate_size, config.hidden_size)
        self.LayerNorm = nn.LayerNorm((config.hidden_size,), eps=config.layer_norm_eps)
        self.dropout = nn.Dropout(p=config.hidden_dropout_prob)

    def forward(self, hidden_states, input_tensor):
        """
        Constructs the output of the Nezha model by applying a series of operations on the hidden states and input tensor.

        Args:
            self (NezhaOutput): An instance of the NezhaOutput class.
            hidden_states (Tensor): The hidden states of the model.
                It should have dimensions (batch_size, sequence_length, hidden_size).
            input_tensor (Tensor): The input tensor to be added to the hidden states.
                It should have the same dimensions as hidden_states.

        Returns:
            None.

        Raises:
            None.
        """
        hidden_states = self.dense(hidden_states)
        hidden_states = self.dropout(hidden_states)
        hidden_states = self.LayerNorm(hidden_states + input_tensor)
        return hidden_states

mindnlp.transformers.models.nezha.nezha.NezhaOutput.__init__(config)

Initializes a new instance of the NezhaOutput class.

PARAMETER DESCRIPTION
self

The object instance.
config

An instance of the configuration class containing the model's configuration parameters.

  • Type: Any
  • Purpose: Specifies the configuration settings for the NezhaOutput instance.
  • Restrictions: None
RETURNS DESCRIPTION

None
Source code in mindnlp/transformers/models/nezha/nezha.py 
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
def __init__(self, config):
    """
    Initializes a new instance of the NezhaOutput class.

    Args:
        self: The object instance.
        config:
            An instance of the configuration class containing the model's configuration parameters.

            - Type: Any
            - Purpose: Specifies the configuration settings for the NezhaOutput instance.
            - Restrictions: None

    Returns:
        None

    Raises:
        None
    """
    super().__init__()
    self.dense = nn.Linear(config.intermediate_size, config.hidden_size)
    self.LayerNorm = nn.LayerNorm((config.hidden_size,), eps=config.layer_norm_eps)
    self.dropout = nn.Dropout(p=config.hidden_dropout_prob)

mindnlp.transformers.models.nezha.nezha.NezhaOutput.forward(hidden_states, input_tensor)

Constructs the output of the Nezha model by applying a series of operations on the hidden states and input tensor.

PARAMETER DESCRIPTION
self

An instance of the NezhaOutput class.

TYPE: NezhaOutput

hidden_states

The hidden states of the model. It should have dimensions (batch_size, sequence_length, hidden_size).

TYPE: Tensor

input_tensor

The input tensor to be added to the hidden states. It should have the same dimensions as hidden_states.

TYPE: Tensor

RETURNS DESCRIPTION

None.
Source code in mindnlp/transformers/models/nezha/nezha.py 
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
def forward(self, hidden_states, input_tensor):
    """
    Constructs the output of the Nezha model by applying a series of operations on the hidden states and input tensor.

    Args:
        self (NezhaOutput): An instance of the NezhaOutput class.
        hidden_states (Tensor): The hidden states of the model.
            It should have dimensions (batch_size, sequence_length, hidden_size).
        input_tensor (Tensor): The input tensor to be added to the hidden states.
            It should have the same dimensions as hidden_states.

    Returns:
        None.

    Raises:
        None.
    """
    hidden_states = self.dense(hidden_states)
    hidden_states = self.dropout(hidden_states)
    hidden_states = self.LayerNorm(hidden_states + input_tensor)
    return hidden_states

mindnlp.transformers.models.nezha.nezha.NezhaPooler

Bases: Module

Nezha Pooler

Source code in mindnlp/transformers/models/nezha/nezha.py 
880
881
882
883
884
885
886
887
888
889
890
891
892
893
894
895
896
897
898
899
900
901
902
903
904
905
906
907
908
909
910
911
912
913
914
915
916
917
918
919
920
921
922
923
class NezhaPooler(nn.Module):
    """Nezha Pooler"""
    def __init__(self, config):
        """
        Initializes the NezhaPooler class.

        Args:
            self (NezhaPooler): The instance of the NezhaPooler class.
            config (object): An object containing configuration parameters for the NezhaPooler.
                This parameter is used to configure the dense layer and activation function.
                It should have a property 'hidden_size' to specify the size of the hidden layer.

        Returns:
            None.

        Raises:
            None.
        """
        super().__init__()
        self.dense = nn.Linear(config.hidden_size, config.hidden_size)
        self.activation = nn.Tanh()

    def forward(self, hidden_states):
        """
        Constructs the pooled output from the given hidden states.

        Args:
            self (NezhaPooler): An instance of the NezhaPooler class.
            hidden_states (Tensor): A tensor containing the hidden states.
                Shape: (batch_size, sequence_length, hidden_size)

        Returns:
            Tensor: A tensor representing the pooled output.
                Shape: (batch_size, hidden_size)

        Raises:
            None.
        """
        # We "pool" the model by simply taking the hidden state corresponding
        # to the first token.
        first_token_tensor = hidden_states[:, 0]
        pooled_output = self.dense(first_token_tensor)
        pooled_output = self.activation(pooled_output)
        return pooled_output

mindnlp.transformers.models.nezha.nezha.NezhaPooler.__init__(config)

Initializes the NezhaPooler class.

PARAMETER DESCRIPTION
self

The instance of the NezhaPooler class.

TYPE: NezhaPooler

config

An object containing configuration parameters for the NezhaPooler. This parameter is used to configure the dense layer and activation function. It should have a property 'hidden_size' to specify the size of the hidden layer.

TYPE: object

RETURNS DESCRIPTION

None.
Source code in mindnlp/transformers/models/nezha/nezha.py 
882
883
884
885
886
887
888
889
890
891
892
893
894
895
896
897
898
899
900
def __init__(self, config):
    """
    Initializes the NezhaPooler class.

    Args:
        self (NezhaPooler): The instance of the NezhaPooler class.
        config (object): An object containing configuration parameters for the NezhaPooler.
            This parameter is used to configure the dense layer and activation function.
            It should have a property 'hidden_size' to specify the size of the hidden layer.

    Returns:
        None.

    Raises:
        None.
    """
    super().__init__()
    self.dense = nn.Linear(config.hidden_size, config.hidden_size)
    self.activation = nn.Tanh()

mindnlp.transformers.models.nezha.nezha.NezhaPooler.forward(hidden_states)

Constructs the pooled output from the given hidden states.

PARAMETER DESCRIPTION
self

An instance of the NezhaPooler class.

TYPE: NezhaPooler

hidden_states

A tensor containing the hidden states. Shape: (batch_size, sequence_length, hidden_size)

TYPE: Tensor

RETURNS DESCRIPTION
Tensor

A tensor representing the pooled output. Shape: (batch_size, hidden_size)
Source code in mindnlp/transformers/models/nezha/nezha.py 
902
903
904
905
906
907
908
909
910
911
912
913
914
915
916
917
918
919
920
921
922
923
def forward(self, hidden_states):
    """
    Constructs the pooled output from the given hidden states.

    Args:
        self (NezhaPooler): An instance of the NezhaPooler class.
        hidden_states (Tensor): A tensor containing the hidden states.
            Shape: (batch_size, sequence_length, hidden_size)

    Returns:
        Tensor: A tensor representing the pooled output.
            Shape: (batch_size, hidden_size)

    Raises:
        None.
    """
    # We "pool" the model by simply taking the hidden state corresponding
    # to the first token.
    first_token_tensor = hidden_states[:, 0]
    pooled_output = self.dense(first_token_tensor)
    pooled_output = self.activation(pooled_output)
    return pooled_output

mindnlp.transformers.models.nezha.nezha.NezhaPreTrainedModel

Bases: PreTrainedModel

An abstract class to handle weights initialization and a simple interface for downloading and loading pretrained models.

Source code in mindnlp/transformers/models/nezha/nezha.py 
1156
1157
1158
1159
1160
1161
1162
1163
1164
1165
1166
1167
1168
1169
1170
1171
1172
1173
1174
1175
1176
1177
1178
1179
1180
1181
1182
1183
1184
1185
1186
1187
1188
1189
1190
1191
1192
1193
1194
1195
1196
1197
1198
1199
1200
1201
1202
1203
1204
1205
1206
1207
1208
1209
1210
1211
1212
1213
1214
1215
1216
1217
1218
1219
1220
1221
1222
1223
1224
1225
1226
1227
1228
1229
1230
1231
1232
1233
1234
1235
1236
1237
1238
1239
1240
1241
1242
1243
1244
1245
1246
1247
1248
1249
1250
1251
1252
1253
1254
1255
1256
1257
1258
1259
1260
1261
1262
1263
1264
1265
1266
1267
1268
1269
1270
1271
class NezhaPreTrainedModel(PreTrainedModel):
    """
    An abstract class to handle weights initialization and a simple interface for downloading and loading pretrained
    models.
    """
    config_class = NezhaConfig
    base_model_prefix = "nezha"
    supports_gradient_checkpointing = True

    _keys_to_ignore_on_load_missing = [r"positions_encoding"]

    def _init_weights(self, cell):
        """Initialize the weights"""
        if isinstance(cell, nn.Linear):
            cell.weight.set_data(initializer(Normal(self.config.initializer_range),
                                                    cell.weight.shape, cell.weight.dtype))
            if cell.bias:
                cell.bias.set_data(initializer('zeros', cell.bias.shape, cell.bias.dtype))
        elif isinstance(cell, nn.Embedding):
            weight = initializer(Normal(self.config.initializer_range),
                                                 cell.weight.shape,
                                                 cell.weight.dtype)
            if cell.padding_idx is not None:
                weight[cell.padding_idx] = 0
            cell.weight.set_data(weight)
        elif isinstance(cell, nn.LayerNorm):
            cell.weight.set_data(initializer('ones', cell.weight.shape, cell.weight.dtype))
            cell.bias.set_data(initializer('zeros', cell.bias.shape, cell.bias.dtype))

    # TODO
    def get_input_embeddings(self):
        """
        Method to get the input embeddings for NezhaPreTrainedModel.

        Args:
            self: NezhaPreTrainedModel object. The instance of the NezhaPreTrainedModel class.

        Returns:
            None.

        Raises:
            None.
        """

    # TODO
    def get_position_embeddings(self):
        """
        This method retrieves the position embeddings for NezhaPreTrainedModel.

        Args:
            self: The instance of the NezhaPreTrainedModel class.

        Returns:
            None.

        Raises:
            None.
        """

    # TODO
    def resize_position_embeddings(self):
        """
        Method to resize the position embeddings of the NezhaPreTrainedModel.

        Args:
            self: NezhaPreTrainedModel, The instance of the NezhaPreTrainedModel class.
                This parameter is used to access and modify the position embeddings of the model.

        Returns:
            None: This method does not return any value. It modifies the position embeddings in place.

        Raises:
            None.
        """

    # TODO
    def set_input_embeddings(self):
        """
        This method sets the input embeddings for the NezhaPreTrainedModel.

        Args:
            self:
                The instance of the NezhaPreTrainedModel class.

                - Type: NezhaPreTrainedModel
                - Purpose: To access and modify the attributes and methods of the NezhaPreTrainedModel instance.

        Returns:
            None.

        Raises:
            None.
        """

    # TODO
    def post_init(self):
        """
        This method is part of the NezhaPreTrainedModel class and is called 'post_init'.

        Args:
            self: An instance of the NezhaPreTrainedModel class.

        Returns:
            None.

        Raises:
            None.

        Description:
            This method is a placeholder and does not perform any specific operations.
            It is called automatically after the initialization of an instance of the NezhaPreTrainedModel class.
            It can be overridden in child classes to add custom initialization logic or perform additional setup steps.

        Note that the 'self' parameter is automatically passed to the method and does not need to be provided explicitly
        when calling the method.
        """

mindnlp.transformers.models.nezha.nezha.NezhaPreTrainedModel.get_input_embeddings()

Method to get the input embeddings for NezhaPreTrainedModel.

PARAMETER DESCRIPTION
self

NezhaPreTrainedModel object. The instance of the NezhaPreTrainedModel class.
RETURNS DESCRIPTION

None.
Source code in mindnlp/transformers/models/nezha/nezha.py 
1186
1187
1188
1189
1190
1191
1192
1193
1194
1195
1196
1197
1198
def get_input_embeddings(self):
    """
    Method to get the input embeddings for NezhaPreTrainedModel.

    Args:
        self: NezhaPreTrainedModel object. The instance of the NezhaPreTrainedModel class.

    Returns:
        None.

    Raises:
        None.
    """

mindnlp.transformers.models.nezha.nezha.NezhaPreTrainedModel.get_position_embeddings()

This method retrieves the position embeddings for NezhaPreTrainedModel.

PARAMETER DESCRIPTION
self

The instance of the NezhaPreTrainedModel class.
RETURNS DESCRIPTION

None.
Source code in mindnlp/transformers/models/nezha/nezha.py 
1201
1202
1203
1204
1205
1206
1207
1208
1209
1210
1211
1212
1213
def get_position_embeddings(self):
    """
    This method retrieves the position embeddings for NezhaPreTrainedModel.

    Args:
        self: The instance of the NezhaPreTrainedModel class.

    Returns:
        None.

    Raises:
        None.
    """

mindnlp.transformers.models.nezha.nezha.NezhaPreTrainedModel.post_init()

This method is part of the NezhaPreTrainedModel class and is called 'post_init'.

PARAMETER DESCRIPTION
self

An instance of the NezhaPreTrainedModel class.
RETURNS DESCRIPTION

None.
Description

This method is a placeholder and does not perform any specific operations. It is called automatically after the initialization of an instance of the NezhaPreTrainedModel class. It can be overridden in child classes to add custom initialization logic or perform additional setup steps.

Note that the 'self' parameter is automatically passed to the method and does not need to be provided explicitly when calling the method.

Source code in mindnlp/transformers/models/nezha/nezha.py 
1251
1252
1253
1254
1255
1256
1257
1258
1259
1260
1261
1262
1263
1264
1265
1266
1267
1268
1269
1270
1271
def post_init(self):
    """
    This method is part of the NezhaPreTrainedModel class and is called 'post_init'.

    Args:
        self: An instance of the NezhaPreTrainedModel class.

    Returns:
        None.

    Raises:
        None.

    Description:
        This method is a placeholder and does not perform any specific operations.
        It is called automatically after the initialization of an instance of the NezhaPreTrainedModel class.
        It can be overridden in child classes to add custom initialization logic or perform additional setup steps.

    Note that the 'self' parameter is automatically passed to the method and does not need to be provided explicitly
    when calling the method.
    """

mindnlp.transformers.models.nezha.nezha.NezhaPreTrainedModel.resize_position_embeddings()

Method to resize the position embeddings of the NezhaPreTrainedModel.

PARAMETER DESCRIPTION
self

NezhaPreTrainedModel, The instance of the NezhaPreTrainedModel class. This parameter is used to access and modify the position embeddings of the model.
RETURNS DESCRIPTION
None

This method does not return any value. It modifies the position embeddings in place.
Source code in mindnlp/transformers/models/nezha/nezha.py 
1216
1217
1218
1219
1220
1221
1222
1223
1224
1225
1226
1227
1228
1229
def resize_position_embeddings(self):
    """
    Method to resize the position embeddings of the NezhaPreTrainedModel.

    Args:
        self: NezhaPreTrainedModel, The instance of the NezhaPreTrainedModel class.
            This parameter is used to access and modify the position embeddings of the model.

    Returns:
        None: This method does not return any value. It modifies the position embeddings in place.

    Raises:
        None.
    """

mindnlp.transformers.models.nezha.nezha.NezhaPreTrainedModel.set_input_embeddings()

This method sets the input embeddings for the NezhaPreTrainedModel.

PARAMETER DESCRIPTION
self

The instance of the NezhaPreTrainedModel class.

  • Type: NezhaPreTrainedModel
  • Purpose: To access and modify the attributes and methods of the NezhaPreTrainedModel instance.
RETURNS DESCRIPTION

None.
Source code in mindnlp/transformers/models/nezha/nezha.py 
1232
1233
1234
1235
1236
1237
1238
1239
1240
1241
1242
1243
1244
1245
1246
1247
1248
def set_input_embeddings(self):
    """
    This method sets the input embeddings for the NezhaPreTrainedModel.

    Args:
        self:
            The instance of the NezhaPreTrainedModel class.

            - Type: NezhaPreTrainedModel
            - Purpose: To access and modify the attributes and methods of the NezhaPreTrainedModel instance.

    Returns:
        None.

    Raises:
        None.
    """

mindnlp.transformers.models.nezha.nezha.NezhaPreTrainingHeads

Bases: Module

Nezha PreTrainingHeads

Source code in mindnlp/transformers/models/nezha/nezha.py 
1111
1112
1113
1114
1115
1116
1117
1118
1119
1120
1121
1122
1123
1124
1125
1126
1127
1128
1129
1130
1131
1132
1133
1134
1135
1136
1137
1138
1139
1140
1141
1142
1143
1144
1145
1146
1147
1148
1149
1150
1151
1152
1153
class NezhaPreTrainingHeads(nn.Module):
    """Nezha PreTrainingHeads"""
    def __init__(self, config):
        """
        Initializes the NezhaPreTrainingHeads class.

        Args:
            self (NezhaPreTrainingHeads): The instance of the NezhaPreTrainingHeads class.
            config: Configuration object containing settings for the NezhaPreTrainingHeads.
                It is expected to be a dictionary-like object.

        Returns:
            None.

        Raises:
            None.
        """
        super().__init__()
        self.predictions = NezhaLMPredictionHead(config)
        self.seq_relationship = nn.Linear(config.hidden_size, 2)

    def forward(self, sequence_output, pooled_output):
        """
        This method forwards Nezha pre-training heads.

        Args:
            self (object): The instance of the NezhaPreTrainingHeads class.
            sequence_output (object): The output of the sequence.
            pooled_output (object): The pooled output.

        Returns:
            tuple:
                A tuple containing 'prediction_scores' and 'seq_relationship_score'.

                - prediction_scores (object): The prediction scores based on the sequence_output.
                - seq_relationship_score (object): The sequence relationship score based on the pooled_output.

        Raises:
            None
        """
        prediction_scores = self.predictions(sequence_output)
        seq_relationship_score = self.seq_relationship(pooled_output)
        return prediction_scores, seq_relationship_score

mindnlp.transformers.models.nezha.nezha.NezhaPreTrainingHeads.__init__(config)

Initializes the NezhaPreTrainingHeads class.

PARAMETER DESCRIPTION
self

The instance of the NezhaPreTrainingHeads class.

TYPE: NezhaPreTrainingHeads

config

Configuration object containing settings for the NezhaPreTrainingHeads. It is expected to be a dictionary-like object.
RETURNS DESCRIPTION

None.
Source code in mindnlp/transformers/models/nezha/nezha.py 
1113
1114
1115
1116
1117
1118
1119
1120
1121
1122
1123
1124
1125
1126
1127
1128
1129
1130
def __init__(self, config):
    """
    Initializes the NezhaPreTrainingHeads class.

    Args:
        self (NezhaPreTrainingHeads): The instance of the NezhaPreTrainingHeads class.
        config: Configuration object containing settings for the NezhaPreTrainingHeads.
            It is expected to be a dictionary-like object.

    Returns:
        None.

    Raises:
        None.
    """
    super().__init__()
    self.predictions = NezhaLMPredictionHead(config)
    self.seq_relationship = nn.Linear(config.hidden_size, 2)

mindnlp.transformers.models.nezha.nezha.NezhaPreTrainingHeads.forward(sequence_output, pooled_output)

This method forwards Nezha pre-training heads.

PARAMETER DESCRIPTION
self

The instance of the NezhaPreTrainingHeads class.

TYPE: object

sequence_output

The output of the sequence.

TYPE: object

pooled_output

The pooled output.

TYPE: object

RETURNS DESCRIPTION
tuple

A tuple containing 'prediction_scores' and 'seq_relationship_score'.

  • prediction_scores (object): The prediction scores based on the sequence_output.
  • seq_relationship_score (object): The sequence relationship score based on the pooled_output.
Source code in mindnlp/transformers/models/nezha/nezha.py 
1132
1133
1134
1135
1136
1137
1138
1139
1140
1141
1142
1143
1144
1145
1146
1147
1148
1149
1150
1151
1152
1153
def forward(self, sequence_output, pooled_output):
    """
    This method forwards Nezha pre-training heads.

    Args:
        self (object): The instance of the NezhaPreTrainingHeads class.
        sequence_output (object): The output of the sequence.
        pooled_output (object): The pooled output.

    Returns:
        tuple:
            A tuple containing 'prediction_scores' and 'seq_relationship_score'.

            - prediction_scores (object): The prediction scores based on the sequence_output.
            - seq_relationship_score (object): The sequence relationship score based on the pooled_output.

    Raises:
        None
    """
    prediction_scores = self.predictions(sequence_output)
    seq_relationship_score = self.seq_relationship(pooled_output)
    return prediction_scores, seq_relationship_score

mindnlp.transformers.models.nezha.nezha.NezhaPredictionHeadTransform

Bases: Module

Nezha Predicton Head Transform

Source code in mindnlp/transformers/models/nezha/nezha.py 
926
927
928
929
930
931
932
933
934
935
936
937
938
939
940
941
942
943
944
945
946
947
948
949
950
951
952
953
954
955
956
957
958
959
960
961
962
963
964
965
966
967
968
969
970
971
972
class NezhaPredictionHeadTransform(nn.Module):
    """Nezha Predicton Head Transform"""
    def __init__(self, config):
        """
        Initializes the NezhaPredictionHeadTransform class.

        Args:
            self: The object instance.
            config:
                An instance of the configuration class that contains the parameters for the head transformation.

                - Type: Any
                - Purpose: Specifies the configuration for the head transformation.
                - Restrictions: None

        Returns:
            None

        Raises:
            None
        """
        super().__init__()
        self.dense = nn.Linear(config.hidden_size, config.hidden_size)
        if isinstance(config.hidden_act, str):
            self.transform_act_fn = ACT2FN[config.hidden_act]
        else:
            self.transform_act_fn = config.hidden_act
        self.LayerNorm = nn.LayerNorm((config.hidden_size,), eps=config.layer_norm_eps)

    def forward(self, hidden_states):
        """
        Constructs the NezhaPredictionHeadTransform.

        Args:
            self: An instance of the NezhaPredictionHeadTransform class.
            hidden_states (tensor): The hidden states to be transformed.

        Returns:
            None

        Raises:
            None
        """
        hidden_states = self.dense(hidden_states)
        hidden_states = self.transform_act_fn(hidden_states)
        hidden_states = self.LayerNorm(hidden_states)
        return hidden_states

mindnlp.transformers.models.nezha.nezha.NezhaPredictionHeadTransform.__init__(config)

Initializes the NezhaPredictionHeadTransform class.

PARAMETER DESCRIPTION
self

The object instance.
config

An instance of the configuration class that contains the parameters for the head transformation.

  • Type: Any
  • Purpose: Specifies the configuration for the head transformation.
  • Restrictions: None
RETURNS DESCRIPTION

None
Source code in mindnlp/transformers/models/nezha/nezha.py 
928
929
930
931
932
933
934
935
936
937
938
939
940
941
942
943
944
945
946
947
948
949
950
951
952
953
def __init__(self, config):
    """
    Initializes the NezhaPredictionHeadTransform class.

    Args:
        self: The object instance.
        config:
            An instance of the configuration class that contains the parameters for the head transformation.

            - Type: Any
            - Purpose: Specifies the configuration for the head transformation.
            - Restrictions: None

    Returns:
        None

    Raises:
        None
    """
    super().__init__()
    self.dense = nn.Linear(config.hidden_size, config.hidden_size)
    if isinstance(config.hidden_act, str):
        self.transform_act_fn = ACT2FN[config.hidden_act]
    else:
        self.transform_act_fn = config.hidden_act
    self.LayerNorm = nn.LayerNorm((config.hidden_size,), eps=config.layer_norm_eps)

mindnlp.transformers.models.nezha.nezha.NezhaPredictionHeadTransform.forward(hidden_states)

Constructs the NezhaPredictionHeadTransform.

PARAMETER DESCRIPTION
self

An instance of the NezhaPredictionHeadTransform class.
hidden_states

The hidden states to be transformed.

TYPE: tensor

RETURNS DESCRIPTION

None
Source code in mindnlp/transformers/models/nezha/nezha.py 
955
956
957
958
959
960
961
962
963
964
965
966
967
968
969
970
971
972
def forward(self, hidden_states):
    """
    Constructs the NezhaPredictionHeadTransform.

    Args:
        self: An instance of the NezhaPredictionHeadTransform class.
        hidden_states (tensor): The hidden states to be transformed.

    Returns:
        None

    Raises:
        None
    """
    hidden_states = self.dense(hidden_states)
    hidden_states = self.transform_act_fn(hidden_states)
    hidden_states = self.LayerNorm(hidden_states)
    return hidden_states

mindnlp.transformers.models.nezha.nezha.NezhaRelativePositionsEncoding

Bases: Module

Implement the Functional Relative Position Encoding

Source code in mindnlp/transformers/models/nezha/nezha.py 
 47
 48
 49
 50
 51
 52
 53
 54
 55
 56
 57
 58
 59
 60
 61
 62
 63
 64
 65
 66
 67
 68
 69
 70
 71
 72
 73
 74
 75
 76
 77
 78
 79
 80
 81
 82
 83
 84
 85
 86
 87
 88
 89
 90
 91
 92
 93
 94
 95
 96
 97
 98
 99
100
101
102
103
104
105
106
107
108
109
class NezhaRelativePositionsEncoding(nn.Module):
    """Implement the Functional Relative Position Encoding"""
    def __init__(self, length, depth, max_relative_position=127):
        """
        Initializes the NezhaRelativePositionsEncoding object with the specified parameters.

        Args:
            self (object): The instance of the NezhaRelativePositionsEncoding class.
            length (int): The length of the input sequence.
            depth (int): The depth of the embeddings table.
            max_relative_position (int, optional): The maximum allowed relative position. Defaults to 127.

        Returns:
            None.

        Raises:
            ValueError: If length or depth is not an integer.
            ValueError: If max_relative_position is not an integer.
            ValueError: If max_relative_position is less than 1.
        """
        super().__init__()
        vocab_size = max_relative_position * 2 + 1
        range_vec = ops.arange(length)
        range_mat = ops.tile(range_vec, (length, 1))
        distance_mat = range_mat - ops.t(range_mat)
        distance_mat_clipped = ops.clamp(distance_mat, -max_relative_position,
                                        max_relative_position)
        final_mat = distance_mat_clipped + max_relative_position

        # TODO: use numpy to avoid setitem(mindspore not support complete `view`.)
        embeddings_table = ops.zeros((vocab_size, depth))
        # position = ops.arange(0, vocab_size, dtype=mindspore.float32).expand_dims(1)
        # div_term = ops.exp(ops.arange(0, depth, 2).astype(mindspore.float32) * (-math.log(10000.0)) / depth)
        # embeddings_table[:, 0::2] = ops.sin(position * div_term)
        # embeddings_table[:, 1::2] = ops.cos(position * div_term)

        flat_relative_positions_matrix = final_mat.view(-1)
        on_value, off_value = Tensor(1.0, mindspore.float32), Tensor(0.0, mindspore.float32)
        one_hot_relative_positions_matrix = ops.one_hot(
            flat_relative_positions_matrix, vocab_size, on_value, off_value, axis=-1
        ).astype(mindspore.float32)
        positions_encoding = ops.matmul(one_hot_relative_positions_matrix, embeddings_table)
        my_shape = list(final_mat.shape)
        my_shape.append(depth)
        self.positions_encoding = Parameter(positions_encoding.view(tuple(my_shape)), requires_grad=False)

    def forward(self, length):
        """
        Constructs a relative positions encoding matrix of specified length.

        Args:
            self (NezhaRelativePositionsEncoding): The instance of the NezhaRelativePositionsEncoding class.
            length (int): The length of the positions encoding matrix to be forwarded.
                Must be a non-negative integer.

        Returns:
            None: The method modifies the internal state of the NezhaRelativePositionsEncoding instance.

        Raises:
            IndexError: If the length provided is greater than the dimensions of the positions_encoding matrix.
            ValueError: If the length provided is a negative integer.
        """
        return self.positions_encoding[:length, :length, :]

mindnlp.transformers.models.nezha.nezha.NezhaRelativePositionsEncoding.__init__(length, depth, max_relative_position=127)

Initializes the NezhaRelativePositionsEncoding object with the specified parameters.

PARAMETER DESCRIPTION
self

The instance of the NezhaRelativePositionsEncoding class.

TYPE: object

length

The length of the input sequence.

TYPE: int

depth

The depth of the embeddings table.

TYPE: int

max_relative_position

The maximum allowed relative position. Defaults to 127.

TYPE: int DEFAULT: 127

RETURNS DESCRIPTION

None.
RAISES DESCRIPTION
ValueError

If length or depth is not an integer.
ValueError

If max_relative_position is not an integer.
ValueError

If max_relative_position is less than 1.
Source code in mindnlp/transformers/models/nezha/nezha.py 
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
def __init__(self, length, depth, max_relative_position=127):
    """
    Initializes the NezhaRelativePositionsEncoding object with the specified parameters.

    Args:
        self (object): The instance of the NezhaRelativePositionsEncoding class.
        length (int): The length of the input sequence.
        depth (int): The depth of the embeddings table.
        max_relative_position (int, optional): The maximum allowed relative position. Defaults to 127.

    Returns:
        None.

    Raises:
        ValueError: If length or depth is not an integer.
        ValueError: If max_relative_position is not an integer.
        ValueError: If max_relative_position is less than 1.
    """
    super().__init__()
    vocab_size = max_relative_position * 2 + 1
    range_vec = ops.arange(length)
    range_mat = ops.tile(range_vec, (length, 1))
    distance_mat = range_mat - ops.t(range_mat)
    distance_mat_clipped = ops.clamp(distance_mat, -max_relative_position,
                                    max_relative_position)
    final_mat = distance_mat_clipped + max_relative_position

    # TODO: use numpy to avoid setitem(mindspore not support complete `view`.)
    embeddings_table = ops.zeros((vocab_size, depth))
    # position = ops.arange(0, vocab_size, dtype=mindspore.float32).expand_dims(1)
    # div_term = ops.exp(ops.arange(0, depth, 2).astype(mindspore.float32) * (-math.log(10000.0)) / depth)
    # embeddings_table[:, 0::2] = ops.sin(position * div_term)
    # embeddings_table[:, 1::2] = ops.cos(position * div_term)

    flat_relative_positions_matrix = final_mat.view(-1)
    on_value, off_value = Tensor(1.0, mindspore.float32), Tensor(0.0, mindspore.float32)
    one_hot_relative_positions_matrix = ops.one_hot(
        flat_relative_positions_matrix, vocab_size, on_value, off_value, axis=-1
    ).astype(mindspore.float32)
    positions_encoding = ops.matmul(one_hot_relative_positions_matrix, embeddings_table)
    my_shape = list(final_mat.shape)
    my_shape.append(depth)
    self.positions_encoding = Parameter(positions_encoding.view(tuple(my_shape)), requires_grad=False)

mindnlp.transformers.models.nezha.nezha.NezhaRelativePositionsEncoding.forward(length)

Constructs a relative positions encoding matrix of specified length.

PARAMETER DESCRIPTION
self

The instance of the NezhaRelativePositionsEncoding class.

TYPE: NezhaRelativePositionsEncoding

length

The length of the positions encoding matrix to be forwarded. Must be a non-negative integer.

TYPE: int

RETURNS DESCRIPTION
None

The method modifies the internal state of the NezhaRelativePositionsEncoding instance.
RAISES DESCRIPTION
IndexError

If the length provided is greater than the dimensions of the positions_encoding matrix.
ValueError

If the length provided is a negative integer.
Source code in mindnlp/transformers/models/nezha/nezha.py 
 93
 94
 95
 96
 97
 98
 99
100
101
102
103
104
105
106
107
108
109
def forward(self, length):
    """
    Constructs a relative positions encoding matrix of specified length.

    Args:
        self (NezhaRelativePositionsEncoding): The instance of the NezhaRelativePositionsEncoding class.
        length (int): The length of the positions encoding matrix to be forwarded.
            Must be a non-negative integer.

    Returns:
        None: The method modifies the internal state of the NezhaRelativePositionsEncoding instance.

    Raises:
        IndexError: If the length provided is greater than the dimensions of the positions_encoding matrix.
        ValueError: If the length provided is a negative integer.
    """
    return self.positions_encoding[:length, :length, :]

mindnlp.transformers.models.nezha.nezha.NezhaSelfAttention

Bases: Module

Self attention layer for NEZHA

Source code in mindnlp/transformers/models/nezha/nezha.py 
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
class NezhaSelfAttention(nn.Module):
    """Self attention layer for NEZHA"""
    def __init__(self, config):
        '''
        This method initializes the NezhaSelfAttention class.

        Args:
            self: The instance of the class.
            config: An object containing the configuration parameters for the self-attention mechanism.
                It should include the following attributes:

                - hidden_size (int): The size of the hidden layers.
                - num_attention_heads (int): The number of attention heads.
                - attention_probs_dropout_prob (float): The dropout probability for the attention probabilities.
                - max_position_embeddings (int): The maximum number of positions for positional encodings.
                - max_relative_position (int): The maximum relative position for the relative positions encoding.
                - is_decoder (bool): Indicates whether the self-attention mechanism is used as part of a decoder.

        Returns:
            None.

        Raises:
            ValueError: If the hidden_size is not a multiple of the number of attention heads.
        '''
        super().__init__()
        if config.hidden_size % config.num_attention_heads != 0:
            raise ValueError(
                f"The hidden size ({config.hidden_size}) is not a multiple of the number of attention "
                f"heads ({config.num_attention_heads})"
            )

        self.num_attention_heads = config.num_attention_heads
        self.attention_head_size = int(config.hidden_size / config.num_attention_heads)
        self.all_head_size = self.num_attention_heads * self.attention_head_size

        self.query = nn.Linear(config.hidden_size, self.all_head_size)
        self.key = nn.Linear(config.hidden_size, self.all_head_size)
        self.value = nn.Linear(config.hidden_size, self.all_head_size)

        self.dropout = nn.Dropout(p=config.attention_probs_dropout_prob)
        self.relative_positions_encoding = NezhaRelativePositionsEncoding(
            length=config.max_position_embeddings,
            depth=self.attention_head_size,
            max_relative_position=config.max_relative_position,
        )
        self.is_decoder = config.is_decoder

    def transpose_for_scores(self, input_x: Tensor) -> Tensor:
        """transpose for scores"""
        new_x_shape = input_x.shape[:-1] + (self.num_attention_heads, self.attention_head_size)
        input_x = input_x.view(tuple(new_x_shape))
        return input_x.permute(0, 2, 1, 3)

    def forward(self, hidden_states, attention_mask = None, head_mask = None,
                  encoder_hidden_states = None, encoder_attention_mask = None,
                  past_key_value = None, output_attentions = False):
        """
        This method 'forward' is defined within the class 'NezhaSelfAttention' and is responsible for performing
        self-attention computations. It takes the following parameters:

        Args:
            self: The instance of the class.
            hidden_states: Tensor, required. The input tensor containing the hidden states for the
                self-attention mechanism.
            attention_mask: Tensor, optional. A 2D tensor providing the attention mask to be applied during
                self-attention computation. Default is None.
            head_mask: Tensor, optional. A 2D tensor representing the head mask for controlling which heads are
                active during self-attention. Default is None.
            encoder_hidden_states: Tensor, optional. The hidden states from the encoder if this is a cross-attention
                operation. Default is None.
            encoder_attention_mask: Tensor, optional. A 2D tensor providing the attention mask for encoder_hidden_states.
                Default is None.
            past_key_value: Tuple of Tensors, optional. The previous key and value tensors from the past self-attention
                computation. Default is None.
            output_attentions: Bool, optional. Flag indicating whether to output attention scores. Default is False.

        Returns:
            Tuple of Tensors or Tuple of (Tensor, Tensor, Tuple of Tensors):
                The output of the self-attention mechanism. If output_attentions is True, returns a tuple containing
                the context_layer and attention_probs. If self.is_decoder is True, the output also includes the
                past_key_value.

        Raises:
            ValueError: If the dimensions or types of the input tensors are incompatible.
            RuntimeError: If any runtime error occurs during the self-attention computation.
            AssertionError: If the conditions for past_key_value are not met.
        """
        mixed_query_layer = self.query(hidden_states)

        # If this is instantiated as a cross-attention module, the keys
        # and values come from an encoder; the attention mask needs to be
        # such that the encoder's padding tokens are not attended to.
        is_cross_attention = encoder_hidden_states is not None

        if is_cross_attention and past_key_value is not None:
            key_layer = past_key_value[0]
            value_layer = past_key_value[1]
            attention_mask = encoder_attention_mask
        elif is_cross_attention:
            key_layer = self.transpose_for_scores(self.key(encoder_hidden_states))
            value_layer = self.transpose_for_scores(self.value(encoder_hidden_states))
            attention_mask = encoder_attention_mask
        elif past_key_value is not None:
            key_layer = self.transpose_for_scores(self.key(hidden_states))
            value_layer = self.transpose_for_scores(self.value(hidden_states))
            key_layer = ops.cat([past_key_value[0], key_layer], axis=2)
            value_layer = ops.cat([past_key_value[1], value_layer], axis=2)
        else:
            key_layer = self.transpose_for_scores(self.key(hidden_states))
            value_layer = self.transpose_for_scores(self.value(hidden_states))

        query_layer = self.transpose_for_scores(mixed_query_layer)

        if self.is_decoder:
            past_key_value = (key_layer, value_layer)

        attention_scores = ops.matmul(query_layer, key_layer.swapaxes(-1, -2))

        batch_size, num_attention_heads, from_seq_length, to_seq_length = attention_scores.shape
        relations_keys = self.relative_positions_encoding(to_seq_length)
        query_layer_t = query_layer.permute(2, 0, 1, 3)

        query_layer_r = query_layer_t.view(
            from_seq_length, batch_size * num_attention_heads, self.attention_head_size
        )
        key_position_scores = ops.matmul(query_layer_r, relations_keys.permute(0, 2, 1))
        key_position_scores_r = key_position_scores.view(
            from_seq_length, batch_size, num_attention_heads, from_seq_length
        )
        key_position_scores_r_t = key_position_scores_r.permute(1, 2, 0, 3)
        attention_scores = attention_scores + key_position_scores_r_t
        attention_scores = attention_scores / math.sqrt(self.attention_head_size)

        if attention_mask is not None:
            # Apply the attention mask is (precomputed for all layers in NezhaModel forward() function)
            attention_scores = attention_scores + attention_mask

        # Normalize the attention scores to probabilities.
        attention_probs = ops.softmax(attention_scores, axis=-1)

        # This is actually dropping out entire tokens to attend to, which might
        # seem a bit unusual, but is taken from the original Transformer paper.
        attention_probs = self.dropout(attention_probs)

        if head_mask is not None:
            attention_probs = attention_probs * head_mask

        context_layer = ops.matmul(attention_probs, value_layer)
        relations_values = self.relative_positions_encoding(to_seq_length)
        attention_probs_t = attention_probs.permute(2, 0, 1, 3)
        attentions_probs_r = attention_probs_t.view(
            from_seq_length, batch_size * num_attention_heads, to_seq_length
        )
        value_position_scores = ops.matmul(attentions_probs_r, relations_values)
        value_position_scores_r = value_position_scores.view(
            from_seq_length, batch_size, num_attention_heads, self.attention_head_size
        )
        value_position_scores_r_t = value_position_scores_r.permute(1, 2, 0, 3)
        context_layer = context_layer + value_position_scores_r_t
        context_layer = context_layer.permute(0, 2, 1, 3)
        new_context_layer_shape = context_layer.shape[:-2] + (self.all_head_size,)
        context_layer = context_layer.view(new_context_layer_shape)
        outputs = (context_layer, attention_probs) if output_attentions else (context_layer,)

        if self.is_decoder:
            outputs = outputs + (past_key_value,)
        return outputs

mindnlp.transformers.models.nezha.nezha.NezhaSelfAttention.__init__(config)

This method initializes the NezhaSelfAttention class.

PARAMETER DESCRIPTION
self

The instance of the class.
config

An object containing the configuration parameters for the self-attention mechanism. It should include the following attributes:

  • hidden_size (int): The size of the hidden layers.
  • num_attention_heads (int): The number of attention heads.
  • attention_probs_dropout_prob (float): The dropout probability for the attention probabilities.
  • max_position_embeddings (int): The maximum number of positions for positional encodings.
  • max_relative_position (int): The maximum relative position for the relative positions encoding.
  • is_decoder (bool): Indicates whether the self-attention mechanism is used as part of a decoder.
RETURNS DESCRIPTION

None.
RAISES DESCRIPTION
ValueError

If the hidden_size is not a multiple of the number of attention heads.
Source code in mindnlp/transformers/models/nezha/nezha.py 
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
def __init__(self, config):
    '''
    This method initializes the NezhaSelfAttention class.

    Args:
        self: The instance of the class.
        config: An object containing the configuration parameters for the self-attention mechanism.
            It should include the following attributes:

            - hidden_size (int): The size of the hidden layers.
            - num_attention_heads (int): The number of attention heads.
            - attention_probs_dropout_prob (float): The dropout probability for the attention probabilities.
            - max_position_embeddings (int): The maximum number of positions for positional encodings.
            - max_relative_position (int): The maximum relative position for the relative positions encoding.
            - is_decoder (bool): Indicates whether the self-attention mechanism is used as part of a decoder.

    Returns:
        None.

    Raises:
        ValueError: If the hidden_size is not a multiple of the number of attention heads.
    '''
    super().__init__()
    if config.hidden_size % config.num_attention_heads != 0:
        raise ValueError(
            f"The hidden size ({config.hidden_size}) is not a multiple of the number of attention "
            f"heads ({config.num_attention_heads})"
        )

    self.num_attention_heads = config.num_attention_heads
    self.attention_head_size = int(config.hidden_size / config.num_attention_heads)
    self.all_head_size = self.num_attention_heads * self.attention_head_size

    self.query = nn.Linear(config.hidden_size, self.all_head_size)
    self.key = nn.Linear(config.hidden_size, self.all_head_size)
    self.value = nn.Linear(config.hidden_size, self.all_head_size)

    self.dropout = nn.Dropout(p=config.attention_probs_dropout_prob)
    self.relative_positions_encoding = NezhaRelativePositionsEncoding(
        length=config.max_position_embeddings,
        depth=self.attention_head_size,
        max_relative_position=config.max_relative_position,
    )
    self.is_decoder = config.is_decoder

mindnlp.transformers.models.nezha.nezha.NezhaSelfAttention.forward(hidden_states, attention_mask=None, head_mask=None, encoder_hidden_states=None, encoder_attention_mask=None, past_key_value=None, output_attentions=False)

This method 'forward' is defined within the class 'NezhaSelfAttention' and is responsible for performing self-attention computations. It takes the following parameters:

PARAMETER DESCRIPTION
self

The instance of the class.
hidden_states

Tensor, required. The input tensor containing the hidden states for the self-attention mechanism.
attention_mask

Tensor, optional. A 2D tensor providing the attention mask to be applied during self-attention computation. Default is None.

DEFAULT: None

head_mask

Tensor, optional. A 2D tensor representing the head mask for controlling which heads are active during self-attention. Default is None.

DEFAULT: None

encoder_hidden_states

Tensor, optional. The hidden states from the encoder if this is a cross-attention operation. Default is None.

DEFAULT: None

encoder_attention_mask

Tensor, optional. A 2D tensor providing the attention mask for encoder_hidden_states. Default is None.

DEFAULT: None

past_key_value

Tuple of Tensors, optional. The previous key and value tensors from the past self-attention computation. Default is None.

DEFAULT: None

output_attentions

Bool, optional. Flag indicating whether to output attention scores. Default is False.

DEFAULT: False

RETURNS DESCRIPTION

Tuple of Tensors or Tuple of (Tensor, Tensor, Tuple of Tensors): The output of the self-attention mechanism. If output_attentions is True, returns a tuple containing the context_layer and attention_probs. If self.is_decoder is True, the output also includes the past_key_value.
RAISES DESCRIPTION
ValueError

If the dimensions or types of the input tensors are incompatible.
RuntimeError

If any runtime error occurs during the self-attention computation.
AssertionError

If the conditions for past_key_value are not met.
Source code in mindnlp/transformers/models/nezha/nezha.py 
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
def forward(self, hidden_states, attention_mask = None, head_mask = None,
              encoder_hidden_states = None, encoder_attention_mask = None,
              past_key_value = None, output_attentions = False):
    """
    This method 'forward' is defined within the class 'NezhaSelfAttention' and is responsible for performing
    self-attention computations. It takes the following parameters:

    Args:
        self: The instance of the class.
        hidden_states: Tensor, required. The input tensor containing the hidden states for the
            self-attention mechanism.
        attention_mask: Tensor, optional. A 2D tensor providing the attention mask to be applied during
            self-attention computation. Default is None.
        head_mask: Tensor, optional. A 2D tensor representing the head mask for controlling which heads are
            active during self-attention. Default is None.
        encoder_hidden_states: Tensor, optional. The hidden states from the encoder if this is a cross-attention
            operation. Default is None.
        encoder_attention_mask: Tensor, optional. A 2D tensor providing the attention mask for encoder_hidden_states.
            Default is None.
        past_key_value: Tuple of Tensors, optional. The previous key and value tensors from the past self-attention
            computation. Default is None.
        output_attentions: Bool, optional. Flag indicating whether to output attention scores. Default is False.

    Returns:
        Tuple of Tensors or Tuple of (Tensor, Tensor, Tuple of Tensors):
            The output of the self-attention mechanism. If output_attentions is True, returns a tuple containing
            the context_layer and attention_probs. If self.is_decoder is True, the output also includes the
            past_key_value.

    Raises:
        ValueError: If the dimensions or types of the input tensors are incompatible.
        RuntimeError: If any runtime error occurs during the self-attention computation.
        AssertionError: If the conditions for past_key_value are not met.
    """
    mixed_query_layer = self.query(hidden_states)

    # If this is instantiated as a cross-attention module, the keys
    # and values come from an encoder; the attention mask needs to be
    # such that the encoder's padding tokens are not attended to.
    is_cross_attention = encoder_hidden_states is not None

    if is_cross_attention and past_key_value is not None:
        key_layer = past_key_value[0]
        value_layer = past_key_value[1]
        attention_mask = encoder_attention_mask
    elif is_cross_attention:
        key_layer = self.transpose_for_scores(self.key(encoder_hidden_states))
        value_layer = self.transpose_for_scores(self.value(encoder_hidden_states))
        attention_mask = encoder_attention_mask
    elif past_key_value is not None:
        key_layer = self.transpose_for_scores(self.key(hidden_states))
        value_layer = self.transpose_for_scores(self.value(hidden_states))
        key_layer = ops.cat([past_key_value[0], key_layer], axis=2)
        value_layer = ops.cat([past_key_value[1], value_layer], axis=2)
    else:
        key_layer = self.transpose_for_scores(self.key(hidden_states))
        value_layer = self.transpose_for_scores(self.value(hidden_states))

    query_layer = self.transpose_for_scores(mixed_query_layer)

    if self.is_decoder:
        past_key_value = (key_layer, value_layer)

    attention_scores = ops.matmul(query_layer, key_layer.swapaxes(-1, -2))

    batch_size, num_attention_heads, from_seq_length, to_seq_length = attention_scores.shape
    relations_keys = self.relative_positions_encoding(to_seq_length)
    query_layer_t = query_layer.permute(2, 0, 1, 3)

    query_layer_r = query_layer_t.view(
        from_seq_length, batch_size * num_attention_heads, self.attention_head_size
    )
    key_position_scores = ops.matmul(query_layer_r, relations_keys.permute(0, 2, 1))
    key_position_scores_r = key_position_scores.view(
        from_seq_length, batch_size, num_attention_heads, from_seq_length
    )
    key_position_scores_r_t = key_position_scores_r.permute(1, 2, 0, 3)
    attention_scores = attention_scores + key_position_scores_r_t
    attention_scores = attention_scores / math.sqrt(self.attention_head_size)

    if attention_mask is not None:
        # Apply the attention mask is (precomputed for all layers in NezhaModel forward() function)
        attention_scores = attention_scores + attention_mask

    # Normalize the attention scores to probabilities.
    attention_probs = ops.softmax(attention_scores, axis=-1)

    # This is actually dropping out entire tokens to attend to, which might
    # seem a bit unusual, but is taken from the original Transformer paper.
    attention_probs = self.dropout(attention_probs)

    if head_mask is not None:
        attention_probs = attention_probs * head_mask

    context_layer = ops.matmul(attention_probs, value_layer)
    relations_values = self.relative_positions_encoding(to_seq_length)
    attention_probs_t = attention_probs.permute(2, 0, 1, 3)
    attentions_probs_r = attention_probs_t.view(
        from_seq_length, batch_size * num_attention_heads, to_seq_length
    )
    value_position_scores = ops.matmul(attentions_probs_r, relations_values)
    value_position_scores_r = value_position_scores.view(
        from_seq_length, batch_size, num_attention_heads, self.attention_head_size
    )
    value_position_scores_r_t = value_position_scores_r.permute(1, 2, 0, 3)
    context_layer = context_layer + value_position_scores_r_t
    context_layer = context_layer.permute(0, 2, 1, 3)
    new_context_layer_shape = context_layer.shape[:-2] + (self.all_head_size,)
    context_layer = context_layer.view(new_context_layer_shape)
    outputs = (context_layer, attention_probs) if output_attentions else (context_layer,)

    if self.is_decoder:
        outputs = outputs + (past_key_value,)
    return outputs

mindnlp.transformers.models.nezha.nezha.NezhaSelfAttention.transpose_for_scores(input_x)

transpose for scores

Source code in mindnlp/transformers/models/nezha/nezha.py 
250
251
252
253
254
def transpose_for_scores(self, input_x: Tensor) -> Tensor:
    """transpose for scores"""
    new_x_shape = input_x.shape[:-1] + (self.num_attention_heads, self.attention_head_size)
    input_x = input_x.view(tuple(new_x_shape))
    return input_x.permute(0, 2, 1, 3)

mindnlp.transformers.models.nezha.nezha.NezhaSelfOutput

Bases: Module

NezhaSelfOutput

Source code in mindnlp/transformers/models/nezha/nezha.py 
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
class NezhaSelfOutput(nn.Module):
    """NezhaSelfOutput"""
    def __init__(self, config):
        """
        Initializes a new instance of the NezhaSelfOutput class.

        Args:
            self (NezhaSelfOutput): The object instance.
            config:
                A configuration object that contains the settings for the NezhaSelfOutput class.

                - hidden_size (int): The size of the hidden state.
                - layer_norm_eps (float): The epsilon value for layer normalization.
                - hidden_dropout_prob (float): The dropout probability for the hidden state.

        Returns:
            None

        Raises:
            None
        """
        super().__init__()
        self.dense = nn.Linear(config.hidden_size, config.hidden_size)
        self.LayerNorm = nn.LayerNorm((config.hidden_size,), eps=config.layer_norm_eps)
        self.dropout = nn.Dropout(p=config.hidden_dropout_prob)

    def forward(self, hidden_states, input_tensor):
        """
        Constructs the self-attention output of the Nezha model.

        Args:
            self (NezhaSelfOutput): An instance of the NezhaSelfOutput class.
            hidden_states (torch.Tensor):
                A tensor representing the hidden states.

                - Shape: (batch_size, sequence_length, hidden_size).
                - Purpose: The hidden states of the previous layer in the Nezha model.
                - Restrictions: None.
            input_tensor (torch.Tensor):
                A tensor representing the input.

                - Shape: (batch_size, sequence_length, hidden_size).
                - Purpose: The input tensor to be added to the hidden states.
                - Restrictions: None.

        Returns:
            torch.Tensor:
                A tensor representing the forwarded self-attention output.

                - Shape: (batch_size, sequence_length, hidden_size).
                - Purpose: The self-attention output of the Nezha model.

        Raises:
            None.
        """
        hidden_states = self.dense(hidden_states)
        hidden_states = self.dropout(hidden_states)
        hidden_states = self.LayerNorm(hidden_states + input_tensor)
        return hidden_states

mindnlp.transformers.models.nezha.nezha.NezhaSelfOutput.__init__(config)

Initializes a new instance of the NezhaSelfOutput class.

PARAMETER DESCRIPTION
self

The object instance.

TYPE: NezhaSelfOutput

config

A configuration object that contains the settings for the NezhaSelfOutput class.

  • hidden_size (int): The size of the hidden state.
  • layer_norm_eps (float): The epsilon value for layer normalization.
  • hidden_dropout_prob (float): The dropout probability for the hidden state.
RETURNS DESCRIPTION

None
Source code in mindnlp/transformers/models/nezha/nezha.py 
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
def __init__(self, config):
    """
    Initializes a new instance of the NezhaSelfOutput class.

    Args:
        self (NezhaSelfOutput): The object instance.
        config:
            A configuration object that contains the settings for the NezhaSelfOutput class.

            - hidden_size (int): The size of the hidden state.
            - layer_norm_eps (float): The epsilon value for layer normalization.
            - hidden_dropout_prob (float): The dropout probability for the hidden state.

    Returns:
        None

    Raises:
        None
    """
    super().__init__()
    self.dense = nn.Linear(config.hidden_size, config.hidden_size)
    self.LayerNorm = nn.LayerNorm((config.hidden_size,), eps=config.layer_norm_eps)
    self.dropout = nn.Dropout(p=config.hidden_dropout_prob)

mindnlp.transformers.models.nezha.nezha.NezhaSelfOutput.forward(hidden_states, input_tensor)

Constructs the self-attention output of the Nezha model.

PARAMETER DESCRIPTION
self

An instance of the NezhaSelfOutput class.

TYPE: NezhaSelfOutput

hidden_states

A tensor representing the hidden states.

  • Shape: (batch_size, sequence_length, hidden_size).
  • Purpose: The hidden states of the previous layer in the Nezha model.
  • Restrictions: None.

TYPE: Tensor

input_tensor

A tensor representing the input.

  • Shape: (batch_size, sequence_length, hidden_size).
  • Purpose: The input tensor to be added to the hidden states.
  • Restrictions: None.

TYPE: Tensor

RETURNS DESCRIPTION

torch.Tensor: A tensor representing the forwarded self-attention output.

  • Shape: (batch_size, sequence_length, hidden_size).
  • Purpose: The self-attention output of the Nezha model.
Source code in mindnlp/transformers/models/nezha/nezha.py 
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
def forward(self, hidden_states, input_tensor):
    """
    Constructs the self-attention output of the Nezha model.

    Args:
        self (NezhaSelfOutput): An instance of the NezhaSelfOutput class.
        hidden_states (torch.Tensor):
            A tensor representing the hidden states.

            - Shape: (batch_size, sequence_length, hidden_size).
            - Purpose: The hidden states of the previous layer in the Nezha model.
            - Restrictions: None.
        input_tensor (torch.Tensor):
            A tensor representing the input.

            - Shape: (batch_size, sequence_length, hidden_size).
            - Purpose: The input tensor to be added to the hidden states.
            - Restrictions: None.

    Returns:
        torch.Tensor:
            A tensor representing the forwarded self-attention output.

            - Shape: (batch_size, sequence_length, hidden_size).
            - Purpose: The self-attention output of the Nezha model.

    Raises:
        None.
    """
    hidden_states = self.dense(hidden_states)
    hidden_states = self.dropout(hidden_states)
    hidden_states = self.LayerNorm(hidden_states + input_tensor)
    return hidden_states

mindnlp.transformers.models.nezha.nezha_config

model nezha config

mindnlp.transformers.models.nezha.nezha_config.NezhaConfig

Bases: PretrainedConfig

Configuration for Nezha

Source code in mindnlp/transformers/models/nezha/nezha_config.py 
 31
 32
 33
 34
 35
 36
 37
 38
 39
 40
 41
 42
 43
 44
 45
 46
 47
 48
 49
 50
 51
 52
 53
 54
 55
 56
 57
 58
 59
 60
 61
 62
 63
 64
 65
 66
 67
 68
 69
 70
 71
 72
 73
 74
 75
 76
 77
 78
 79
 80
 81
 82
 83
 84
 85
 86
 87
 88
 89
 90
 91
 92
 93
 94
 95
 96
 97
 98
 99
100
101
102
103
104
class NezhaConfig(PretrainedConfig):
    """
    Configuration for Nezha
    """
    def __init__(
        self,
        vocab_size=21128,
        hidden_size=768,
        num_hidden_layers=12,
        num_attention_heads=12,
        intermediate_size=3072,
        hidden_act="gelu",
        hidden_dropout_prob=0.1,
        attention_probs_dropout_prob=0.1,
        max_position_embeddings=512,
        max_relative_position=64,
        type_vocab_size=2,
        initializer_range=0.02,
        layer_norm_eps=1e-12,
        classifier_dropout=0.1,
        pad_token_id=0,
        bos_token_id=2,
        eos_token_id=3,
        use_cache=True,
        **kwargs,
    ):
        '''
        Initializes a new instance of the NezhaConfig class.

        Args:
            self: The instance of the class.
            vocab_size (int, optional): The size of the vocabulary. Defaults to 21128.
            hidden_size (int, optional): The size of the hidden layers. Defaults to 768.
            num_hidden_layers (int, optional): The number of hidden layers. Defaults to 12.
            num_attention_heads (int, optional): The number of attention heads. Defaults to 12.
            intermediate_size (int, optional): The size of the intermediate layers. Defaults to 3072.
            hidden_act (str, optional): The activation function for the hidden layers. Defaults to 'gelu'.
            hidden_dropout_prob (float, optional): The dropout probability for the hidden layers. Defaults to 0.1.
            attention_probs_dropout_prob (float, optional): The dropout probability for the attention probabilities.
                Defaults to 0.1.
            max_position_embeddings (int, optional): The maximum number of positional embeddings. Defaults to 512.
            max_relative_position (int, optional): The maximum relative position. Defaults to 64.
            type_vocab_size (int, optional): The size of the type vocabulary. Defaults to 2.
            initializer_range (float, optional): The range for the initializer. Defaults to 0.02.
            layer_norm_eps (float, optional): The epsilon value for layer normalization. Defaults to 1e-12.
            classifier_dropout (float, optional): The dropout probability for the classifier. Defaults to 0.1.
            pad_token_id (int, optional): The ID of the padding token. Defaults to 0.
            bos_token_id (int, optional): The ID of the beginning-of-sentence token. Defaults to 2.
            eos_token_id (int, optional): The ID of the end-of-sentence token. Defaults to 3.
            use_cache (bool, optional): Whether to use caching. Defaults to True.

        Returns:
            None.

        Raises:
            None.
        '''
        super().__init__(pad_token_id=pad_token_id, bos_token_id=bos_token_id, eos_token_id=eos_token_id, **kwargs)

        self.vocab_size = vocab_size
        self.hidden_size = hidden_size
        self.num_hidden_layers = num_hidden_layers
        self.num_attention_heads = num_attention_heads
        self.hidden_act = hidden_act
        self.intermediate_size = intermediate_size
        self.hidden_dropout_prob = hidden_dropout_prob
        self.attention_probs_dropout_prob = attention_probs_dropout_prob
        self.max_position_embeddings = max_position_embeddings
        self.max_relative_position = max_relative_position
        self.type_vocab_size = type_vocab_size
        self.initializer_range = initializer_range
        self.layer_norm_eps = layer_norm_eps
        self.classifier_dropout = classifier_dropout
        self.use_cache = use_cache

mindnlp.transformers.models.nezha.nezha_config.NezhaConfig.__init__(vocab_size=21128, hidden_size=768, num_hidden_layers=12, num_attention_heads=12, intermediate_size=3072, hidden_act='gelu', hidden_dropout_prob=0.1, attention_probs_dropout_prob=0.1, max_position_embeddings=512, max_relative_position=64, type_vocab_size=2, initializer_range=0.02, layer_norm_eps=1e-12, classifier_dropout=0.1, pad_token_id=0, bos_token_id=2, eos_token_id=3, use_cache=True, **kwargs)

Initializes a new instance of the NezhaConfig class.

PARAMETER DESCRIPTION
self

The instance of the class.
vocab_size

The size of the vocabulary. Defaults to 21128.

TYPE: int DEFAULT: 21128

hidden_size

The size of the hidden layers. Defaults to 768.

TYPE: int DEFAULT: 768

num_hidden_layers

The number of hidden layers. Defaults to 12.

TYPE: int DEFAULT: 12

num_attention_heads

The number of attention heads. Defaults to 12.

TYPE: int DEFAULT: 12

intermediate_size

The size of the intermediate layers. Defaults to 3072.

TYPE: int DEFAULT: 3072

hidden_act

The activation function for the hidden layers. Defaults to 'gelu'.

TYPE: str DEFAULT: 'gelu'

hidden_dropout_prob

The dropout probability for the hidden layers. Defaults to 0.1.

TYPE: float DEFAULT: 0.1

attention_probs_dropout_prob

The dropout probability for the attention probabilities. Defaults to 0.1.

TYPE: float DEFAULT: 0.1

max_position_embeddings

The maximum number of positional embeddings. Defaults to 512.

TYPE: int DEFAULT: 512

max_relative_position

The maximum relative position. Defaults to 64.

TYPE: int DEFAULT: 64

type_vocab_size

The size of the type vocabulary. Defaults to 2.

TYPE: int DEFAULT: 2

initializer_range

The range for the initializer. Defaults to 0.02.

TYPE: float DEFAULT: 0.02

layer_norm_eps

The epsilon value for layer normalization. Defaults to 1e-12.

TYPE: float DEFAULT: 1e-12

classifier_dropout

The dropout probability for the classifier. Defaults to 0.1.

TYPE: float DEFAULT: 0.1

pad_token_id

The ID of the padding token. Defaults to 0.

TYPE: int DEFAULT: 0

bos_token_id

The ID of the beginning-of-sentence token. Defaults to 2.

TYPE: int DEFAULT: 2

eos_token_id

The ID of the end-of-sentence token. Defaults to 3.

TYPE: int DEFAULT: 3

use_cache

Whether to use caching. Defaults to True.

TYPE: bool DEFAULT: True

RETURNS DESCRIPTION

None.
Source code in mindnlp/transformers/models/nezha/nezha_config.py 
 35
 36
 37
 38
 39
 40
 41
 42
 43
 44
 45
 46
 47
 48
 49
 50
 51
 52
 53
 54
 55
 56
 57
 58
 59
 60
 61
 62
 63
 64
 65
 66
 67
 68
 69
 70
 71
 72
 73
 74
 75
 76
 77
 78
 79
 80
 81
 82
 83
 84
 85
 86
 87
 88
 89
 90
 91
 92
 93
 94
 95
 96
 97
 98
 99
100
101
102
103
104
def __init__(
    self,
    vocab_size=21128,
    hidden_size=768,
    num_hidden_layers=12,
    num_attention_heads=12,
    intermediate_size=3072,
    hidden_act="gelu",
    hidden_dropout_prob=0.1,