deberta_v2

mindnlp.transformers.models.deberta_v2.configuration_deberta_v2

DeBERTa-v2 model configuration

mindnlp.transformers.models.deberta_v2.configuration_deberta_v2.DebertaV2Config

Bases: PretrainedConfig

This is the configuration class to store the configuration of a [DebertaV2Model]. It is used to instantiate a DeBERTa-v2 model according to the specified arguments, defining the model architecture. Instantiating a configuration with the defaults will yield a similar configuration to that of the DeBERTa microsoft/deberta-v2-xlarge architecture.

Configuration objects inherit from [PretrainedConfig] and can be used to control the model outputs. Read the documentation from [PretrainedConfig] for more information.

PARAMETER	DESCRIPTION
vocab_size	Vocabulary size of the DeBERTa-v2 model. Defines the number of different tokens that can be represented by the inputs_ids passed when calling [DebertaV2Model]. TYPE: `int`, *optional*, defaults to 128100 DEFAULT: 128100
hidden_size	Dimensionality of the encoder layers and the pooler layer. TYPE: `int`, *optional*, defaults to 1536 DEFAULT: 1536
num_hidden_layers	Number of hidden layers in the Transformer encoder. TYPE: `int`, *optional*, defaults to 24 DEFAULT: 24
num_attention_heads	Number of attention heads for each attention layer in the Transformer encoder. TYPE: `int`, *optional*, defaults to 24 DEFAULT: 24
intermediate_size	Dimensionality of the "intermediate" (often named feed-forward) layer in the Transformer encoder. TYPE: `int`, *optional*, defaults to 6144 DEFAULT: 6144
hidden_act	The non-linear activation function (function or string) in the encoder and pooler. If string, "gelu", "relu", "silu", "gelu", "tanh", "gelu_fast", "mish", "linear", "sigmoid" and "gelu_new" are supported. TYPE: `str` or `Callable`, *optional*, defaults to `"gelu"` DEFAULT: 'gelu'
hidden_dropout_prob	The dropout probability for all fully connected layers in the embeddings, encoder, and pooler. TYPE: `float`, *optional*, defaults to 0.1 DEFAULT: 0.1
attention_probs_dropout_prob	The dropout ratio for the attention probabilities. TYPE: `float`, *optional*, defaults to 0.1 DEFAULT: 0.1
max_position_embeddings	The maximum sequence length that this model might ever be used with. Typically set this to something large just in case (e.g., 512 or 1024 or 2048). TYPE: `int`, *optional*, defaults to 512 DEFAULT: 512
type_vocab_size	The vocabulary size of the token_type_ids passed when calling [DebertaModel] or [TFDebertaModel]. TYPE: `int`, *optional*, defaults to 0 DEFAULT: 0
initializer_range	The standard deviation of the truncated_normal_initializer for initializing all weight matrices. TYPE: `float`, *optional*, defaults to 0.02 DEFAULT: 0.02
layer_norm_eps	The epsilon used by the layer normalization layers. TYPE: `float`, *optional*, defaults to 1e-7 DEFAULT: 1e-07
relative_attention	Whether use relative position encoding. TYPE: `bool`, *optional*, defaults to `True` DEFAULT: False
max_relative_positions	The range of relative positions [-max_position_embeddings, max_position_embeddings]. Use the same value as max_position_embeddings. TYPE: `int`, *optional*, defaults to -1 DEFAULT: -1
pad_token_id	The value used to pad input_ids. TYPE: `int`, *optional*, defaults to 0 DEFAULT: 0
position_biased_input	Whether add absolute position embedding to content embedding. TYPE: `bool`, *optional*, defaults to `True` DEFAULT: True
pos_att_type	The type of relative position attention, it can be a combination of ["p2c", "c2p"], e.g. ["p2c"], ["p2c", "c2p"], ["p2c", "c2p"]. TYPE: `List[str]`, *optional* DEFAULT: None
layer_norm_eps	The epsilon used by the layer normalization layers. TYPE: `float`, optional, defaults to 1e-12 DEFAULT: 1e-07

Example

>>> from transformers import DebertaV2Config, DebertaV2Model
...
>>> # Initializing a DeBERTa-v2 microsoft/deberta-v2-xlarge style configuration
>>> configuration = DebertaV2Config()
...
>>> # Initializing a model (with random weights) from the microsoft/deberta-v2-xlarge style configuration
>>> model = DebertaV2Model(configuration)
...
>>> # Accessing the model configuration
>>> configuration = model.config

Source code in mindnlp\transformers\models\deberta_v2\configuration_deberta_v2.py

class DebertaV2Config(PretrainedConfig):
    r"""
    This is the configuration class to store the configuration of a [`DebertaV2Model`]. It is used to instantiate a
    DeBERTa-v2 model according to the specified arguments, defining the model architecture. Instantiating a
    configuration with the defaults will yield a similar configuration to that of the DeBERTa
    [microsoft/deberta-v2-xlarge](https://huggingface.co/microsoft/deberta-v2-xlarge) architecture.

    Configuration objects inherit from [`PretrainedConfig`] and can be used to control the model outputs. Read the
    documentation from [`PretrainedConfig`] for more information.

    Arguments:
        vocab_size (`int`, *optional*, defaults to 128100):
            Vocabulary size of the DeBERTa-v2 model. Defines the number of different tokens that can be represented by
            the `inputs_ids` passed when calling [`DebertaV2Model`].
        hidden_size (`int`, *optional*, defaults to 1536):
            Dimensionality of the encoder layers and the pooler layer.
        num_hidden_layers (`int`, *optional*, defaults to 24):
            Number of hidden layers in the Transformer encoder.
        num_attention_heads (`int`, *optional*, defaults to 24):
            Number of attention heads for each attention layer in the Transformer encoder.
        intermediate_size (`int`, *optional*, defaults to 6144):
            Dimensionality of the "intermediate" (often named feed-forward) layer in the Transformer encoder.
        hidden_act (`str` or `Callable`, *optional*, defaults to `"gelu"`):
            The non-linear activation function (function or string) in the encoder and pooler. If string, `"gelu"`,
            `"relu"`, `"silu"`, `"gelu"`, `"tanh"`, `"gelu_fast"`, `"mish"`, `"linear"`, `"sigmoid"` and `"gelu_new"`
            are supported.
        hidden_dropout_prob (`float`, *optional*, defaults to 0.1):
            The dropout probability for all fully connected layers in the embeddings, encoder, and pooler.
        attention_probs_dropout_prob (`float`, *optional*, defaults to 0.1):
            The dropout ratio for the attention probabilities.
        max_position_embeddings (`int`, *optional*, defaults to 512):
            The maximum sequence length that this model might ever be used with. Typically set this to something large
            just in case (e.g., 512 or 1024 or 2048).
        type_vocab_size (`int`, *optional*, defaults to 0):
            The vocabulary size of the `token_type_ids` passed when calling [`DebertaModel`] or [`TFDebertaModel`].
        initializer_range (`float`, *optional*, defaults to 0.02):
            The standard deviation of the truncated_normal_initializer for initializing all weight matrices.
        layer_norm_eps (`float`, *optional*, defaults to 1e-7):
            The epsilon used by the layer normalization layers.
        relative_attention (`bool`, *optional*, defaults to `True`):
            Whether use relative position encoding.
        max_relative_positions (`int`, *optional*, defaults to -1):
            The range of relative positions `[-max_position_embeddings, max_position_embeddings]`. Use the same value
            as `max_position_embeddings`.
        pad_token_id (`int`, *optional*, defaults to 0):
            The value used to pad input_ids.
        position_biased_input (`bool`, *optional*, defaults to `True`):
            Whether add absolute position embedding to content embedding.
        pos_att_type (`List[str]`, *optional*):
            The type of relative position attention, it can be a combination of `["p2c", "c2p"]`, e.g. `["p2c"]`,
            `["p2c", "c2p"]`, `["p2c", "c2p"]`.
        layer_norm_eps (`float`, optional, defaults to 1e-12):
            The epsilon used by the layer normalization layers.

    Example:
        ```python
        >>> from transformers import DebertaV2Config, DebertaV2Model
        ...
        >>> # Initializing a DeBERTa-v2 microsoft/deberta-v2-xlarge style configuration
        >>> configuration = DebertaV2Config()
        ...
        >>> # Initializing a model (with random weights) from the microsoft/deberta-v2-xlarge style configuration
        >>> model = DebertaV2Model(configuration)
        ...
        >>> # Accessing the model configuration
        >>> configuration = model.config
        ```
    """

    model_type = "deberta-v2"

    def __init__(
        self,
        vocab_size=128100,
        hidden_size=1536,
        num_hidden_layers=24,
        num_attention_heads=24,
        intermediate_size=6144,
        hidden_act="gelu",
        hidden_dropout_prob=0.1,
        attention_probs_dropout_prob=0.1,
        max_position_embeddings=512,
        type_vocab_size=0,
        initializer_range=0.02,
        layer_norm_eps=1e-7,
        relative_attention=False,
        max_relative_positions=-1,
        pad_token_id=0,
        position_biased_input=True,
        pos_att_type=None,
        pooler_dropout=0,
        pooler_hidden_act="gelu",
        **kwargs,
    ):
        super().__init__(**kwargs)

        self.hidden_size = hidden_size
        self.num_hidden_layers = num_hidden_layers
        self.num_attention_heads = num_attention_heads
        self.intermediate_size = intermediate_size
        self.hidden_act = hidden_act
        self.hidden_dropout_prob = hidden_dropout_prob
        self.attention_probs_dropout_prob = attention_probs_dropout_prob
        self.max_position_embeddings = max_position_embeddings
        self.type_vocab_size = type_vocab_size
        self.initializer_range = initializer_range
        self.relative_attention = relative_attention
        self.max_relative_positions = max_relative_positions
        self.pad_token_id = pad_token_id
        self.position_biased_input = position_biased_input

        # Backwards compatibility
        if isinstance(pos_att_type, str):
            pos_att_type = [x.strip() for x in pos_att_type.lower().split("|")]

        self.pos_att_type = pos_att_type
        self.vocab_size = vocab_size
        self.layer_norm_eps = layer_norm_eps

        self.pooler_hidden_size = kwargs.get("pooler_hidden_size", hidden_size)
        self.pooler_dropout = pooler_dropout
        self.pooler_hidden_act = pooler_hidden_act

mindnlp.transformers.models.deberta_v2.modeling_deberta_v2

MindSpore DeBERTa-v2 model.

mindnlp.transformers.models.deberta_v2.modeling_deberta_v2.DebertaV2Embeddings

Bases: Module

Construct the embeddings from word, position and token_type embeddings.

Source code in mindnlp\transformers\models\deberta_v2\modeling_deberta_v2.py

class DebertaV2Embeddings(nn.Module):
    """Construct the embeddings from word, position and token_type embeddings."""

    def __init__(self, config):
        super().__init__()
        pad_token_id = getattr(config, "pad_token_id", 0)
        self.embedding_size = getattr(config, "embedding_size", config.hidden_size)
        self.word_embeddings = nn.Embedding(config.vocab_size, self.embedding_size, padding_idx=pad_token_id)

        self.position_biased_input = getattr(config, "position_biased_input", True)
        if not self.position_biased_input:
            self.position_embeddings = None
        else:
            self.position_embeddings = nn.Embedding(config.max_position_embeddings, self.embedding_size)

        if config.type_vocab_size > 0:
            self.token_type_embeddings = nn.Embedding(config.type_vocab_size, self.embedding_size)

        if self.embedding_size != config.hidden_size:
            self.embed_proj = nn.Linear(self.embedding_size, config.hidden_size, bias=False)
        self.LayerNorm = LayerNorm(config.hidden_size, config.layer_norm_eps)
        self.dropout = StableDropout(config.hidden_dropout_prob)
        self.config = config

        # position_ids (1, len position emb) is contiguous in memory and exported when serialized
        self.register_buffer(
            "position_ids", ops.arange(config.max_position_embeddings).broadcast_to((1, -1)), persistent=False
        )

    def forward(self, input_ids=None, token_type_ids=None, position_ids=None, mask=None, inputs_embeds=None):
        if input_ids is not None:
            input_shape = input_ids.shape
        else:
            input_shape = inputs_embeds.shape[:-1]

        seq_length = input_shape[1]

        if position_ids is None:
            position_ids = self.position_ids[:, :seq_length]

        if token_type_ids is None:
            token_type_ids = ops.zeros(input_shape, dtype=mindspore.int64)

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

        if self.position_embeddings is not None:
            position_embeddings = self.position_embeddings(position_ids.long())
        else:
            position_embeddings = ops.zeros_like(inputs_embeds)

        embeddings = inputs_embeds
        if self.position_biased_input:
            embeddings += position_embeddings
        if self.config.type_vocab_size > 0:
            token_type_embeddings = self.token_type_embeddings(token_type_ids)
            embeddings += token_type_embeddings

        if self.embedding_size != self.config.hidden_size:
            embeddings = self.embed_proj(embeddings)

        embeddings = self.LayerNorm(embeddings)

        if mask is not None:
            if mask.dim() != embeddings.dim():
                if mask.dim() == 4:
                    mask = mask.squeeze(1).squeeze(1)
                mask = mask.unsqueeze(2)
            mask = mask.to(embeddings.dtype)

            embeddings = embeddings * mask

        embeddings = self.dropout(embeddings)
        return embeddings

mindnlp.transformers.models.deberta_v2.modeling_deberta_v2.DebertaV2Encoder

Bases: Module

Modified BertEncoder with relative position bias support

Source code in mindnlp\transformers\models\deberta_v2\modeling_deberta_v2.py

class DebertaV2Encoder(nn.Module):
    """Modified BertEncoder with relative position bias support"""

    def __init__(self, config):
        super().__init__()

        self.layer = nn.ModuleList([DebertaV2Layer(config) for _ in range(config.num_hidden_layers)])
        self.relative_attention = getattr(config, "relative_attention", False)

        if self.relative_attention:
            self.max_relative_positions = getattr(config, "max_relative_positions", -1)
            if self.max_relative_positions < 1:
                self.max_relative_positions = config.max_position_embeddings

            self.position_buckets = getattr(config, "position_buckets", -1)
            pos_ebd_size = self.max_relative_positions * 2

            if self.position_buckets > 0:
                pos_ebd_size = self.position_buckets * 2

            self.rel_embeddings = nn.Embedding(pos_ebd_size, config.hidden_size)

        self.norm_rel_ebd = [x.strip() for x in getattr(config, "norm_rel_ebd", "none").lower().split("|")]

        if "layer_norm" in self.norm_rel_ebd:
            self.LayerNorm = LayerNorm(config.hidden_size, config.layer_norm_eps, elementwise_affine=True)

        self.conv = ConvLayer(config) if getattr(config, "conv_kernel_size", 0) > 0 else None
        self.gradient_checkpointing = False

    def get_rel_embedding(self):
        rel_embeddings = self.rel_embeddings.weight if self.relative_attention else None
        if rel_embeddings is not None and ("layer_norm" in self.norm_rel_ebd):
            rel_embeddings = self.LayerNorm(rel_embeddings)
        return rel_embeddings

    def get_attention_mask(self, attention_mask):
        if attention_mask.dim() <= 2:
            extended_attention_mask = attention_mask.unsqueeze(1).unsqueeze(2)
            attention_mask = extended_attention_mask * extended_attention_mask.squeeze(-2).unsqueeze(-1)
        elif attention_mask.dim() == 3:
            attention_mask = attention_mask.unsqueeze(1)

        return attention_mask

    def get_rel_pos(self, hidden_states, query_states=None, relative_pos=None):
        if self.relative_attention and relative_pos is None:
            q = query_states.shape[-2] if query_states is not None else hidden_states.shape[-2]
            relative_pos = build_relative_position(
                q,
                hidden_states.shape[-2],
                bucket_size=self.position_buckets,
                max_position=self.max_relative_positions,
            )
        return relative_pos

    def forward(
        self,
        hidden_states,
        attention_mask,
        output_hidden_states=True,
        output_attentions=False,
        query_states=None,
        relative_pos=None,
        return_dict=True,
    ):
        if attention_mask.dim() <= 2:
            input_mask = attention_mask
        else:
            input_mask = attention_mask.sum(-2) > 0
        attention_mask = self.get_attention_mask(attention_mask)
        relative_pos = self.get_rel_pos(hidden_states, query_states, relative_pos)

        all_hidden_states = () if output_hidden_states else None
        all_attentions = () if output_attentions else None

        if isinstance(hidden_states, Sequence):
            next_kv = hidden_states[0]
        else:
            next_kv = hidden_states
        rel_embeddings = self.get_rel_embedding()
        output_states = next_kv
        for i, layer_module in enumerate(self.layer):
            if output_hidden_states:
                all_hidden_states = all_hidden_states + (output_states,)

            if self.gradient_checkpointing and self.training:
                output_states = self._gradient_checkpointing_func(
                    layer_module.__call__,
                    next_kv,
                    attention_mask,
                    query_states,
                    relative_pos,
                    rel_embeddings,
                    output_attentions,
                )
            else:
                output_states = layer_module(
                    next_kv,
                    attention_mask,
                    query_states=query_states,
                    relative_pos=relative_pos,
                    rel_embeddings=rel_embeddings,
                    output_attentions=output_attentions,
                )

            if output_attentions:
                output_states, att_m = output_states

            if i == 0 and self.conv is not None:
                output_states = self.conv(hidden_states, output_states, input_mask)

            if query_states is not None:
                query_states = output_states
                if isinstance(hidden_states, Sequence):
                    next_kv = hidden_states[i + 1] if i + 1 < len(self.layer) else None
            else:
                next_kv = output_states

            if output_attentions:
                all_attentions = all_attentions + (att_m,)

        if output_hidden_states:
            all_hidden_states = all_hidden_states + (output_states,)

        if not return_dict:
            return tuple(v for v in [output_states, all_hidden_states, all_attentions] if v is not None)
        return BaseModelOutput(
            last_hidden_state=output_states, hidden_states=all_hidden_states, attentions=all_attentions
        )

mindnlp.transformers.models.deberta_v2.modeling_deberta_v2.DebertaV2ForMaskedLM

Bases: DebertaV2PreTrainedModel

Source code in mindnlp\transformers\models\deberta_v2\modeling_deberta_v2.py

class DebertaV2ForMaskedLM(DebertaV2PreTrainedModel):
    _tied_weights_keys = ["cls.predictions.decoder.weight", "cls.predictions.decoder.bias"]

    def __init__(self, config):
        super().__init__(config)

        self.deberta = DebertaV2Model(config)
        self.cls = DebertaV2OnlyMLMHead(config)

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

    def get_output_embeddings(self):
        return self.cls.predictions.decoder

    def set_output_embeddings(self, new_embeddings):
        self.cls.predictions.decoder = new_embeddings
        self.cls.predictions.bias = new_embeddings.bias

    # Copied from transformers.models.deberta.modeling_deberta.DebertaForMaskedLM.forward with Deberta->DebertaV2
    def forward(
        self,
        input_ids: Optional[mindspore.Tensor] = None,
        attention_mask: Optional[mindspore.Tensor] = None,
        token_type_ids: Optional[mindspore.Tensor] = None,
        position_ids: Optional[mindspore.Tensor] = None,
        inputs_embeds: Optional[mindspore.Tensor] = None,
        labels: Optional[mindspore.Tensor] = None,
        output_attentions: Optional[bool] = None,
        output_hidden_states: Optional[bool] = None,
        return_dict: Optional[bool] = None,
    ) -> Union[Tuple, MaskedLMOutput]:
        r"""
        labels (`mindspore.Tensor` of shape `(batch_size, sequence_length)`, *optional*):
            Labels for computing the masked language modeling loss. Indices should be in `[-100, 0, ...,
            config.vocab_size]` (see `input_ids` docstring) Tokens with indices set to `-100` are ignored (masked), the
            loss is only computed for the tokens with labels in `[0, ..., config.vocab_size]`
        """

        return_dict = return_dict if return_dict is not None else self.config.use_return_dict

        outputs = self.deberta(
            input_ids,
            attention_mask=attention_mask,
            token_type_ids=token_type_ids,
            position_ids=position_ids,
            inputs_embeds=inputs_embeds,
            output_attentions=output_attentions,
            output_hidden_states=output_hidden_states,
            return_dict=return_dict,
        )

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

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

        if not return_dict:
            output = (prediction_scores,) + outputs[1:]
            return ((masked_lm_loss,) + output) if masked_lm_loss is not None else output

        return MaskedLMOutput(
            loss=masked_lm_loss,
            logits=prediction_scores,
            hidden_states=outputs.hidden_states,
            attentions=outputs.attentions,
        )

mindnlp.transformers.models.deberta_v2.modeling_deberta_v2.DebertaV2ForMaskedLM.forward(input_ids=None, attention_mask=None, token_type_ids=None, position_ids=None, inputs_embeds=None, labels=None, output_attentions=None, output_hidden_states=None, return_dict=None)

labels (mindspore.Tensor of shape (batch_size, sequence_length), optional): Labels for computing the masked language modeling loss. Indices should be in [-100, 0, ..., config.vocab_size] (see input_ids docstring) Tokens with indices set to -100 are ignored (masked), the loss is only computed for the tokens with labels in [0, ..., config.vocab_size]

Source code in mindnlp\transformers\models\deberta_v2\modeling_deberta_v2.py

def forward(
    self,
    input_ids: Optional[mindspore.Tensor] = None,
    attention_mask: Optional[mindspore.Tensor] = None,
    token_type_ids: Optional[mindspore.Tensor] = None,
    position_ids: Optional[mindspore.Tensor] = None,
    inputs_embeds: Optional[mindspore.Tensor] = None,
    labels: Optional[mindspore.Tensor] = None,
    output_attentions: Optional[bool] = None,
    output_hidden_states: Optional[bool] = None,
    return_dict: Optional[bool] = None,
) -> Union[Tuple, MaskedLMOutput]:
    r"""
    labels (`mindspore.Tensor` of shape `(batch_size, sequence_length)`, *optional*):
        Labels for computing the masked language modeling loss. Indices should be in `[-100, 0, ...,
        config.vocab_size]` (see `input_ids` docstring) Tokens with indices set to `-100` are ignored (masked), the
        loss is only computed for the tokens with labels in `[0, ..., config.vocab_size]`
    """

    return_dict = return_dict if return_dict is not None else self.config.use_return_dict

    outputs = self.deberta(
        input_ids,
        attention_mask=attention_mask,
        token_type_ids=token_type_ids,
        position_ids=position_ids,
        inputs_embeds=inputs_embeds,
        output_attentions=output_attentions,
        output_hidden_states=output_hidden_states,
        return_dict=return_dict,
    )

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

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

    if not return_dict:
        output = (prediction_scores,) + outputs[1:]
        return ((masked_lm_loss,) + output) if masked_lm_loss is not None else output

    return MaskedLMOutput(
        loss=masked_lm_loss,
        logits=prediction_scores,
        hidden_states=outputs.hidden_states,
        attentions=outputs.attentions,
    )

mindnlp.transformers.models.deberta_v2.modeling_deberta_v2.DebertaV2ForMultipleChoice

Bases: DebertaV2PreTrainedModel

Source code in mindnlp\transformers\models\deberta_v2\modeling_deberta_v2.py

class DebertaV2ForMultipleChoice(DebertaV2PreTrainedModel):
    def __init__(self, config):
        super().__init__(config)

        num_labels = getattr(config, "num_labels", 2)
        self.num_labels = num_labels

        self.deberta = DebertaV2Model(config)
        self.pooler = ContextPooler(config)
        output_dim = self.pooler.output_dim

        self.classifier = nn.Linear(output_dim, 1)
        drop_out = getattr(config, "cls_dropout", None)
        drop_out = self.config.hidden_dropout_prob if drop_out is None else drop_out
        self.dropout = StableDropout(drop_out)

        self.init_weights()

    def get_input_embeddings(self):
        return self.deberta.get_input_embeddings()

    def set_input_embeddings(self, new_embeddings):
        self.deberta.set_input_embeddings(new_embeddings)

    def forward(
        self,
        input_ids: Optional[mindspore.Tensor] = None,
        attention_mask: Optional[mindspore.Tensor] = None,
        token_type_ids: Optional[mindspore.Tensor] = None,
        position_ids: Optional[mindspore.Tensor] = None,
        inputs_embeds: Optional[mindspore.Tensor] = None,
        labels: Optional[mindspore.Tensor] = None,
        output_attentions: Optional[bool] = None,
        output_hidden_states: Optional[bool] = None,
        return_dict: Optional[bool] = None,
    ) -> Union[Tuple, MultipleChoiceModelOutput]:
        r"""
        labels (`mindspore.Tensor` 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)
        """
        return_dict = return_dict if return_dict is not None else self.config.use_return_dict
        num_choices = input_ids.shape[1] if input_ids is not None else inputs_embeds.shape[1]

        flat_input_ids = input_ids.view(-1, input_ids.shape[-1]) if input_ids is not None else None
        flat_position_ids = position_ids.view(-1, position_ids.shape[-1]) if position_ids is not None else None
        flat_token_type_ids = token_type_ids.view(-1, token_type_ids.shape[-1]) if token_type_ids is not None else None
        flat_attention_mask = attention_mask.view(-1, attention_mask.shape[-1]) if attention_mask is not None else None
        flat_inputs_embeds = (
            inputs_embeds.view(-1, inputs_embeds.shape[-2], inputs_embeds.shape[-1])
            if inputs_embeds is not None
            else None
        )

        outputs = self.deberta(
            flat_input_ids,
            position_ids=flat_position_ids,
            token_type_ids=flat_token_type_ids,
            attention_mask=flat_attention_mask,
            inputs_embeds=flat_inputs_embeds,
            output_attentions=output_attentions,
            output_hidden_states=output_hidden_states,
            return_dict=return_dict,
        )

        encoder_layer = outputs[0]
        pooled_output = self.pooler(encoder_layer)
        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 = CrossEntropyLoss()
            loss = loss_fct(reshaped_logits, labels)

        if not return_dict:
            output = (reshaped_logits,) + outputs[1:]
            return ((loss,) + output) if loss is not None else output

        return MultipleChoiceModelOutput(
            loss=loss,
            logits=reshaped_logits,
            hidden_states=outputs.hidden_states,
            attentions=outputs.attentions,
        )

mindnlp.transformers.models.deberta_v2.modeling_deberta_v2.DebertaV2ForMultipleChoice.forward(input_ids=None, attention_mask=None, token_type_ids=None, position_ids=None, inputs_embeds=None, labels=None, output_attentions=None, output_hidden_states=None, return_dict=None)

labels (mindspore.Tensor 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)

Source code in mindnlp\transformers\models\deberta_v2\modeling_deberta_v2.py

def forward(
    self,
    input_ids: Optional[mindspore.Tensor] = None,
    attention_mask: Optional[mindspore.Tensor] = None,
    token_type_ids: Optional[mindspore.Tensor] = None,
    position_ids: Optional[mindspore.Tensor] = None,
    inputs_embeds: Optional[mindspore.Tensor] = None,
    labels: Optional[mindspore.Tensor] = None,
    output_attentions: Optional[bool] = None,
    output_hidden_states: Optional[bool] = None,
    return_dict: Optional[bool] = None,
) -> Union[Tuple, MultipleChoiceModelOutput]:
    r"""
    labels (`mindspore.Tensor` 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)
    """
    return_dict = return_dict if return_dict is not None else self.config.use_return_dict
    num_choices = input_ids.shape[1] if input_ids is not None else inputs_embeds.shape[1]

    flat_input_ids = input_ids.view(-1, input_ids.shape[-1]) if input_ids is not None else None
    flat_position_ids = position_ids.view(-1, position_ids.shape[-1]) if position_ids is not None else None
    flat_token_type_ids = token_type_ids.view(-1, token_type_ids.shape[-1]) if token_type_ids is not None else None
    flat_attention_mask = attention_mask.view(-1, attention_mask.shape[-1]) if attention_mask is not None else None
    flat_inputs_embeds = (
        inputs_embeds.view(-1, inputs_embeds.shape[-2], inputs_embeds.shape[-1])
        if inputs_embeds is not None
        else None
    )

    outputs = self.deberta(
        flat_input_ids,
        position_ids=flat_position_ids,
        token_type_ids=flat_token_type_ids,
        attention_mask=flat_attention_mask,
        inputs_embeds=flat_inputs_embeds,
        output_attentions=output_attentions,
        output_hidden_states=output_hidden_states,
        return_dict=return_dict,
    )

    encoder_layer = outputs[0]
    pooled_output = self.pooler(encoder_layer)
    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 = CrossEntropyLoss()
        loss = loss_fct(reshaped_logits, labels)

    if not return_dict:
        output = (reshaped_logits,) + outputs[1:]
        return ((loss,) + output) if loss is not None else output

    return MultipleChoiceModelOutput(
        loss=loss,
        logits=reshaped_logits,
        hidden_states=outputs.hidden_states,
        attentions=outputs.attentions,
    )

mindnlp.transformers.models.deberta_v2.modeling_deberta_v2.DebertaV2ForQuestionAnswering

Bases: DebertaV2PreTrainedModel

Source code in mindnlp\transformers\models\deberta_v2\modeling_deberta_v2.py

class DebertaV2ForQuestionAnswering(DebertaV2PreTrainedModel):
    def __init__(self, config):
        super().__init__(config)
        self.num_labels = config.num_labels

        self.deberta = DebertaV2Model(config)
        self.qa_outputs = nn.Linear(config.hidden_size, config.num_labels)

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

    # Copied from transformers.models.deberta.modeling_deberta.DebertaForQuestionAnswering.forward with Deberta->DebertaV2
    def forward(
        self,
        input_ids: Optional[mindspore.Tensor] = None,
        attention_mask: Optional[mindspore.Tensor] = None,
        token_type_ids: Optional[mindspore.Tensor] = None,
        position_ids: Optional[mindspore.Tensor] = None,
        inputs_embeds: Optional[mindspore.Tensor] = None,
        start_positions: Optional[mindspore.Tensor] = None,
        end_positions: Optional[mindspore.Tensor] = None,
        output_attentions: Optional[bool] = None,
        output_hidden_states: Optional[bool] = None,
        return_dict: Optional[bool] = None,
    ) -> Union[Tuple, QuestionAnsweringModelOutput]:
        r"""
        start_positions (`mindspore.Tensor` 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` 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.
        """
        return_dict = return_dict if return_dict is not None else self.config.use_return_dict

        outputs = self.deberta(
            input_ids,
            attention_mask=attention_mask,
            token_type_ids=token_type_ids,
            position_ids=position_ids,
            inputs_embeds=inputs_embeds,
            output_attentions=output_attentions,
            output_hidden_states=output_hidden_states,
            return_dict=return_dict,
        )

        sequence_output = outputs[0]

        logits = self.qa_outputs(sequence_output)
        start_logits, end_logits = ops.split(logits, 1, dim=-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 = 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

        if not return_dict:
            output = (start_logits, end_logits) + outputs[1:]
            return ((total_loss,) + output) if total_loss is not None else output

        return QuestionAnsweringModelOutput(
            loss=total_loss,
            start_logits=start_logits,
            end_logits=end_logits,
            hidden_states=outputs.hidden_states,
            attentions=outputs.attentions,
        )

mindnlp.transformers.models.deberta_v2.modeling_deberta_v2.DebertaV2ForQuestionAnswering.forward(input_ids=None, attention_mask=None, token_type_ids=None, position_ids=None, inputs_embeds=None, start_positions=None, end_positions=None, output_attentions=None, output_hidden_states=None, return_dict=None)

start_positions (mindspore.Tensor 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 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.

Source code in mindnlp\transformers\models\deberta_v2\modeling_deberta_v2.py

def forward(
    self,
    input_ids: Optional[mindspore.Tensor] = None,
    attention_mask: Optional[mindspore.Tensor] = None,
    token_type_ids: Optional[mindspore.Tensor] = None,
    position_ids: Optional[mindspore.Tensor] = None,
    inputs_embeds: Optional[mindspore.Tensor] = None,
    start_positions: Optional[mindspore.Tensor] = None,
    end_positions: Optional[mindspore.Tensor] = None,
    output_attentions: Optional[bool] = None,
    output_hidden_states: Optional[bool] = None,
    return_dict: Optional[bool] = None,
) -> Union[Tuple, QuestionAnsweringModelOutput]:
    r"""
    start_positions (`mindspore.Tensor` 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` 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.
    """
    return_dict = return_dict if return_dict is not None else self.config.use_return_dict

    outputs = self.deberta(
        input_ids,
        attention_mask=attention_mask,
        token_type_ids=token_type_ids,
        position_ids=position_ids,
        inputs_embeds=inputs_embeds,
        output_attentions=output_attentions,
        output_hidden_states=output_hidden_states,
        return_dict=return_dict,
    )

    sequence_output = outputs[0]

    logits = self.qa_outputs(sequence_output)
    start_logits, end_logits = ops.split(logits, 1, dim=-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 = 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

    if not return_dict:
        output = (start_logits, end_logits) + outputs[1:]
        return ((total_loss,) + output) if total_loss is not None else output

    return QuestionAnsweringModelOutput(
        loss=total_loss,
        start_logits=start_logits,
        end_logits=end_logits,
        hidden_states=outputs.hidden_states,
        attentions=outputs.attentions,
    )

mindnlp.transformers.models.deberta_v2.modeling_deberta_v2.DebertaV2ForSequenceClassification

Bases: DebertaV2PreTrainedModel

Source code in mindnlp\transformers\models\deberta_v2\modeling_deberta_v2.py

class DebertaV2ForSequenceClassification(DebertaV2PreTrainedModel):
    def __init__(self, config):
        super().__init__(config)

        num_labels = getattr(config, "num_labels", 2)
        self.num_labels = num_labels

        self.deberta = DebertaV2Model(config)
        self.pooler = ContextPooler(config)
        output_dim = self.pooler.output_dim

        self.classifier = nn.Linear(output_dim, num_labels)
        drop_out = getattr(config, "cls_dropout", None)
        drop_out = self.config.hidden_dropout_prob if drop_out is None else drop_out
        self.dropout = StableDropout(drop_out)

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

    def get_input_embeddings(self):
        return self.deberta.get_input_embeddings()

    def set_input_embeddings(self, new_embeddings):
        self.deberta.set_input_embeddings(new_embeddings)

    # Copied from transformers.models.deberta.modeling_deberta.DebertaForSequenceClassification.forward with Deberta->DebertaV2
    def forward(
        self,
        input_ids: Optional[mindspore.Tensor] = None,
        attention_mask: Optional[mindspore.Tensor] = None,
        token_type_ids: Optional[mindspore.Tensor] = None,
        position_ids: Optional[mindspore.Tensor] = None,
        inputs_embeds: Optional[mindspore.Tensor] = None,
        labels: Optional[mindspore.Tensor] = None,
        output_attentions: Optional[bool] = None,
        output_hidden_states: Optional[bool] = None,
        return_dict: Optional[bool] = None,
    ) -> Union[Tuple, SequenceClassifierOutput]:
        r"""
        labels (`mindspore.Tensor` of shape `(batch_size,)`, *optional*):
            Labels for computing the sequence classification/regression loss. Indices should be in `[0, ...,
            config.num_labels - 1]`. If `config.num_labels == 1` a regression loss is computed (Mean-Square loss), If
            `config.num_labels > 1` a classification loss is computed (Cross-Entropy).
        """
        return_dict = return_dict if return_dict is not None else self.config.use_return_dict

        outputs = self.deberta(
            input_ids,
            token_type_ids=token_type_ids,
            attention_mask=attention_mask,
            position_ids=position_ids,
            inputs_embeds=inputs_embeds,
            output_attentions=output_attentions,
            output_hidden_states=output_hidden_states,
            return_dict=return_dict,
        )

        encoder_layer = outputs[0]
        pooled_output = self.pooler(encoder_layer)
        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:
                    # regression task
                    loss_fn = nn.MSELoss()
                    logits = logits.view(-1).to(labels.dtype)
                    loss = loss_fn(logits, labels.view(-1))
                elif labels.dim() == 1 or labels.shape[-1] == 1:
                    label_index = (labels >= 0).nonzero()
                    labels = labels.long()
                    if label_index.shape[0] > 0:
                        labeled_logits = ops.gather(
                            logits, 0, label_index.broadcast_to((label_index.shape[0], logits.shape[1]))
                        )
                        labels = ops.gather(labels, 0, label_index.view(-1))
                        loss_fct = CrossEntropyLoss()
                        loss = loss_fct(labeled_logits.view(-1, self.num_labels).float(), labels.view(-1))
                    else:
                        loss = mindspore.tensor(0).to(logits)
                else:
                    log_softmax = nn.LogSoftmax(-1)
                    loss = -((log_softmax(logits) * labels).sum(-1)).mean()
            elif self.config.problem_type == "regression":
                loss_fct = 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 = CrossEntropyLoss()
                loss = loss_fct(logits.view(-1, self.num_labels), labels.view(-1))
            elif self.config.problem_type == "multi_label_classification":
                loss_fct = BCEWithLogitsLoss()
                loss = loss_fct(logits, labels)
        if not return_dict:
            output = (logits,) + outputs[1:]
            return ((loss,) + output) if loss is not None else output

        return SequenceClassifierOutput(
            loss=loss, logits=logits, hidden_states=outputs.hidden_states, attentions=outputs.attentions
        )

mindnlp.transformers.models.deberta_v2.modeling_deberta_v2.DebertaV2ForSequenceClassification.forward(input_ids=None, attention_mask=None, token_type_ids=None, position_ids=None, inputs_embeds=None, labels=None, output_attentions=None, output_hidden_states=None, return_dict=None)

labels (mindspore.Tensor of shape (batch_size,), optional): Labels for computing the sequence classification/regression loss. Indices should be in [0, ..., config.num_labels - 1]. If config.num_labels == 1 a regression loss is computed (Mean-Square loss), If config.num_labels > 1 a classification loss is computed (Cross-Entropy).

Source code in mindnlp\transformers\models\deberta_v2\modeling_deberta_v2.py

def forward(
    self,
    input_ids: Optional[mindspore.Tensor] = None,
    attention_mask: Optional[mindspore.Tensor] = None,
    token_type_ids: Optional[mindspore.Tensor] = None,
    position_ids: Optional[mindspore.Tensor] = None,
    inputs_embeds: Optional[mindspore.Tensor] = None,
    labels: Optional[mindspore.Tensor] = None,
    output_attentions: Optional[bool] = None,
    output_hidden_states: Optional[bool] = None,
    return_dict: Optional[bool] = None,
) -> Union[Tuple, SequenceClassifierOutput]:
    r"""
    labels (`mindspore.Tensor` of shape `(batch_size,)`, *optional*):
        Labels for computing the sequence classification/regression loss. Indices should be in `[0, ...,
        config.num_labels - 1]`. If `config.num_labels == 1` a regression loss is computed (Mean-Square loss), If
        `config.num_labels > 1` a classification loss is computed (Cross-Entropy).
    """
    return_dict = return_dict if return_dict is not None else self.config.use_return_dict

    outputs = self.deberta(
        input_ids,
        token_type_ids=token_type_ids,
        attention_mask=attention_mask,
        position_ids=position_ids,
        inputs_embeds=inputs_embeds,
        output_attentions=output_attentions,
        output_hidden_states=output_hidden_states,
        return_dict=return_dict,
    )

    encoder_layer = outputs[0]
    pooled_output = self.pooler(encoder_layer)
    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:
                # regression task
                loss_fn = nn.MSELoss()
                logits = logits.view(-1).to(labels.dtype)
                loss = loss_fn(logits, labels.view(-1))
            elif labels.dim() == 1 or labels.shape[-1] == 1:
                label_index = (labels >= 0).nonzero()
                labels = labels.long()
                if label_index.shape[0] > 0:
                    labeled_logits = ops.gather(
                        logits, 0, label_index.broadcast_to((label_index.shape[0], logits.shape[1]))
                    )
                    labels = ops.gather(labels, 0, label_index.view(-1))
                    loss_fct = CrossEntropyLoss()
                    loss = loss_fct(labeled_logits.view(-1, self.num_labels).float(), labels.view(-1))
                else:
                    loss = mindspore.tensor(0).to(logits)
            else:
                log_softmax = nn.LogSoftmax(-1)
                loss = -((log_softmax(logits) * labels).sum(-1)).mean()
        elif self.config.problem_type == "regression":
            loss_fct = 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 = CrossEntropyLoss()
            loss = loss_fct(logits.view(-1, self.num_labels), labels.view(-1))
        elif self.config.problem_type == "multi_label_classification":
            loss_fct = BCEWithLogitsLoss()
            loss = loss_fct(logits, labels)
    if not return_dict:
        output = (logits,) + outputs[1:]
        return ((loss,) + output) if loss is not None else output

    return SequenceClassifierOutput(
        loss=loss, logits=logits, hidden_states=outputs.hidden_states, attentions=outputs.attentions
    )

mindnlp.transformers.models.deberta_v2.modeling_deberta_v2.DebertaV2ForTokenClassification

Bases: DebertaV2PreTrainedModel

Source code in mindnlp\transformers\models\deberta_v2\modeling_deberta_v2.py

class DebertaV2ForTokenClassification(DebertaV2PreTrainedModel):
    def __init__(self, config):
        super().__init__(config)
        self.num_labels = config.num_labels

        self.deberta = DebertaV2Model(config)
        self.dropout = nn.Dropout(config.hidden_dropout_prob)
        self.classifier = nn.Linear(config.hidden_size, config.num_labels)

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

    def forward(
        self,
        input_ids: Optional[mindspore.Tensor] = None,
        attention_mask: Optional[mindspore.Tensor] = None,
        token_type_ids: Optional[mindspore.Tensor] = None,
        position_ids: Optional[mindspore.Tensor] = None,
        inputs_embeds: Optional[mindspore.Tensor] = None,
        labels: Optional[mindspore.Tensor] = None,
        output_attentions: Optional[bool] = None,
        output_hidden_states: Optional[bool] = None,
        return_dict: Optional[bool] = None,
    ) -> Union[Tuple, TokenClassifierOutput]:
        r"""
        labels (`mindspore.Tensor` of shape `(batch_size, sequence_length)`, *optional*):
            Labels for computing the token classification loss. Indices should be in `[0, ..., config.num_labels - 1]`.
        """
        return_dict = return_dict if return_dict is not None else self.config.use_return_dict

        outputs = self.deberta(
            input_ids,
            attention_mask=attention_mask,
            token_type_ids=token_type_ids,
            position_ids=position_ids,
            inputs_embeds=inputs_embeds,
            output_attentions=output_attentions,
            output_hidden_states=output_hidden_states,
            return_dict=return_dict,
        )

        sequence_output = outputs[0]

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

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

        if not return_dict:
            output = (logits,) + outputs[1:]
            return ((loss,) + output) if loss is not None else output

        return TokenClassifierOutput(
            loss=loss, logits=logits, hidden_states=outputs.hidden_states, attentions=outputs.attentions
        )

mindnlp.transformers.models.deberta_v2.modeling_deberta_v2.DebertaV2ForTokenClassification.forward(input_ids=None, attention_mask=None, token_type_ids=None, position_ids=None, inputs_embeds=None, labels=None, output_attentions=None, output_hidden_states=None, return_dict=None)

labels (mindspore.Tensor of shape (batch_size, sequence_length), optional): Labels for computing the token classification loss. Indices should be in [0, ..., config.num_labels - 1].

Source code in mindnlp\transformers\models\deberta_v2\modeling_deberta_v2.py

def forward(
    self,
    input_ids: Optional[mindspore.Tensor] = None,
    attention_mask: Optional[mindspore.Tensor] = None,
    token_type_ids: Optional[mindspore.Tensor] = None,
    position_ids: Optional[mindspore.Tensor] = None,
    inputs_embeds: Optional[mindspore.Tensor] = None,
    labels: Optional[mindspore.Tensor] = None,
    output_attentions: Optional[bool] = None,
    output_hidden_states: Optional[bool] = None,
    return_dict: Optional[bool] = None,
) -> Union[Tuple, TokenClassifierOutput]:
    r"""
    labels (`mindspore.Tensor` of shape `(batch_size, sequence_length)`, *optional*):
        Labels for computing the token classification loss. Indices should be in `[0, ..., config.num_labels - 1]`.
    """
    return_dict = return_dict if return_dict is not None else self.config.use_return_dict

    outputs = self.deberta(
        input_ids,
        attention_mask=attention_mask,
        token_type_ids=token_type_ids,
        position_ids=position_ids,
        inputs_embeds=inputs_embeds,
        output_attentions=output_attentions,
        output_hidden_states=output_hidden_states,
        return_dict=return_dict,
    )

    sequence_output = outputs[0]

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

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

    if not return_dict:
        output = (logits,) + outputs[1:]
        return ((loss,) + output) if loss is not None else output

    return TokenClassifierOutput(
        loss=loss, logits=logits, hidden_states=outputs.hidden_states, attentions=outputs.attentions
    )

mindnlp.transformers.models.deberta_v2.modeling_deberta_v2.DebertaV2Model

Bases: DebertaV2PreTrainedModel

Source code in mindnlp\transformers\models\deberta_v2\modeling_deberta_v2.py

class DebertaV2Model(DebertaV2PreTrainedModel):
    def __init__(self, config):
        super().__init__(config)

        self.embeddings = DebertaV2Embeddings(config)
        self.encoder = DebertaV2Encoder(config)
        self.z_steps = 0
        self.config = config
        # Initialize weights and apply final processing
        self.post_init()

    def get_input_embeddings(self):
        return self.embeddings.word_embeddings

    def set_input_embeddings(self, new_embeddings):
        self.embeddings.word_embeddings = new_embeddings

    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
        """
        raise NotImplementedError("The prune function is not implemented in DeBERTa model.")

    def forward(
        self,
        input_ids: Optional[mindspore.Tensor] = None,
        attention_mask: Optional[mindspore.Tensor] = None,
        token_type_ids: Optional[mindspore.Tensor] = None,
        position_ids: Optional[mindspore.Tensor] = None,
        inputs_embeds: Optional[mindspore.Tensor] = None,
        output_attentions: Optional[bool] = None,
        output_hidden_states: Optional[bool] = None,
        return_dict: Optional[bool] = None,
    ) -> Union[Tuple, BaseModelOutput]:
        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
        )
        return_dict = return_dict if return_dict is not None else self.config.use_return_dict

        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")
        elif input_ids is not None:
            self.warn_if_padding_and_no_attention_mask(input_ids, attention_mask)
            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")


        if attention_mask is None:
            attention_mask = ops.ones(input_shape)
        if token_type_ids is None:
            token_type_ids = ops.zeros(input_shape, dtype=mindspore.int64)

        embedding_output = self.embeddings(
            input_ids=input_ids,
            token_type_ids=token_type_ids,
            position_ids=position_ids,
            mask=attention_mask,
            inputs_embeds=inputs_embeds,
        )

        encoder_outputs = self.encoder(
            embedding_output,
            attention_mask,
            output_hidden_states=True,
            output_attentions=output_attentions,
            return_dict=return_dict,
        )
        encoded_layers = encoder_outputs[1]

        if self.z_steps > 1:
            hidden_states = encoded_layers[-2]
            layers = [self.encoder.layer[-1] for _ in range(self.z_steps)]
            query_states = encoded_layers[-1]
            rel_embeddings = self.encoder.get_rel_embedding()
            attention_mask = self.encoder.get_attention_mask(attention_mask)
            rel_pos = self.encoder.get_rel_pos(embedding_output)
            for layer in layers[1:]:
                query_states = layer(
                    hidden_states,
                    attention_mask,
                    output_attentions=False,
                    query_states=query_states,
                    relative_pos=rel_pos,
                    rel_embeddings=rel_embeddings,
                )
                encoded_layers.append(query_states)

        sequence_output = encoded_layers[-1]

        if not return_dict:
            return (sequence_output,) + encoder_outputs[(1 if output_hidden_states else 2) :]

        return BaseModelOutput(
            last_hidden_state=sequence_output,
            hidden_states=encoder_outputs.hidden_states if output_hidden_states else None,
            attentions=encoder_outputs.attentions,
        )

mindnlp.transformers.models.deberta_v2.modeling_deberta_v2.DebertaV2PreTrainedModel

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\deberta_v2\modeling_deberta_v2.py

class DebertaV2PreTrainedModel(PreTrainedModel):
    """
    An abstract class to handle weights initialization and a simple interface for downloading and loading pretrained
    models.
    """

    config_class = DebertaV2Config
    base_model_prefix = "deberta"
    _keys_to_ignore_on_load_unexpected = ["position_embeddings"]
    supports_gradient_checkpointing = True

    def _init_weights(self, module):
        """Initialize the weights."""
        if isinstance(module, nn.Linear):
            # Slightly different from the TF version which uses truncated_normal for initialization
            # cf https://github.com/pytorch/pytorch/pull/5617
            nn.init.normal_(module.weight, mean=0.0, std=self.config.initializer_range)
            if module.bias is not None:
                nn.init.zeros_(module.bias)
        elif isinstance(module, nn.Embedding):
            nn.init.normal_(module.weight, mean=0.0, std=self.config.initializer_range)
            if module.padding_idx is not None:
                module.weight[module.padding_idx] = 0

mindnlp.transformers.models.deberta_v2.modeling_deberta_v2.DisentangledSelfAttention

Bases: Module

Disentangled self-attention module

PARAMETER	DESCRIPTION
config	A model config class instance with the configuration to build a new model. The schema is similar to BertConfig, for more details, please refer [DebertaV2Config] TYPE: `DebertaV2Config`

Source code in mindnlp\transformers\models\deberta_v2\modeling_deberta_v2.py

class DisentangledSelfAttention(nn.Module):
    """
    Disentangled self-attention module

    Parameters:
        config (`DebertaV2Config`):
            A model config class instance with the configuration to build a new model. The schema is similar to
            *BertConfig*, for more details, please refer [`DebertaV2Config`]

    """

    def __init__(self, config):
        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
        _attention_head_size = config.hidden_size // config.num_attention_heads
        self.attention_head_size = getattr(config, "attention_head_size", _attention_head_size)
        self.all_head_size = self.num_attention_heads * self.attention_head_size
        self.query_proj = nn.Linear(config.hidden_size, self.all_head_size, bias=True)
        self.key_proj = nn.Linear(config.hidden_size, self.all_head_size, bias=True)
        self.value_proj = nn.Linear(config.hidden_size, self.all_head_size, bias=True)

        self.share_att_key = getattr(config, "share_att_key", False)
        self.pos_att_type = config.pos_att_type if config.pos_att_type is not None else []
        self.relative_attention = getattr(config, "relative_attention", False)

        if self.relative_attention:
            self.position_buckets = getattr(config, "position_buckets", -1)
            self.max_relative_positions = getattr(config, "max_relative_positions", -1)
            if self.max_relative_positions < 1:
                self.max_relative_positions = config.max_position_embeddings
            self.pos_ebd_size = self.max_relative_positions
            if self.position_buckets > 0:
                self.pos_ebd_size = self.position_buckets

            self.pos_dropout = StableDropout(config.hidden_dropout_prob)

            if not self.share_att_key:
                if "c2p" in self.pos_att_type:
                    self.pos_key_proj = nn.Linear(config.hidden_size, self.all_head_size, bias=True)
                if "p2c" in self.pos_att_type:
                    self.pos_query_proj = nn.Linear(config.hidden_size, self.all_head_size)

        self.dropout = StableDropout(config.attention_probs_dropout_prob)

    def transpose_for_scores(self, x, attention_heads):
        new_x_shape = x.shape[:-1] + (attention_heads, -1)
        x = x.view(new_x_shape)
        return x.permute(0, 2, 1, 3).view(-1, x.shape[1], x.shape[-1])

    def forward(
        self,
        hidden_states,
        attention_mask,
        output_attentions=False,
        query_states=None,
        relative_pos=None,
        rel_embeddings=None,
    ):
        """
        Call the module

        Args:
            hidden_states (`mindspore.Tensor`):
                Input states to the module usually the output from previous layer, it will be the Q,K and V in
                *Attention(Q,K,V)*

            attention_mask (`mindspore.Tensor`):
                An attention mask matrix of shape [*B*, *N*, *N*] where *B* is the batch size, *N* is the maximum
                sequence length in which element [i,j] = *1* means the *i* th token in the input can attend to the *j*
                th token.

            output_attentions (`bool`, *optional*):
                Whether return the attention matrix.

            query_states (`mindspore.Tensor`, *optional*):
                The *Q* state in *Attention(Q,K,V)*.

            relative_pos (`mindspore.Tensor`):
                The relative position encoding between the tokens in the sequence. It's of shape [*B*, *N*, *N*] with
                values ranging in [*-max_relative_positions*, *max_relative_positions*].

            rel_embeddings (`mindspore.Tensor`):
                The embedding of relative distances. It's a tensor of shape [\\(2 \\times
                \\text{max_relative_positions}\\), *hidden_size*].


        """
        if query_states is None:
            query_states = hidden_states
        query_layer = self.transpose_for_scores(self.query_proj(query_states), self.num_attention_heads)
        key_layer = self.transpose_for_scores(self.key_proj(hidden_states), self.num_attention_heads)
        value_layer = self.transpose_for_scores(self.value_proj(hidden_states), self.num_attention_heads)

        rel_att = None
        # Take the dot product between "query" and "key" to get the raw attention scores.
        scale_factor = 1
        if "c2p" in self.pos_att_type:
            scale_factor += 1
        if "p2c" in self.pos_att_type:
            scale_factor += 1
        scale = ops.sqrt(mindspore.tensor(query_layer.shape[-1], dtype=mindspore.float32) * scale_factor)
        attention_scores = ops.bmm(query_layer, ops.transpose(key_layer, -1, -2) / scale.to(dtype=query_layer.dtype))
        if self.relative_attention:
            rel_embeddings = self.pos_dropout(rel_embeddings)
            rel_att = self.disentangled_attention_bias(
                query_layer, key_layer, relative_pos, rel_embeddings, scale_factor
            )

        if rel_att is not None:
            attention_scores = attention_scores + rel_att
        attention_scores = attention_scores.view(
            -1, self.num_attention_heads, attention_scores.shape[-2], attention_scores.shape[-1]
        )

        # bsz x height x length x dimension
        attention_probs = XSoftmax(-1)(attention_scores, attention_mask)
        attention_probs = self.dropout(attention_probs)
        context_layer = ops.bmm(
            attention_probs.view(-1, attention_probs.shape[-2], attention_probs.shape[-1]), value_layer
        )
        context_layer = (
            context_layer.view(-1, self.num_attention_heads, context_layer.shape[-2], context_layer.shape[-1])
            .permute(0, 2, 1, 3)
        )
        new_context_layer_shape = context_layer.shape[:-2] + (-1,)
        context_layer = context_layer.view(new_context_layer_shape)
        if output_attentions:
            return (context_layer, attention_probs)
        else:
            return context_layer

    def disentangled_attention_bias(self, query_layer, key_layer, relative_pos, rel_embeddings, scale_factor):
        if relative_pos is None:
            q = query_layer.shape[-2]
            relative_pos = build_relative_position(
                q,
                key_layer.shape[-2],
                bucket_size=self.position_buckets,
                max_position=self.max_relative_positions,
            )
        if relative_pos.dim() == 2:
            relative_pos = relative_pos.unsqueeze(0).unsqueeze(0)
        elif relative_pos.dim() == 3:
            relative_pos = relative_pos.unsqueeze(1)
        # bsz x height x query x key
        elif relative_pos.dim() != 4:
            raise ValueError(f"Relative position ids must be of dim 2 or 3 or 4. {relative_pos.dim()}")

        att_span = self.pos_ebd_size
        relative_pos = relative_pos.long()

        rel_embeddings = rel_embeddings[0 : att_span * 2, :].unsqueeze(0)
        if self.share_att_key:
            pos_query_layer = self.transpose_for_scores(
                self.query_proj(rel_embeddings), self.num_attention_heads
            ).repeat(query_layer.shape[0] // self.num_attention_heads, 1, 1)
            pos_key_layer = self.transpose_for_scores(self.key_proj(rel_embeddings), self.num_attention_heads).repeat(
                query_layer.shape[0] // self.num_attention_heads, 1, 1
            )
        else:
            if "c2p" in self.pos_att_type:
                pos_key_layer = self.transpose_for_scores(
                    self.pos_key_proj(rel_embeddings), self.num_attention_heads
                ).repeat(query_layer.shape[0] // self.num_attention_heads, 1, 1)  # .split(self.all_head_size, dim=-1)
            if "p2c" in self.pos_att_type:
                pos_query_layer = self.transpose_for_scores(
                    self.pos_query_proj(rel_embeddings), self.num_attention_heads
                ).repeat(query_layer.shape[0] // self.num_attention_heads, 1, 1)  # .split(self.all_head_size, dim=-1)

        score = 0
        # content->position
        if "c2p" in self.pos_att_type:
            scale = ops.sqrt(mindspore.tensor(pos_key_layer.shape[-1], dtype=mindspore.float32) * scale_factor)
            c2p_att = ops.bmm(query_layer, ops.transpose(pos_key_layer, -1, -2))
            c2p_pos = ops.clamp(relative_pos + att_span, 0, att_span * 2 - 1)
            c2p_att = ops.gather(
                c2p_att,
                dim=-1,
                index=c2p_pos.squeeze(0).broadcast_to([query_layer.shape[0], query_layer.shape[1], relative_pos.shape[-1]]),
            )
            score += c2p_att / scale.to(dtype=c2p_att.dtype)

        # position->content
        if "p2c" in self.pos_att_type:
            scale = ops.sqrt(mindspore.tensor(pos_query_layer.shape[-1], dtype=mindspore.float32) * scale_factor)
            if key_layer.shape[-2] != query_layer.shape[-2]:
                r_pos = build_relative_position(
                    key_layer.shape[-2],
                    key_layer.shape[-2],
                    bucket_size=self.position_buckets,
                    max_position=self.max_relative_positions,
                )
                r_pos = r_pos.unsqueeze(0)
            else:
                r_pos = relative_pos

            p2c_pos = ops.clamp(-r_pos + att_span, 0, att_span * 2 - 1)
            p2c_att = ops.bmm(key_layer, ops.transpose(pos_query_layer, -1, -2))
            p2c_att = ops.transpose(ops.gather(
                p2c_att,
                dim=-1,
                index=p2c_pos.squeeze(0).broadcast_to([query_layer.shape[0], key_layer.shape[-2], key_layer.shape[-2]]),
            ), -1, -2)
            score += p2c_att / scale.to(dtype=p2c_att.dtype)

        return score

mindnlp.transformers.models.deberta_v2.modeling_deberta_v2.DisentangledSelfAttention.forward(hidden_states, attention_mask, output_attentions=False, query_states=None, relative_pos=None, rel_embeddings=None)

Call the module

PARAMETER	DESCRIPTION
hidden_states	Input states to the module usually the output from previous layer, it will be the Q,K and V in Attention(Q,K,V) TYPE: `mindspore.Tensor`
attention_mask	An attention mask matrix of shape [B, N, N] where B is the batch size, N is the maximum sequence length in which element [i,j] = 1 means the i th token in the input can attend to the j th token. TYPE: `mindspore.Tensor`
output_attentions	Whether return the attention matrix. TYPE: `bool`, *optional* DEFAULT: False
query_states	The Q state in Attention(Q,K,V). TYPE: `mindspore.Tensor`, *optional* DEFAULT: None
relative_pos	The relative position encoding between the tokens in the sequence. It's of shape [B, N, N] with values ranging in [-max_relative_positions, max_relative_positions]. TYPE: `mindspore.Tensor` DEFAULT: None
rel_embeddings	The embedding of relative distances. It's a tensor of shape [\(2 \times \text{max_relative_positions}\), hidden_size]. TYPE: `mindspore.Tensor` DEFAULT: None

Source code in mindnlp\transformers\models\deberta_v2\modeling_deberta_v2.py

def forward(
    self,
    hidden_states,
    attention_mask,
    output_attentions=False,
    query_states=None,
    relative_pos=None,
    rel_embeddings=None,
):
    """
    Call the module

    Args:
        hidden_states (`mindspore.Tensor`):
            Input states to the module usually the output from previous layer, it will be the Q,K and V in
            *Attention(Q,K,V)*

        attention_mask (`mindspore.Tensor`):
            An attention mask matrix of shape [*B*, *N*, *N*] where *B* is the batch size, *N* is the maximum
            sequence length in which element [i,j] = *1* means the *i* th token in the input can attend to the *j*
            th token.

        output_attentions (`bool`, *optional*):
            Whether return the attention matrix.

        query_states (`mindspore.Tensor`, *optional*):
            The *Q* state in *Attention(Q,K,V)*.

        relative_pos (`mindspore.Tensor`):
            The relative position encoding between the tokens in the sequence. It's of shape [*B*, *N*, *N*] with
            values ranging in [*-max_relative_positions*, *max_relative_positions*].

        rel_embeddings (`mindspore.Tensor`):
            The embedding of relative distances. It's a tensor of shape [\\(2 \\times
            \\text{max_relative_positions}\\), *hidden_size*].


    """
    if query_states is None:
        query_states = hidden_states
    query_layer = self.transpose_for_scores(self.query_proj(query_states), self.num_attention_heads)
    key_layer = self.transpose_for_scores(self.key_proj(hidden_states), self.num_attention_heads)
    value_layer = self.transpose_for_scores(self.value_proj(hidden_states), self.num_attention_heads)

    rel_att = None
    # Take the dot product between "query" and "key" to get the raw attention scores.
    scale_factor = 1
    if "c2p" in self.pos_att_type:
        scale_factor += 1
    if "p2c" in self.pos_att_type:
        scale_factor += 1
    scale = ops.sqrt(mindspore.tensor(query_layer.shape[-1], dtype=mindspore.float32) * scale_factor)
    attention_scores = ops.bmm(query_layer, ops.transpose(key_layer, -1, -2) / scale.to(dtype=query_layer.dtype))
    if self.relative_attention:
        rel_embeddings = self.pos_dropout(rel_embeddings)
        rel_att = self.disentangled_attention_bias(
            query_layer, key_layer, relative_pos, rel_embeddings, scale_factor
        )

    if rel_att is not None:
        attention_scores = attention_scores + rel_att
    attention_scores = attention_scores.view(
        -1, self.num_attention_heads, attention_scores.shape[-2], attention_scores.shape[-1]
    )

    # bsz x height x length x dimension
    attention_probs = XSoftmax(-1)(attention_scores, attention_mask)
    attention_probs = self.dropout(attention_probs)
    context_layer = ops.bmm(
        attention_probs.view(-1, attention_probs.shape[-2], attention_probs.shape[-1]), value_layer
    )
    context_layer = (
        context_layer.view(-1, self.num_attention_heads, context_layer.shape[-2], context_layer.shape[-1])
        .permute(0, 2, 1, 3)
    )
    new_context_layer_shape = context_layer.shape[:-2] + (-1,)
    context_layer = context_layer.view(new_context_layer_shape)
    if output_attentions:
        return (context_layer, attention_probs)
    else:
        return context_layer

mindnlp.transformers.models.deberta_v2.modeling_deberta_v2.build_relative_position(query_size, key_size, bucket_size=-1, max_position=-1)

Build relative position according to the query and key

We assume the absolute position of query \(P_q\) is range from (0, query_size) and the absolute position of key \(P_k\) is range from (0, key_size), The relative positions from query to key is \(R_{q \rightarrow k} = P_q - P_k\)

PARAMETER	DESCRIPTION
query_size	the length of query TYPE: int
key_size	the length of key TYPE: int
bucket_size	the size of position bucket TYPE: int DEFAULT: -1
max_position	the maximum allowed absolute position TYPE: int DEFAULT: -1

Return

mindspore.Tensor: A tensor with shape [1, query_size, key_size]

Source code in mindnlp\transformers\models\deberta_v2\modeling_deberta_v2.py

def build_relative_position(query_size, key_size, bucket_size=-1, max_position=-1):
    """
    Build relative position according to the query and key

    We assume the absolute position of query \\(P_q\\) is range from (0, query_size) and the absolute position of key
    \\(P_k\\) is range from (0, key_size), The relative positions from query to key is \\(R_{q \\rightarrow k} = P_q -
    P_k\\)

    Args:
        query_size (int): the length of query
        key_size (int): the length of key
        bucket_size (int): the size of position bucket
        max_position (int): the maximum allowed absolute position

    Return:
        `mindspore.Tensor`: A tensor with shape [1, query_size, key_size]
    """

    q_ids = ops.arange(0, query_size)
    k_ids = ops.arange(0, key_size)
    rel_pos_ids = q_ids[:, None] - k_ids[None, :]
    if bucket_size > 0 and max_position > 0:
        rel_pos_ids = make_log_bucket_position(rel_pos_ids, bucket_size, max_position)
    rel_pos_ids = rel_pos_ids.to(mindspore.int64)
    rel_pos_ids = rel_pos_ids[:query_size, :]
    rel_pos_ids = rel_pos_ids.unsqueeze(0)
    return rel_pos_ids

mindnlp.transformers.models.deberta_v2.tokenization_deberta_v2

Tokenization class for model DeBERTa.

mindnlp.transformers.models.deberta_v2.tokenization_deberta_v2.DebertaV2Tokenizer

Bases: PreTrainedTokenizer

Constructs a DeBERTa-v2 tokenizer. Based on SentencePiece.

PARAMETER	DESCRIPTION
vocab_file	SentencePiece file (generally has a .spm extension) that contains the vocabulary necessary to instantiate a tokenizer. TYPE: `str`
do_lower_case	Whether or not to lowercase the input when tokenizing. TYPE: `bool`, *optional*, defaults to `False` DEFAULT: False
bos_token	The beginning of sequence token that was used during pre-training. Can be used a sequence classifier token. When building a sequence using special tokens, this is not the token that is used for the beginning of sequence. The token used is the cls_token. TYPE: `string`, *optional*, defaults to `"[CLS]"` DEFAULT: '[CLS]'
eos_token	The end of sequence token. When building a sequence using special tokens, this is not the token that is used for the end of sequence. The token used is the sep_token. TYPE: `string`, *optional*, defaults to `"[SEP]"` DEFAULT: '[SEP]'
unk_token	The unknown token. A token that is not in the vocabulary cannot be converted to an ID and is set to be this token instead. TYPE: `str`, *optional*, defaults to `"[UNK]"` DEFAULT: '[UNK]'
sep_token	The separator token, which is used when building a sequence from multiple sequences, e.g. two sequences for sequence classification or for a text and a question for question answering. It is also used as the last token of a sequence built with special tokens. TYPE: `str`, *optional*, defaults to `"[SEP]"` DEFAULT: '[SEP]'
pad_token	The token used for padding, for example when batching sequences of different lengths. TYPE: `str`, *optional*, defaults to `"[PAD]"` DEFAULT: '[PAD]'
cls_token	The classifier token which is used when doing sequence classification (classification of the whole sequence instead of per-token classification). It is the first token of the sequence when built with special tokens. TYPE: `str`, *optional*, defaults to `"[CLS]"` DEFAULT: '[CLS]'
mask_token	The token used for masking values. This is the token used when training this model with masked language modeling. This is the token which the model will try to predict. TYPE: `str`, *optional*, defaults to `"[MASK]"` DEFAULT: '[MASK]'
sp_model_kwargs	Will be passed to the SentencePieceProcessor.__init__() method. The Python wrapper for SentencePiece can be used, among other things, to set: enable_sampling: Enable subword regularization. nbest_size: Sampling parameters for unigram. Invalid for BPE-Dropout. nbest_size = {0,1}: No sampling is performed. nbest_size > 1: samples from the nbest_size results. nbest_size < 0: assuming that nbest_size is infinite and samples from the all hypothesis (lattice) using forward-filtering-and-backward-sampling algorithm. alpha: Smoothing parameter for unigram sampling, and dropout probability of merge operations for BPE-dropout. TYPE: `dict`, *optional* DEFAULT: None

Source code in mindnlp\transformers\models\deberta_v2\tokenization_deberta_v2.py

class DebertaV2Tokenizer(PreTrainedTokenizer):
    r"""
    Constructs a DeBERTa-v2 tokenizer. Based on [SentencePiece](https://github.com/google/sentencepiece).

    Args:
        vocab_file (`str`):
            [SentencePiece](https://github.com/google/sentencepiece) file (generally has a *.spm* extension) that
            contains the vocabulary necessary to instantiate a tokenizer.
        do_lower_case (`bool`, *optional*, defaults to `False`):
            Whether or not to lowercase the input when tokenizing.
        bos_token (`string`, *optional*, defaults to `"[CLS]"`):
            The beginning of sequence token that was used during pre-training. Can be used a sequence classifier token.
            When building a sequence using special tokens, this is not the token that is used for the beginning of
            sequence. The token used is the `cls_token`.
        eos_token (`string`, *optional*, defaults to `"[SEP]"`):
            The end of sequence token. When building a sequence using special tokens, this is not the token that is
            used for the end of sequence. The token used is the `sep_token`.
        unk_token (`str`, *optional*, defaults to `"[UNK]"`):
            The unknown token. A token that is not in the vocabulary cannot be converted to an ID and is set to be this
            token instead.
        sep_token (`str`, *optional*, defaults to `"[SEP]"`):
            The separator token, which is used when building a sequence from multiple sequences, e.g. two sequences for
            sequence classification or for a text and a question for question answering. It is also used as the last
            token of a sequence built with special tokens.
        pad_token (`str`, *optional*, defaults to `"[PAD]"`):
            The token used for padding, for example when batching sequences of different lengths.
        cls_token (`str`, *optional*, defaults to `"[CLS]"`):
            The classifier token which is used when doing sequence classification (classification of the whole sequence
            instead of per-token classification). It is the first token of the sequence when built with special tokens.
        mask_token (`str`, *optional*, defaults to `"[MASK]"`):
            The token used for masking values. This is the token used when training this model with masked language
            modeling. This is the token which the model will try to predict.
        sp_model_kwargs (`dict`, *optional*):
            Will be passed to the `SentencePieceProcessor.__init__()` method. The [Python wrapper for
            SentencePiece](https://github.com/google/sentencepiece/tree/master/python) can be used, among other things,
            to set:

            - `enable_sampling`: Enable subword regularization.
            - `nbest_size`: Sampling parameters for unigram. Invalid for BPE-Dropout.

              - `nbest_size = {0,1}`: No sampling is performed.
              - `nbest_size > 1`: samples from the nbest_size results.
              - `nbest_size < 0`: assuming that nbest_size is infinite and samples from the all hypothesis (lattice)
                using forward-filtering-and-backward-sampling algorithm.

            - `alpha`: Smoothing parameter for unigram sampling, and dropout probability of merge operations for
              BPE-dropout.
    """

    vocab_files_names = VOCAB_FILES_NAMES

    def __init__(
        self,
        vocab_file,
        do_lower_case=False,
        split_by_punct=False,
        bos_token="[CLS]",
        eos_token="[SEP]",
        unk_token="[UNK]",
        sep_token="[SEP]",
        pad_token="[PAD]",
        cls_token="[CLS]",
        mask_token="[MASK]",
        sp_model_kwargs: Optional[Dict[str, Any]] = None,
        **kwargs,
    ) -> None:
        self.sp_model_kwargs = {} if sp_model_kwargs is None else sp_model_kwargs

        if not os.path.isfile(vocab_file):
            raise ValueError(
                f"Can't find a vocabulary file at path '{vocab_file}'. To load the vocabulary from a Google pretrained"
                " model use `tokenizer = AutoTokenizer.from_pretrained(PRETRAINED_MODEL_NAME)`"
            )
        self.do_lower_case = do_lower_case
        self.split_by_punct = split_by_punct
        self.vocab_file = vocab_file
        self._tokenizer = SPMTokenizer(
            vocab_file, None, split_by_punct=split_by_punct, sp_model_kwargs=self.sp_model_kwargs
        )
        unk_token = AddedToken(unk_token, normalized=True, special=True) if isinstance(unk_token, str) else unk_token
        super().__init__(
            do_lower_case=do_lower_case,
            bos_token=bos_token,
            eos_token=eos_token,
            unk_token=unk_token,
            sep_token=sep_token,
            pad_token=pad_token,
            cls_token=cls_token,
            mask_token=mask_token,
            split_by_punct=split_by_punct,
            sp_model_kwargs=self.sp_model_kwargs,
            **kwargs,
        )
        self._tokenizer.special_tokens = self.all_special_tokens

    @property
    def vocab_size(self):
        return len(self.vocab)

    @property
    def vocab(self):
        return self._tokenizer.vocab

    def get_vocab(self):
        vocab = self.vocab.copy()
        vocab.update(self.get_added_vocab())
        return vocab

    def _tokenize(self, text: str) -> List[str]:
        """Take as input a string and return a list of strings (tokens) for words/sub-words"""
        if self.do_lower_case:
            text = text.lower()
        return self._tokenizer.tokenize(text)

    def _convert_token_to_id(self, token):
        """Converts a token (str) in an id using the vocab."""
        return self._tokenizer.spm.PieceToId(token)

    def _convert_id_to_token(self, index):
        """Converts an index (integer) in a token (str) using the vocab."""
        return self._tokenizer.spm.IdToPiece(index) if index < self.vocab_size else self.unk_token

    def convert_tokens_to_string(self, tokens):
        """Converts a sequence of tokens (string) in a single string."""
        return self._tokenizer.decode(tokens)

    def build_inputs_with_special_tokens(self, token_ids_0, token_ids_1=None):
        """
        Build model inputs from a sequence or a pair of sequence for sequence classification tasks by concatenating and
        adding special tokens. A DeBERTa sequence has the following format:

        - single sequence: [CLS] X [SEP]
        - pair of sequences: [CLS] A [SEP] B [SEP]

        Args:
            token_ids_0 (`List[int]`):
                List of IDs to which the special tokens will be added.
            token_ids_1 (`List[int]`, *optional*):
                Optional second list of IDs for sequence pairs.

        Returns:
            `List[int]`: List of [input IDs](../glossary#input-ids) with the appropriate special tokens.
        """

        if token_ids_1 is None:
            return [self.cls_token_id] + token_ids_0 + [self.sep_token_id]
        cls = [self.cls_token_id]
        sep = [self.sep_token_id]
        return cls + token_ids_0 + sep + token_ids_1 + sep

    def get_special_tokens_mask(self, token_ids_0, token_ids_1=None, already_has_special_tokens=False):
        """
        Retrieves sequence ids from a token list that has no special tokens added. This method is called when adding
        special tokens using the tokenizer `prepare_for_model` or `encode_plus` methods.

        Args:
            token_ids_0 (`List[int]`):
                List of IDs.
            token_ids_1 (`List[int]`, *optional*):
                Optional second list of IDs for sequence pairs.
            already_has_special_tokens (`bool`, *optional*, defaults to `False`):
                Whether or not the token list is already formatted with special tokens for the model.

        Returns:
            `List[int]`: A list of integers in the range [0, 1]: 1 for a special token, 0 for a sequence token.
        """

        if already_has_special_tokens:
            return super().get_special_tokens_mask(
                token_ids_0=token_ids_0, token_ids_1=token_ids_1, already_has_special_tokens=True
            )

        if token_ids_1 is not None:
            return [1] + ([0] * len(token_ids_0)) + [1] + ([0] * len(token_ids_1)) + [1]
        return [1] + ([0] * len(token_ids_0)) + [1]

    def create_token_type_ids_from_sequences(self, token_ids_0, token_ids_1=None):
        """
        Create a mask from the two sequences passed to be used in a sequence-pair classification task. A DeBERTa
        sequence pair mask has the following format:

        ```
        0 0 0 0 0 0 0 0 0 0 0 1 1 1 1 1 1 1 1 1
        | first sequence    | second sequence |
        ```

        If `token_ids_1` is `None`, this method only returns the first portion of the mask (0s).

        Args:
            token_ids_0 (`List[int]`):
                List of IDs.
            token_ids_1 (`List[int]`, *optional*):
                Optional second list of IDs for sequence pairs.

        Returns:
            `List[int]`: List of [token type IDs](../glossary#token-type-ids) according to the given sequence(s).
        """
        sep = [self.sep_token_id]
        cls = [self.cls_token_id]
        if token_ids_1 is None:
            return len(cls + token_ids_0 + sep) * [0]
        return len(cls + token_ids_0 + sep) * [0] + len(token_ids_1 + sep) * [1]

    def prepare_for_tokenization(self, text, is_split_into_words=False, **kwargs):
        add_prefix_space = kwargs.pop("add_prefix_space", False)
        if is_split_into_words or add_prefix_space:
            text = " " + text
        return (text, kwargs)

    def save_vocabulary(self, save_directory: str, filename_prefix: Optional[str] = None) -> Tuple[str]:
        return self._tokenizer.save_pretrained(save_directory, filename_prefix=filename_prefix)

mindnlp.transformers.models.deberta_v2.tokenization_deberta_v2.DebertaV2Tokenizer.build_inputs_with_special_tokens(token_ids_0, token_ids_1=None)

Build model inputs from a sequence or a pair of sequence for sequence classification tasks by concatenating and adding special tokens. A DeBERTa sequence has the following format:

• single sequence: [CLS] X [SEP]
• pair of sequences: [CLS] A [SEP] B [SEP]

PARAMETER	DESCRIPTION
token_ids_0	List of IDs to which the special tokens will be added. TYPE: `List[int]`
token_ids_1	Optional second list of IDs for sequence pairs. TYPE: `List[int]`, *optional* DEFAULT: None

RETURNS	DESCRIPTION
	List[int]: List of input IDs with the appropriate special tokens.

Source code in mindnlp\transformers\models\deberta_v2\tokenization_deberta_v2.py

def build_inputs_with_special_tokens(self, token_ids_0, token_ids_1=None):
    """
    Build model inputs from a sequence or a pair of sequence for sequence classification tasks by concatenating and
    adding special tokens. A DeBERTa sequence has the following format:

    - single sequence: [CLS] X [SEP]
    - pair of sequences: [CLS] A [SEP] B [SEP]

    Args:
        token_ids_0 (`List[int]`):
            List of IDs to which the special tokens will be added.
        token_ids_1 (`List[int]`, *optional*):
            Optional second list of IDs for sequence pairs.

    Returns:
        `List[int]`: List of [input IDs](../glossary#input-ids) with the appropriate special tokens.
    """

    if token_ids_1 is None:
        return [self.cls_token_id] + token_ids_0 + [self.sep_token_id]
    cls = [self.cls_token_id]
    sep = [self.sep_token_id]
    return cls + token_ids_0 + sep + token_ids_1 + sep

mindnlp.transformers.models.deberta_v2.tokenization_deberta_v2.DebertaV2Tokenizer.convert_tokens_to_string(tokens)

Converts a sequence of tokens (string) in a single string.

Source code in mindnlp\transformers\models\deberta_v2\tokenization_deberta_v2.py

def convert_tokens_to_string(self, tokens):
    """Converts a sequence of tokens (string) in a single string."""
    return self._tokenizer.decode(tokens)

mindnlp.transformers.models.deberta_v2.tokenization_deberta_v2.DebertaV2Tokenizer.create_token_type_ids_from_sequences(token_ids_0, token_ids_1=None)

Create a mask from the two sequences passed to be used in a sequence-pair classification task. A DeBERTa sequence pair mask has the following format:

0 0 0 0 0 0 0 0 0 0 0 1 1 1 1 1 1 1 1 1
| first sequence    | second sequence |

If token_ids_1 is None, this method only returns the first portion of the mask (0s).

PARAMETER	DESCRIPTION
token_ids_0	List of IDs. TYPE: `List[int]`
token_ids_1	Optional second list of IDs for sequence pairs. TYPE: `List[int]`, *optional* DEFAULT: None

RETURNS	DESCRIPTION
	List[int]: List of token type IDs according to the given sequence(s).

Source code in mindnlp\transformers\models\deberta_v2\tokenization_deberta_v2.py

def create_token_type_ids_from_sequences(self, token_ids_0, token_ids_1=None):
    """
    Create a mask from the two sequences passed to be used in a sequence-pair classification task. A DeBERTa
    sequence pair mask has the following format:

    ```
    0 0 0 0 0 0 0 0 0 0 0 1 1 1 1 1 1 1 1 1
    | first sequence    | second sequence |
    ```

    If `token_ids_1` is `None`, this method only returns the first portion of the mask (0s).

    Args:
        token_ids_0 (`List[int]`):
            List of IDs.
        token_ids_1 (`List[int]`, *optional*):
            Optional second list of IDs for sequence pairs.

    Returns:
        `List[int]`: List of [token type IDs](../glossary#token-type-ids) according to the given sequence(s).
    """
    sep = [self.sep_token_id]
    cls = [self.cls_token_id]
    if token_ids_1 is None:
        return len(cls + token_ids_0 + sep) * [0]
    return len(cls + token_ids_0 + sep) * [0] + len(token_ids_1 + sep) * [1]

mindnlp.transformers.models.deberta_v2.tokenization_deberta_v2.DebertaV2Tokenizer.get_special_tokens_mask(token_ids_0, token_ids_1=None, already_has_special_tokens=False)

Retrieves sequence ids from a token list that has no special tokens added. This method is called when adding special tokens using the tokenizer prepare_for_model or encode_plus methods.

PARAMETER	DESCRIPTION
token_ids_0	List of IDs. TYPE: `List[int]`
token_ids_1	Optional second list of IDs for sequence pairs. TYPE: `List[int]`, *optional* DEFAULT: None
already_has_special_tokens	Whether or not the token list is already formatted with special tokens for the model. TYPE: `bool`, *optional*, defaults to `False` DEFAULT: False

RETURNS	DESCRIPTION
	List[int]: A list of integers in the range [0, 1]: 1 for a special token, 0 for a sequence token.

Source code in mindnlp\transformers\models\deberta_v2\tokenization_deberta_v2.py

def get_special_tokens_mask(self, token_ids_0, token_ids_1=None, already_has_special_tokens=False):
    """
    Retrieves sequence ids from a token list that has no special tokens added. This method is called when adding
    special tokens using the tokenizer `prepare_for_model` or `encode_plus` methods.

    Args:
        token_ids_0 (`List[int]`):
            List of IDs.
        token_ids_1 (`List[int]`, *optional*):
            Optional second list of IDs for sequence pairs.
        already_has_special_tokens (`bool`, *optional*, defaults to `False`):
            Whether or not the token list is already formatted with special tokens for the model.

    Returns:
        `List[int]`: A list of integers in the range [0, 1]: 1 for a special token, 0 for a sequence token.
    """

    if already_has_special_tokens:
        return super().get_special_tokens_mask(
            token_ids_0=token_ids_0, token_ids_1=token_ids_1, already_has_special_tokens=True
        )

    if token_ids_1 is not None:
        return [1] + ([0] * len(token_ids_0)) + [1] + ([0] * len(token_ids_1)) + [1]
    return [1] + ([0] * len(token_ids_0)) + [1]

mindnlp.transformers.models.deberta_v2.tokenization_deberta_v2.SPMTokenizer

Constructs a tokenizer based on SentencePiece.

PARAMETER	DESCRIPTION
vocab_file	SentencePiece file (generally has a .spm extension) that contains the vocabulary necessary to instantiate a tokenizer. TYPE: `str`
sp_model_kwargs	Will be passed to the SentencePieceProcessor.__init__() method. The Python wrapper for SentencePiece can be used, among other things, to set: enable_sampling: Enable subword regularization. nbest_size: Sampling parameters for unigram. Invalid for BPE-Dropout. nbest_size = {0,1}: No sampling is performed. nbest_size > 1: samples from the nbest_size results. nbest_size < 0: assuming that nbest_size is infinite and samples from the all hypothesis (lattice) using forward-filtering-and-backward-sampling algorithm. alpha: Smoothing parameter for unigram sampling, and dropout probability of merge operations for BPE-dropout. TYPE: `dict`, *optional* DEFAULT: None

Source code in mindnlp\transformers\models\deberta_v2\tokenization_deberta_v2.py

class SPMTokenizer:
    r"""
    Constructs a tokenizer based on [SentencePiece](https://github.com/google/sentencepiece).

    Args:
        vocab_file (`str`):
            [SentencePiece](https://github.com/google/sentencepiece) file (generally has a *.spm* extension) that
            contains the vocabulary necessary to instantiate a tokenizer.
        sp_model_kwargs (`dict`, *optional*):
            Will be passed to the `SentencePieceProcessor.__init__()` method. The [Python wrapper for
            SentencePiece](https://github.com/google/sentencepiece/tree/master/python) can be used, among other things,
            to set:

            - `enable_sampling`: Enable subword regularization.
            - `nbest_size`: Sampling parameters for unigram. Invalid for BPE-Dropout.

              - `nbest_size = {0,1}`: No sampling is performed.
              - `nbest_size > 1`: samples from the nbest_size results.
              - `nbest_size < 0`: assuming that nbest_size is infinite and samples from the all hypothesis (lattice)
                using forward-filtering-and-backward-sampling algorithm.

            - `alpha`: Smoothing parameter for unigram sampling, and dropout probability of merge operations for
              BPE-dropout.
    """

    def __init__(
        self, vocab_file, special_tokens, split_by_punct=False, sp_model_kwargs: Optional[Dict[str, Any]] = None
    ):
        self.split_by_punct = split_by_punct
        self.vocab_file = vocab_file
        self.sp_model_kwargs = {} if sp_model_kwargs is None else sp_model_kwargs
        spm = sp.SentencePieceProcessor(**self.sp_model_kwargs)
        if not os.path.exists(vocab_file):
            raise FileNotFoundError(f"{vocab_file} does not exist!")
        spm.load(vocab_file)
        bpe_vocab_size = spm.GetPieceSize()
        # Token map
        # <unk> 0+1
        # <s> 1+1
        # </s> 2+1
        self.vocab = {spm.IdToPiece(i): i for i in range(bpe_vocab_size)}
        self.ids_to_tokens = [spm.IdToPiece(i) for i in range(bpe_vocab_size)]
        # self.vocab['[PAD]'] = 0
        # self.vocab['[CLS]'] = 1
        # self.vocab['[SEP]'] = 2
        # self.vocab['[UNK]'] = 3

        self.spm = spm
        self.special_tokens = special_tokens

    def __getstate__(self):
        state = self.__dict__.copy()
        state["spm"] = None
        return state

    def __setstate__(self, d):
        self.__dict__ = d

        # for backward compatibility
        if not hasattr(self, "sp_model_kwargs"):
            self.sp_model_kwargs = {}

        self.spm = sp.SentencePieceProcessor(**self.sp_model_kwargs)
        self.spm.Load(self.vocab_file)

    def tokenize(self, text):
        return self._encode_as_pieces(text)

    def convert_ids_to_tokens(self, ids):
        tokens = []
        for i in ids:
            tokens.append(self.ids_to_tokens[i])
        return tokens

    def decode(self, tokens, start=-1, end=-1, raw_text=None):
        if raw_text is None:
            current_sub_tokens = []
            out_string = ""
            prev_is_special = False
            for token in tokens:
                # make sure that special tokens are not decoded using sentencepiece model
                if token in self.special_tokens:
                    if not prev_is_special:
                        out_string += " "
                    out_string += self.spm.decode_pieces(current_sub_tokens) + token
                    prev_is_special = True
                    current_sub_tokens = []
                else:
                    current_sub_tokens.append(token)
                    prev_is_special = False
            out_string += self.spm.decode_pieces(current_sub_tokens)
            return out_string.strip()
        else:
            words = self.split_to_words(raw_text)
            word_tokens = [self.tokenize(w) for w in words]
            token2words = [0] * len(tokens)
            tid = 0
            for i, w in enumerate(word_tokens):
                for k, t in enumerate(w):
                    token2words[tid] = i
                    tid += 1
            word_start = token2words[start]
            word_end = token2words[end] if end < len(tokens) else len(words)
            text = "".join(words[word_start:word_end])
            return text

    # TODO add a deprecation cycle as this can have different behaviour from our API
    def add_special_token(self, token):
        if token not in self.special_tokens:
            self.special_tokens.append(token)
            if token not in self.vocab:
                self.vocab[token] = len(self.vocab) - 1
                self.ids_to_tokens.append(token)
        return self.id(token)

    def part_of_whole_word(self, token, is_bos=False):
        logger.warning_once(
            "The `DebertaTokenizer.part_of_whole_word` method is deprecated.35`"
        )
        if is_bos:
            return True
        if (
            len(token) == 1
            and (_is_whitespace(list(token)[0]) or _is_control(list(token)[0]) or _is_punctuation(list(token)[0]))
        ) or token in self.special_tokens:
            return False

        word_start = b"\xe2\x96\x81".decode("utf-8")
        return not token.startswith(word_start)

    def pad(self):
        return "[PAD]"

    def bos(self):
        return "[CLS]"

    def eos(self):
        return "[SEP]"

    def unk(self):
        return "[UNK]"

    def mask(self):
        return "[MASK]"

    def sym(self, id):
        return self.ids_to_tokens[id]

    def id(self, sym):
        logger.warning_once(
            "The `DebertaTokenizer.id` method is deprecated.35`"
        )
        return self.vocab[sym] if sym in self.vocab else 1

    def _encode_as_pieces(self, text):
        text = convert_to_unicode(text)
        if self.split_by_punct:
            words = self._run_split_on_punc(text)
            pieces = [self.spm.encode(w, out_type=str) for w in words]
            return [p for w in pieces for p in w]
        else:
            return self.spm.encode(text, out_type=str)

    def split_to_words(self, text):
        pieces = self._encode_as_pieces(text)
        word_start = b"\xe2\x96\x81".decode("utf-8")
        words = []
        offset = 0
        prev_end = 0
        for i, p in enumerate(pieces):
            if p.startswith(word_start):
                if offset > prev_end:
                    words.append(text[prev_end:offset])
                prev_end = offset
                w = p.replace(word_start, "")
            else:
                w = p
            try:
                s = text.index(w, offset)
                pn = ""
                k = i + 1
                while k < len(pieces):
                    pn = pieces[k].replace(word_start, "")
                    if len(pn) > 0:
                        break
                    k += 1

                if len(pn) > 0 and pn in text[offset:s]:
                    offset = offset + 1
                else:
                    offset = s + len(w)
            except Exception:
                offset = offset + 1

        if prev_end < offset:
            words.append(text[prev_end:offset])

        return words

    def _run_split_on_punc(self, text):
        """Splits punctuation on a piece of text."""
        chars = list(text)
        i = 0
        start_new_word = True
        output = []
        while i < len(chars):
            char = chars[i]
            if _is_punctuation(char):
                output.append([char])
                start_new_word = True
            else:
                if start_new_word:
                    output.append([])
                start_new_word = False
                output[-1].append(char)
            i += 1

        return ["".join(x) for x in output]

    def save_pretrained(self, path: str, filename_prefix: str = None):
        filename = VOCAB_FILES_NAMES[list(VOCAB_FILES_NAMES.keys())[0]]
        if filename_prefix is not None:
            filename = filename_prefix + "-" + filename
        full_path = os.path.join(path, filename)
        with open(full_path, "wb") as fs:
            fs.write(self.spm.serialized_model_proto())
        return (full_path,)

mindnlp.transformers.models.deberta_v2.tokenization_deberta_v2.convert_to_unicode(text)

Converts text to Unicode (if it's not already), assuming utf-8 input.

Source code in mindnlp\transformers\models\deberta_v2\tokenization_deberta_v2.py

def convert_to_unicode(text):
    """Converts `text` to Unicode (if it's not already), assuming utf-8 input."""
    if isinstance(text, str):
        return text
    elif isinstance(text, bytes):
        return text.decode("utf-8", "ignore")
    else:
        raise ValueError(f"Unsupported string type: {type(text)}")

mindnlp.transformers.models.deberta_v2.tokenization_deberta_v2_fast

Fast Tokenization class for model DeBERTa.

mindnlp.transformers.models.deberta_v2.tokenization_deberta_v2_fast.DebertaV2TokenizerFast

Bases: PreTrainedTokenizerFast

Constructs a DeBERTa-v2 fast tokenizer. Based on SentencePiece.

PARAMETER	DESCRIPTION
vocab_file	SentencePiece file (generally has a .spm extension) that contains the vocabulary necessary to instantiate a tokenizer. TYPE: `str` DEFAULT: None
do_lower_case	Whether or not to lowercase the input when tokenizing. TYPE: `bool`, *optional*, defaults to `False` DEFAULT: False
bos_token	The beginning of sequence token that was used during pre-training. Can be used a sequence classifier token. When building a sequence using special tokens, this is not the token that is used for the beginning of sequence. The token used is the cls_token. TYPE: `string`, *optional*, defaults to `"[CLS]"` DEFAULT: '[CLS]'
eos_token	The end of sequence token. When building a sequence using special tokens, this is not the token that is used for the end of sequence. The token used is the sep_token. TYPE: `string`, *optional*, defaults to `"[SEP]"` DEFAULT: '[SEP]'
unk_token	The unknown token. A token that is not in the vocabulary cannot be converted to an ID and is set to be this token instead. TYPE: `str`, *optional*, defaults to `"[UNK]"` DEFAULT: '[UNK]'
sep_token	The separator token, which is used when building a sequence from multiple sequences, e.g. two sequences for sequence classification or for a text and a question for question answering. It is also used as the last token of a sequence built with special tokens. TYPE: `str`, *optional*, defaults to `"[SEP]"` DEFAULT: '[SEP]'
pad_token	The token used for padding, for example when batching sequences of different lengths. TYPE: `str`, *optional*, defaults to `"[PAD]"` DEFAULT: '[PAD]'
cls_token	The classifier token which is used when doing sequence classification (classification of the whole sequence instead of per-token classification). It is the first token of the sequence when built with special tokens. TYPE: `str`, *optional*, defaults to `"[CLS]"` DEFAULT: '[CLS]'
mask_token	The token used for masking values. This is the token used when training this model with masked language modeling. This is the token which the model will try to predict. TYPE: `str`, *optional*, defaults to `"[MASK]"` DEFAULT: '[MASK]'
sp_model_kwargs	Will be passed to the SentencePieceProcessor.__init__() method. The Python wrapper for SentencePiece can be used, among other things, to set: enable_sampling: Enable subword regularization. nbest_size: Sampling parameters for unigram. Invalid for BPE-Dropout. nbest_size = {0,1}: No sampling is performed. nbest_size > 1: samples from the nbest_size results. nbest_size < 0: assuming that nbest_size is infinite and samples from the all hypothesis (lattice) using forward-filtering-and-backward-sampling algorithm. alpha: Smoothing parameter for unigram sampling, and dropout probability of merge operations for BPE-dropout. TYPE: `dict`, *optional*

Source code in mindnlp\transformers\models\deberta_v2\tokenization_deberta_v2_fast.py

class DebertaV2TokenizerFast(PreTrainedTokenizerFast):
    r"""
    Constructs a DeBERTa-v2 fast tokenizer. Based on [SentencePiece](https://github.com/google/sentencepiece).

    Args:
        vocab_file (`str`):
            [SentencePiece](https://github.com/google/sentencepiece) file (generally has a *.spm* extension) that
            contains the vocabulary necessary to instantiate a tokenizer.
        do_lower_case (`bool`, *optional*, defaults to `False`):
            Whether or not to lowercase the input when tokenizing.
        bos_token (`string`, *optional*, defaults to `"[CLS]"`):
            The beginning of sequence token that was used during pre-training. Can be used a sequence classifier token.
            When building a sequence using special tokens, this is not the token that is used for the beginning of
            sequence. The token used is the `cls_token`.
        eos_token (`string`, *optional*, defaults to `"[SEP]"`):
            The end of sequence token. When building a sequence using special tokens, this is not the token that is
            used for the end of sequence. The token used is the `sep_token`.
        unk_token (`str`, *optional*, defaults to `"[UNK]"`):
            The unknown token. A token that is not in the vocabulary cannot be converted to an ID and is set to be this
            token instead.
        sep_token (`str`, *optional*, defaults to `"[SEP]"`):
            The separator token, which is used when building a sequence from multiple sequences, e.g. two sequences for
            sequence classification or for a text and a question for question answering. It is also used as the last
            token of a sequence built with special tokens.
        pad_token (`str`, *optional*, defaults to `"[PAD]"`):
            The token used for padding, for example when batching sequences of different lengths.
        cls_token (`str`, *optional*, defaults to `"[CLS]"`):
            The classifier token which is used when doing sequence classification (classification of the whole sequence
            instead of per-token classification). It is the first token of the sequence when built with special tokens.
        mask_token (`str`, *optional*, defaults to `"[MASK]"`):
            The token used for masking values. This is the token used when training this model with masked language
            modeling. This is the token which the model will try to predict.
        sp_model_kwargs (`dict`, *optional*):
            Will be passed to the `SentencePieceProcessor.__init__()` method. The [Python wrapper for
            SentencePiece](https://github.com/google/sentencepiece/tree/master/python) can be used, among other things,
            to set:

            - `enable_sampling`: Enable subword regularization.
            - `nbest_size`: Sampling parameters for unigram. Invalid for BPE-Dropout.

              - `nbest_size = {0,1}`: No sampling is performed.
              - `nbest_size > 1`: samples from the nbest_size results.
              - `nbest_size < 0`: assuming that nbest_size is infinite and samples from the all hypothesis (lattice)
                using forward-filtering-and-backward-sampling algorithm.

            - `alpha`: Smoothing parameter for unigram sampling, and dropout probability of merge operations for
              BPE-dropout.
    """

    vocab_files_names = VOCAB_FILES_NAMES
    slow_tokenizer_class = DebertaV2Tokenizer

    def __init__(
        self,
        vocab_file=None,
        tokenizer_file=None,
        do_lower_case=False,
        split_by_punct=False,
        bos_token="[CLS]",
        eos_token="[SEP]",
        unk_token="[UNK]",
        sep_token="[SEP]",
        pad_token="[PAD]",
        cls_token="[CLS]",
        mask_token="[MASK]",
        **kwargs,
    ) -> None:
        super().__init__(
            vocab_file,
            tokenizer_file=tokenizer_file,
            do_lower_case=do_lower_case,
            bos_token=bos_token,
            eos_token=eos_token,
            unk_token=unk_token,
            sep_token=sep_token,
            pad_token=pad_token,
            cls_token=cls_token,
            mask_token=mask_token,
            split_by_punct=split_by_punct,
            **kwargs,
        )

        self.do_lower_case = do_lower_case
        self.split_by_punct = split_by_punct
        self.vocab_file = vocab_file

    @property
    def can_save_slow_tokenizer(self) -> bool:
        return os.path.isfile(self.vocab_file) if self.vocab_file else False

    def build_inputs_with_special_tokens(self, token_ids_0, token_ids_1=None):
        """
        Build model inputs from a sequence or a pair of sequence for sequence classification tasks by concatenating and
        adding special tokens. A DeBERTa sequence has the following format:

        - single sequence: [CLS] X [SEP]
        - pair of sequences: [CLS] A [SEP] B [SEP]

        Args:
            token_ids_0 (`List[int]`):
                List of IDs to which the special tokens will be added.
            token_ids_1 (`List[int]`, *optional*):
                Optional second list of IDs for sequence pairs.

        Returns:
            `List[int]`: List of [input IDs](../glossary#input-ids) with the appropriate special tokens.
        """

        if token_ids_1 is None:
            return [self.cls_token_id] + token_ids_0 + [self.sep_token_id]
        cls = [self.cls_token_id]
        sep = [self.sep_token_id]
        return cls + token_ids_0 + sep + token_ids_1 + sep

    def get_special_tokens_mask(self, token_ids_0, token_ids_1=None, already_has_special_tokens=False):
        """
        Retrieves sequence ids from a token list that has no special tokens added. This method is called when adding
        special tokens using the tokenizer `prepare_for_model` or `encode_plus` methods.

        Args:
            token_ids_0 (`List[int]`):
                List of IDs.
            token_ids_1 (`List[int]`, *optional*):
                Optional second list of IDs for sequence pairs.
            already_has_special_tokens (`bool`, *optional*, defaults to `False`):
                Whether or not the token list is already formatted with special tokens for the model.

        Returns:
            `List[int]`: A list of integers in the range [0, 1]: 1 for a special token, 0 for a sequence token.
        """

        if already_has_special_tokens:
            return super().get_special_tokens_mask(
                token_ids_0=token_ids_0, token_ids_1=token_ids_1, already_has_special_tokens=True
            )

        if token_ids_1 is not None:
            return [1] + ([0] * len(token_ids_0)) + [1] + ([0] * len(token_ids_1)) + [1]
        return [1] + ([0] * len(token_ids_0)) + [1]

    def create_token_type_ids_from_sequences(self, token_ids_0, token_ids_1=None):
        """
        Create a mask from the two sequences passed to be used in a sequence-pair classification task. A DeBERTa
        sequence pair mask has the following format:

        ```
        0 0 0 0 0 0 0 0 0 0 0 1 1 1 1 1 1 1 1 1
        | first sequence    | second sequence |
        ```

        If `token_ids_1` is `None`, this method only returns the first portion of the mask (0s).

        Args:
            token_ids_0 (`List[int]`):
                List of IDs.
            token_ids_1 (`List[int]`, *optional*):
                Optional second list of IDs for sequence pairs.

        Returns:
            `List[int]`: List of [token type IDs](../glossary#token-type-ids) according to the given sequence(s).
        """
        sep = [self.sep_token_id]
        cls = [self.cls_token_id]
        if token_ids_1 is None:
            return len(cls + token_ids_0 + sep) * [0]
        return len(cls + token_ids_0 + sep) * [0] + len(token_ids_1 + sep) * [1]

    def save_vocabulary(self, save_directory: str, filename_prefix: Optional[str] = None) -> Tuple[str]:
        if not self.can_save_slow_tokenizer:
            raise ValueError(
                "Your fast tokenizer does not have the necessary information to save the vocabulary for a slow "
                "tokenizer."
            )

        if not os.path.isdir(save_directory):
            logger.error(f"Vocabulary path ({save_directory}) should be a directory")
            return
        out_vocab_file = os.path.join(
            save_directory, (filename_prefix + "-" if filename_prefix else "") + VOCAB_FILES_NAMES["vocab_file"]
        )

        if os.path.abspath(self.vocab_file) != os.path.abspath(out_vocab_file):
            copyfile(self.vocab_file, out_vocab_file)

        return (out_vocab_file,)

mindnlp.transformers.models.deberta_v2.tokenization_deberta_v2_fast.DebertaV2TokenizerFast.build_inputs_with_special_tokens(token_ids_0, token_ids_1=None)

Build model inputs from a sequence or a pair of sequence for sequence classification tasks by concatenating and adding special tokens. A DeBERTa sequence has the following format:

• single sequence: [CLS] X [SEP]
• pair of sequences: [CLS] A [SEP] B [SEP]

PARAMETER	DESCRIPTION
token_ids_0	List of IDs to which the special tokens will be added. TYPE: `List[int]`
token_ids_1	Optional second list of IDs for sequence pairs. TYPE: `List[int]`, *optional* DEFAULT: None

RETURNS	DESCRIPTION
	List[int]: List of input IDs with the appropriate special tokens.

Source code in mindnlp\transformers\models\deberta_v2\tokenization_deberta_v2_fast.py

def build_inputs_with_special_tokens(self, token_ids_0, token_ids_1=None):
    """
    Build model inputs from a sequence or a pair of sequence for sequence classification tasks by concatenating and
    adding special tokens. A DeBERTa sequence has the following format:

    - single sequence: [CLS] X [SEP]
    - pair of sequences: [CLS] A [SEP] B [SEP]

    Args:
        token_ids_0 (`List[int]`):
            List of IDs to which the special tokens will be added.
        token_ids_1 (`List[int]`, *optional*):
            Optional second list of IDs for sequence pairs.

    Returns:
        `List[int]`: List of [input IDs](../glossary#input-ids) with the appropriate special tokens.
    """

    if token_ids_1 is None:
        return [self.cls_token_id] + token_ids_0 + [self.sep_token_id]
    cls = [self.cls_token_id]
    sep = [self.sep_token_id]
    return cls + token_ids_0 + sep + token_ids_1 + sep

mindnlp.transformers.models.deberta_v2.tokenization_deberta_v2_fast.DebertaV2TokenizerFast.create_token_type_ids_from_sequences(token_ids_0, token_ids_1=None)

Create a mask from the two sequences passed to be used in a sequence-pair classification task. A DeBERTa sequence pair mask has the following format:

0 0 0 0 0 0 0 0 0 0 0 1 1 1 1 1 1 1 1 1
| first sequence    | second sequence |

If token_ids_1 is None, this method only returns the first portion of the mask (0s).

PARAMETER	DESCRIPTION
token_ids_0	List of IDs. TYPE: `List[int]`
token_ids_1	Optional second list of IDs for sequence pairs. TYPE: `List[int]`, *optional* DEFAULT: None

RETURNS	DESCRIPTION
	List[int]: List of token type IDs according to the given sequence(s).

Source code in mindnlp\transformers\models\deberta_v2\tokenization_deberta_v2_fast.py

def create_token_type_ids_from_sequences(self, token_ids_0, token_ids_1=None):
    """
    Create a mask from the two sequences passed to be used in a sequence-pair classification task. A DeBERTa
    sequence pair mask has the following format:

    ```
    0 0 0 0 0 0 0 0 0 0 0 1 1 1 1 1 1 1 1 1
    | first sequence    | second sequence |
    ```

    If `token_ids_1` is `None`, this method only returns the first portion of the mask (0s).

    Args:
        token_ids_0 (`List[int]`):
            List of IDs.
        token_ids_1 (`List[int]`, *optional*):
            Optional second list of IDs for sequence pairs.

    Returns:
        `List[int]`: List of [token type IDs](../glossary#token-type-ids) according to the given sequence(s).
    """
    sep = [self.sep_token_id]
    cls = [self.cls_token_id]
    if token_ids_1 is None:
        return len(cls + token_ids_0 + sep) * [0]
    return len(cls + token_ids_0 + sep) * [0] + len(token_ids_1 + sep) * [1]

mindnlp.transformers.models.deberta_v2.tokenization_deberta_v2_fast.DebertaV2TokenizerFast.get_special_tokens_mask(token_ids_0, token_ids_1=None, already_has_special_tokens=False)

Retrieves sequence ids from a token list that has no special tokens added. This method is called when adding special tokens using the tokenizer prepare_for_model or encode_plus methods.

PARAMETER	DESCRIPTION
token_ids_0	List of IDs. TYPE: `List[int]`
token_ids_1	Optional second list of IDs for sequence pairs. TYPE: `List[int]`, *optional* DEFAULT: None
already_has_special_tokens	Whether or not the token list is already formatted with special tokens for the model. TYPE: `bool`, *optional*, defaults to `False` DEFAULT: False

RETURNS	DESCRIPTION
	List[int]: A list of integers in the range [0, 1]: 1 for a special token, 0 for a sequence token.

Source code in mindnlp\transformers\models\deberta_v2\tokenization_deberta_v2_fast.py

def get_special_tokens_mask(self, token_ids_0, token_ids_1=None, already_has_special_tokens=False):
    """
    Retrieves sequence ids from a token list that has no special tokens added. This method is called when adding
    special tokens using the tokenizer `prepare_for_model` or `encode_plus` methods.

    Args:
        token_ids_0 (`List[int]`):
            List of IDs.
        token_ids_1 (`List[int]`, *optional*):
            Optional second list of IDs for sequence pairs.
        already_has_special_tokens (`bool`, *optional*, defaults to `False`):
            Whether or not the token list is already formatted with special tokens for the model.

    Returns:
        `List[int]`: A list of integers in the range [0, 1]: 1 for a special token, 0 for a sequence token.
    """

    if already_has_special_tokens:
        return super().get_special_tokens_mask(
            token_ids_0=token_ids_0, token_ids_1=token_ids_1, already_has_special_tokens=True
        )

    if token_ids_1 is not None:
        return [1] + ([0] * len(token_ids_0)) + [1] + ([0] * len(token_ids_1)) + [1]
    return [1] + ([0] * len(token_ids_0)) + [1]