Skip to content

phi3

mindnlp.transformers.models.phi3.configuration_phi3

Phi-3 model configuration

mindnlp.transformers.models.phi3.configuration_phi3.Phi3Config

Bases: PretrainedConfig

This is the configuration class to store the configuration of a [Phi3Model]. It is used to instantiate a Phi-3 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 microsoft/Phi-3-mini-4k-instruct.

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 Phi-3 model. Defines the number of different tokens that can be represented by the inputs_ids passed when calling [Phi3Model].

TYPE: `int`, *optional*, defaults to 32064 DEFAULT: 32064

hidden_size

Dimension of the hidden representations.

TYPE: `int`, *optional*, defaults to 3072 DEFAULT: 3072

intermediate_size

Dimension of the MLP representations.

TYPE: `int`, *optional*, defaults to 8192 DEFAULT: 8192

num_hidden_layers

Number of hidden layers in the Transformer decoder.

TYPE: `int`, *optional*, defaults to 32 DEFAULT: 32

num_attention_heads

Number of attention heads for each attention layer in the Transformer decoder.

TYPE: `int`, *optional*, defaults to 32 DEFAULT: 32

num_key_value_heads

This is the number of key_value heads that should be used to implement Grouped Query Attention. If num_key_value_heads=num_attention_heads, the model will use Multi Head Attention (MHA), if num_key_value_heads=1 the model will use Multi Query Attention (MQA) otherwise GQA is used. When converting a multi-head checkpoint to a GQA checkpoint, each group key and value head should be forwarded by meanpooling all the original heads within that group. For more details checkout this paper. If it is not specified, will default to num_attention_heads.

TYPE: `int`, *optional* DEFAULT: None

resid_pdrop

Dropout probability for mlp outputs.

TYPE: `float`, *optional*, defaults to 0.0 DEFAULT: 0.0

embd_pdrop

The dropout ratio for the embeddings.

TYPE: `int`, *optional*, defaults to 0.0 DEFAULT: 0.0

attention_dropout

The dropout ratio after computing the attention scores.

TYPE: `float`, *optional*, defaults to 0.0 DEFAULT: 0.0

hidden_act

The non-linear activation function (function or string) in the decoder.

TYPE: `str` or `function`, *optional*, defaults to `"silu"` DEFAULT: 'silu'

max_position_embeddings

The maximum sequence length that this model might ever be used with.

TYPE: `int`, *optional*, defaults to 4096 DEFAULT: 4096

original_max_position_embeddings

The maximum sequence length that this model was trained with. This is used to determine the size of the original RoPE embeddings when using long scaling.

TYPE: `int`, *optional*, defaults to 4096 DEFAULT: 4096

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

rms_norm_eps

The epsilon value used for the RMSNorm.

TYPE: `float`, *optional*, defaults to 1e-05 DEFAULT: 1e-05

use_cache

Whether or not the model should return the last key/values attentions (not used by all models). Only relevant if config.is_decoder=True. Whether to tie weight embeddings or not.

TYPE: `bool`, *optional*, defaults to `True` DEFAULT: True

tie_word_embeddings

Whether to tie weight embeddings

TYPE: `bool`, *optional*, defaults to `False` DEFAULT: False

rope_theta

The base period of the RoPE embeddings.

TYPE: `float`, *optional*, defaults to 10000.0 DEFAULT: 10000.0

rope_scaling

The scaling strategy for the RoPE embeddings. If None, no scaling is applied. If a dictionary, it must contain the following keys: type, short_factor and long_factor. The type must be either su or yarn and the short_factor and long_factor must be lists of numbers with the same length as the hidden size divided by the number of attention heads divided by 2.

TYPE: `dict`, *optional* DEFAULT: None

bos_token_id

The id of the "beginning-of-sequence" token.

TYPE: `int`, *optional*, defaults to 1 DEFAULT: 1

eos_token_id

The id of the "end-of-sequence" token.

TYPE: `int`, *optional*, defaults to 32000 DEFAULT: 32000

pad_token_id

The id of the padding token.

TYPE: `int`, *optional*, defaults to 32000 DEFAULT: 32000

sliding_window

Sliding window attention window size. If None, no sliding window is applied.

TYPE: `int`, *optional* DEFAULT: None

Example 
>>> from transformers import Phi3Model, Phi3Config
...
>>> # Initializing a Phi-3 style configuration
>>> configuration = Phi3Config.from_pretrained("microsoft/Phi-3-mini-4k-instruct")
...
>>> # Initializing a model from the configuration
>>> model = Phi3Model(configuration)
...
>>> # Accessing the model configuration
>>> configuration = model.config
Source code in mindnlp/transformers/models/phi3/configuration_phi3.py 
 30
 31
 32
 33
 34
 35
 36
 37
 38
 39
 40
 41
 42
 43
 44
 45
 46
 47
 48
 49
 50
 51
 52
 53
 54
 55
 56
 57
 58
 59
 60
 61
 62
 63
 64
 65
 66
 67
 68
 69
 70
 71
 72
 73
 74
 75
 76
 77
 78
 79
 80
 81
 82
 83
 84
 85
 86
 87
 88
 89
 90
 91
 92
 93
 94
 95
 96
 97
 98
 99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
class Phi3Config(PretrainedConfig):
    r"""
    This is the configuration class to store the configuration of a [`Phi3Model`]. It is used to instantiate a Phi-3
    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
    [microsoft/Phi-3-mini-4k-instruct](https://huggingface.co/microsoft/Phi-3-mini-4k-instruct).

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

    Args:
        vocab_size (`int`, *optional*, defaults to 32064):
            Vocabulary size of the Phi-3 model. Defines the number of different tokens that can be represented by the
            `inputs_ids` passed when calling [`Phi3Model`].
        hidden_size (`int`, *optional*, defaults to 3072):
            Dimension of the hidden representations.
        intermediate_size (`int`, *optional*, defaults to 8192):
            Dimension of the MLP representations.
        num_hidden_layers (`int`, *optional*, defaults to 32):
            Number of hidden layers in the Transformer decoder.
        num_attention_heads (`int`, *optional*, defaults to 32):
            Number of attention heads for each attention layer in the Transformer decoder.
        num_key_value_heads (`int`, *optional*):
            This is the number of key_value heads that should be used to implement Grouped Query Attention. If
            `num_key_value_heads=num_attention_heads`, the model will use Multi Head Attention (MHA), if
            `num_key_value_heads=1` the model will use Multi Query Attention (MQA) otherwise GQA is used. When
            converting a multi-head checkpoint to a GQA checkpoint, each group key and value head should be forwarded
            by meanpooling all the original heads within that group. For more details checkout [this
            paper](https://arxiv.org/pdf/2305.13245.pdf). If it is not specified, will default to
            `num_attention_heads`.
        resid_pdrop (`float`, *optional*, defaults to 0.0):
            Dropout probability for mlp outputs.
        embd_pdrop (`int`, *optional*, defaults to 0.0):
            The dropout ratio for the embeddings.
        attention_dropout (`float`, *optional*, defaults to 0.0):
            The dropout ratio after computing the attention scores.
        hidden_act (`str` or `function`, *optional*, defaults to `"silu"`):
            The non-linear activation function (function or string) in the decoder.
        max_position_embeddings (`int`, *optional*, defaults to 4096):
            The maximum sequence length that this model might ever be used with.
        original_max_position_embeddings (`int`, *optional*, defaults to 4096):
            The maximum sequence length that this model was trained with. This is used to determine the size of the
            original RoPE embeddings when using long scaling.
        initializer_range (`float`, *optional*, defaults to 0.02):
            The standard deviation of the truncated_normal_initializer for initializing all weight matrices.
        rms_norm_eps (`float`, *optional*, defaults to 1e-05):
            The epsilon value used for the RMSNorm.
        use_cache (`bool`, *optional*, defaults to `True`):
            Whether or not the model should return the last key/values attentions (not used by all models). Only
            relevant if `config.is_decoder=True`. Whether to tie weight embeddings or not.
        tie_word_embeddings (`bool`, *optional*, defaults to `False`):
            Whether to tie weight embeddings
        rope_theta (`float`, *optional*, defaults to 10000.0):
            The base period of the RoPE embeddings.
        rope_scaling (`dict`, *optional*):
            The scaling strategy for the RoPE embeddings. If `None`, no scaling is applied. If a dictionary, it must
            contain the following keys: `type`, `short_factor` and `long_factor`. The `type` must be either `su` or `yarn` and
            the `short_factor` and `long_factor` must be lists of numbers with the same length as the hidden size
            divided by the number of attention heads divided by 2.
        bos_token_id (`int`, *optional*, defaults to 1):
            The id of the "beginning-of-sequence" token.
        eos_token_id (`int`, *optional*, defaults to 32000):
            The id of the "end-of-sequence" token.
        pad_token_id (`int`, *optional*, defaults to 32000):
            The id of the padding token.
        sliding_window (`int`, *optional*):
            Sliding window attention window size. If `None`, no sliding window is applied.

    Example:
        ```python
        >>> from transformers import Phi3Model, Phi3Config
        ...
        >>> # Initializing a Phi-3 style configuration
        >>> configuration = Phi3Config.from_pretrained("microsoft/Phi-3-mini-4k-instruct")
        ...
        >>> # Initializing a model from the configuration
        >>> model = Phi3Model(configuration)
        ...
        >>> # Accessing the model configuration
        >>> configuration = model.config
        ```
    """
    model_type = "phi3"
    keys_to_ignore_at_inference = ["past_key_values"]

    def __init__(
        self,
        vocab_size=32064,
        hidden_size=3072,
        intermediate_size=8192,
        num_hidden_layers=32,
        num_attention_heads=32,
        num_key_value_heads=None,
        resid_pdrop=0.0,
        embd_pdrop=0.0,
        attention_dropout=0.0,
        hidden_act="silu",
        max_position_embeddings=4096,
        original_max_position_embeddings=4096,
        initializer_range=0.02,
        rms_norm_eps=1e-5,
        use_cache=True,
        tie_word_embeddings=False,
        rope_theta=10000.0,
        rope_scaling=None,
        bos_token_id=1,
        eos_token_id=32000,
        pad_token_id=32000,
        sliding_window=None,
        **kwargs,
    ):
        """
        This method initializes an instance of the Phi3Config class with the provided parameters.

        Args:
            self: The instance of the class.
            vocab_size (int): The size of the vocabulary. Default is 32064.
            hidden_size (int): The size of the hidden layers in the model. Default is 3072.
            intermediate_size (int): The size of the intermediate layers in the model. Default is 8192.
            num_hidden_layers (int): The number of hidden layers in the model. Default is 32.
            num_attention_heads (int): The number of attention heads. Default is 32.
            num_key_value_heads (int): The number of key and value heads. If None, it defaults to num_attention_heads.
            resid_pdrop (float): The dropout probability for residual connections. Default is 0.0.
            embd_pdrop (float): The dropout probability for the embeddings. Default is 0.0.
            attention_dropout (float): The dropout probability for attention layers. Default is 0.0.
            hidden_act (str): The activation function for the hidden layers. Default is 'silu'.
            max_position_embeddings (int): The maximum position embeddings. Default is 4096.
            original_max_position_embeddings (int): The original maximum position embeddings. Default is 4096.
            initializer_range (float): The range for parameter initializations. Default is 0.02.
            rms_norm_eps (float): The epsilon value for RMS normalization. Default is 1e-05.
            use_cache (bool): Indicates whether caching is used. Default is True.
            tie_word_embeddings (bool): Indicates whether word embeddings are tied. Default is False.
            rope_theta (float): The theta value for ROPE. Default is 10000.0.
            rope_scaling (float): The scaling factor for ROPE.
            bos_token_id (int): The beginning of sequence token id. Default is 1.
            eos_token_id (int): The end of sequence token id. Default is 32000.
            pad_token_id (int): The padding token id. Default is 32000.
            sliding_window: Not specified.

        Returns:
            None.

        Raises:
            ValueError: If rope_scaling is provided without rope_theta or vice versa.
            TypeError: If any of the input parameters have an unexpected type.
        """
        self.vocab_size = vocab_size
        self.hidden_size = hidden_size
        self.intermediate_size = intermediate_size
        self.num_hidden_layers = num_hidden_layers
        self.num_attention_heads = num_attention_heads

        if num_key_value_heads is None:
            num_key_value_heads = num_attention_heads

        self.num_key_value_heads = num_key_value_heads
        self.resid_pdrop = resid_pdrop
        self.embd_pdrop = embd_pdrop
        self.attention_dropout = attention_dropout
        self.hidden_act = hidden_act
        self.max_position_embeddings = max_position_embeddings
        self.original_max_position_embeddings = original_max_position_embeddings
        self.initializer_range = initializer_range
        self.rms_norm_eps = rms_norm_eps
        self.use_cache = use_cache
        self.rope_theta = rope_theta
        self.rope_scaling = rope_scaling
        self._rope_scaling_validation()
        self.sliding_window = sliding_window

        super().__init__(
            bos_token_id=bos_token_id,
            eos_token_id=eos_token_id,
            pad_token_id=pad_token_id,
            tie_word_embeddings=tie_word_embeddings,
            **kwargs,
        )

    def _rope_scaling_validation(self):
        """
        Validate the `rope_scaling` configuration.
        """
        if self.rope_scaling is None:
            return

        if not isinstance(self.rope_scaling, dict) or len(self.rope_scaling) != 3:
            raise ValueError(
                "`rope_scaling` must be a dictionary with three fields, `type`, `short_factor` and `long_factor`, "
                f"got {self.rope_scaling}"
            )
        rope_scaling_type = self.rope_scaling.get("type", None)
        rope_scaling_short_factor = self.rope_scaling.get("short_factor", None)
        rope_scaling_long_factor = self.rope_scaling.get("long_factor", None)
        if rope_scaling_type is None or rope_scaling_type not in ["su", "yarn"]:
            raise ValueError(f"`rope_scaling`'s type field must be one of ['su', 'yarn'], got {rope_scaling_type}")
        if not (
            isinstance(rope_scaling_short_factor, list)
            and all(isinstance(x, (int, float)) for x in rope_scaling_short_factor)
        ):
            raise ValueError(
                f"`rope_scaling`'s short_factor field must be a list of numbers, got {rope_scaling_short_factor}"
            )
        if not len(rope_scaling_short_factor) == self.hidden_size // self.num_attention_heads // 2:
            raise ValueError(
                f"`rope_scaling`'s short_factor field must have length {self.hidden_size // self.num_attention_heads // 2}, got {len(rope_scaling_short_factor)}"
            )
        if not (
            isinstance(rope_scaling_long_factor, list)
            and all(isinstance(x, (int, float)) for x in rope_scaling_long_factor)
        ):
            raise ValueError(
                f"`rope_scaling`'s long_factor field must be a list of numbers, got {rope_scaling_long_factor}"
            )
        if not len(rope_scaling_long_factor) == self.hidden_size // self.num_attention_heads // 2:
            raise ValueError(
                f"`rope_scaling`'s long_factor field must have length {self.hidden_size // self.num_attention_heads // 2}, got {len(rope_scaling_long_factor)}"
            )

mindnlp.transformers.models.phi3.configuration_phi3.Phi3Config.__init__(vocab_size=32064, hidden_size=3072, intermediate_size=8192, num_hidden_layers=32, num_attention_heads=32, num_key_value_heads=None, resid_pdrop=0.0, embd_pdrop=0.0, attention_dropout=0.0, hidden_act='silu', max_position_embeddings=4096, original_max_position_embeddings=4096, initializer_range=0.02, rms_norm_eps=1e-05, use_cache=True, tie_word_embeddings=False, rope_theta=10000.0, rope_scaling=None, bos_token_id=1, eos_token_id=32000, pad_token_id=32000, sliding_window=None, **kwargs)

This method initializes an instance of the Phi3Config class with the provided parameters.

PARAMETER DESCRIPTION
self

The instance of the class.
vocab_size

The size of the vocabulary. Default is 32064.

TYPE: int DEFAULT: 32064

hidden_size

The size of the hidden layers in the model. Default is 3072.

TYPE: int DEFAULT: 3072

intermediate_size

The size of the intermediate layers in the model. Default is 8192.

TYPE: int DEFAULT: 8192

num_hidden_layers

The number of hidden layers in the model. Default is 32.

TYPE: int DEFAULT: 32

num_attention_heads

The number of attention heads. Default is 32.

TYPE: int DEFAULT: 32

num_key_value_heads

The number of key and value heads. If None, it defaults to num_attention_heads.

TYPE: int DEFAULT: None

resid_pdrop

The dropout probability for residual connections. Default is 0.0.

TYPE: float DEFAULT: 0.0

embd_pdrop

The dropout probability for the embeddings. Default is 0.0.

TYPE: float DEFAULT: 0.0

attention_dropout

The dropout probability for attention layers. Default is 0.0.

TYPE: float DEFAULT: 0.0

hidden_act

The activation function for the hidden layers. Default is 'silu'.

TYPE: str DEFAULT: 'silu'

max_position_embeddings

The maximum position embeddings. Default is 4096.

TYPE: int DEFAULT: 4096

original_max_position_embeddings

The original maximum position embeddings. Default is 4096.

TYPE: int DEFAULT: 4096

initializer_range

The range for parameter initializations. Default is 0.02.

TYPE: float DEFAULT: 0.02

rms_norm_eps

The epsilon value for RMS normalization. Default is 1e-05.

TYPE: float DEFAULT: 1e-05

use_cache

Indicates whether caching is used. Default is True.

TYPE: bool DEFAULT: True

tie_word_embeddings

Indicates whether word embeddings are tied. Default is False.

TYPE: bool DEFAULT: False

rope_theta

The theta value for ROPE. Default is 10000.0.

TYPE: float DEFAULT: 10000.0

rope_scaling

The scaling factor for ROPE.

TYPE: float DEFAULT: None

bos_token_id

The beginning of sequence token id. Default is 1.

TYPE: int DEFAULT: 1

eos_token_id

The end of sequence token id. Default is 32000.

TYPE: int DEFAULT: 32000

pad_token_id

The padding token id. Default is 32000.

TYPE: int DEFAULT: 32000

sliding_window

Not specified.

DEFAULT: None

RETURNS DESCRIPTION

None.
RAISES DESCRIPTION
ValueError

If rope_scaling is provided without rope_theta or vice versa.
TypeError

If any of the input parameters have an unexpected type.
Source code in mindnlp/transformers/models/phi3/configuration_phi3.py 
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
def __init__(
    self,
    vocab_size=32064,
    hidden_size=3072,
    intermediate_size=8192,
    num_hidden_layers=32,
    num_attention_heads=32,
    num_key_value_heads=None,
    resid_pdrop=0.0,
    embd_pdrop=0.0,
    attention_dropout=0.0,
    hidden_act="silu",
    max_position_embeddings=4096,
    original_max_position_embeddings=4096,
    initializer_range=0.02,
    rms_norm_eps=1e-5,
    use_cache=True,
    tie_word_embeddings=False,
    rope_theta=10000.0,
    rope_scaling=None,
    bos_token_id=1,
    eos_token_id=32000,
    pad_token_id=32000,
    sliding_window=None,
    **kwargs,
):
    """
    This method initializes an instance of the Phi3Config class with the provided parameters.

    Args:
        self: The instance of the class.
        vocab_size (int): The size of the vocabulary. Default is 32064.
        hidden_size (int): The size of the hidden layers in the model. Default is 3072.
        intermediate_size (int): The size of the intermediate layers in the model. Default is 8192.
        num_hidden_layers (int): The number of hidden layers in the model. Default is 32.
        num_attention_heads (int): The number of attention heads. Default is 32.
        num_key_value_heads (int): The number of key and value heads. If None, it defaults to num_attention_heads.
        resid_pdrop (float): The dropout probability for residual connections. Default is 0.0.
        embd_pdrop (float): The dropout probability for the embeddings. Default is 0.0.
        attention_dropout (float): The dropout probability for attention layers. Default is 0.0.
        hidden_act (str): The activation function for the hidden layers. Default is 'silu'.
        max_position_embeddings (int): The maximum position embeddings. Default is 4096.
        original_max_position_embeddings (int): The original maximum position embeddings. Default is 4096.
        initializer_range (float): The range for parameter initializations. Default is 0.02.
        rms_norm_eps (float): The epsilon value for RMS normalization. Default is 1e-05.
        use_cache (bool): Indicates whether caching is used. Default is True.
        tie_word_embeddings (bool): Indicates whether word embeddings are tied. Default is False.
        rope_theta (float): The theta value for ROPE. Default is 10000.0.
        rope_scaling (float): The scaling factor for ROPE.
        bos_token_id (int): The beginning of sequence token id. Default is 1.
        eos_token_id (int): The end of sequence token id. Default is 32000.
        pad_token_id (int): The padding token id. Default is 32000.
        sliding_window: Not specified.

    Returns:
        None.

    Raises:
        ValueError: If rope_scaling is provided without rope_theta or vice versa.
        TypeError: If any of the input parameters have an unexpected type.
    """
    self.vocab_size = vocab_size
    self.hidden_size = hidden_size
    self.intermediate_size = intermediate_size
    self.num_hidden_layers = num_hidden_layers
    self.num_attention_heads = num_attention_heads

    if num_key_value_heads is None:
        num_key_value_heads = num_attention_heads

    self.num_key_value_heads = num_key_value_heads
    self.resid_pdrop = resid_pdrop
    self.embd_pdrop = embd_pdrop
    self.attention_dropout = attention_dropout
    self.hidden_act = hidden_act
    self.max_position_embeddings = max_position_embeddings
    self.original_max_position_embeddings = original_max_position_embeddings
    self.initializer_range = initializer_range
    self.rms_norm_eps = rms_norm_eps
    self.use_cache = use_cache
    self.rope_theta = rope_theta
    self.rope_scaling = rope_scaling
    self._rope_scaling_validation()
    self.sliding_window = sliding_window

    super().__init__(
        bos_token_id=bos_token_id,
        eos_token_id=eos_token_id,
        pad_token_id=pad_token_id,
        tie_word_embeddings=tie_word_embeddings,
        **kwargs,
    )

mindnlp.transformers.models.phi3.modeling_phi3

MindSpore Phi-3 model.

mindnlp.transformers.models.phi3.modeling_phi3.Phi3Attention

Bases: Module

Multi-headed attention from 'Attention Is All You Need' paper

Source code in mindnlp/transformers/models/phi3/modeling_phi3.py 
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
class Phi3Attention(nn.Module):
    """Multi-headed attention from 'Attention Is All You Need' paper"""
    def __init__(self, config: Phi3Config, layer_idx: Optional[int] = None):
        """
        Initializes an instance of the `Phi3Attention` class.

        Args:
            self: The instance of the `Phi3Attention` class.
            config (Phi3Config): An instance of the `Phi3Config` class containing the configuration settings
                for the attention layer.
            layer_idx (Optional[int]): The index of the layer. Defaults to None.

        Returns:
            None

        Raises:
            ValueError: If `hidden_size` is not divisible by `num_heads`.

        Notes:
            - Instantiating `Phi3Attention` without passing a `layer_idx` is not recommended and may lead to
            errors during the forward call if caching is used. It is advised to provide a `layer_idx` when
            creating this class.
            - The `Phi3Attention` class expects `hidden_size` to be divisible by `num_heads`.

            - The following attributes are initialized within the `__init__` method:

                - `self.config`: An instance of the `Phi3Config` class containing the configuration settings.
                - `self.layer_idx`: The index of the layer.
                - `self.attention_dropout`: The dropout rate for attention.
                - `self.hidden_size`: The hidden size of the layer.
                - `self.num_heads`: The number of attention heads.
                - `self.head_dim`: The dimension of each attention head.
                - `self.num_key_value_heads`: The number of key-value attention heads.
                - `self.num_key_value_groups`: The number of groups formed by key-value attention heads.
                - `self.max_position_embeddings`: The maximum number of position embeddings.
                - `self.original_max_position_embeddings`: The original maximum number of position embeddings.
                - `self.rope_theta`: The theta value for relative position encoding.
                - `self.rope_scaling`: The scaling factor for relative position encoding.
                - `self.is_causal`: A boolean indicating if the attention is causal.
                - `self.o_proj`: A fully connected layer for projecting the output.
                - `self.qkv_proj`: A fully connected layer for projecting the queries, keys, and values.
                - `self._init_rope()`: A private method for initializing the relative position encoding.

        """
        super().__init__()
        self.config = config
        self.layer_idx = layer_idx
        if layer_idx is None:
            logger.warning_once(
                f"Instantiating {self.__class__.__name__} without passing a `layer_idx` is not recommended and will "
                "lead to errors during the forward call if caching is used. Please make sure to provide a `layer_idx` "
                "when creating this class."
            )

        self.attention_dropout = config.attention_dropout
        self.hidden_size = config.hidden_size
        self.num_heads = config.num_attention_heads
        self.head_dim = self.hidden_size // self.num_heads
        self.num_key_value_heads = config.num_key_value_heads
        self.num_key_value_groups = self.num_heads // self.num_key_value_heads
        self.max_position_embeddings = config.max_position_embeddings
        self.original_max_position_embeddings = config.original_max_position_embeddings
        self.rope_theta = config.rope_theta
        self.rope_scaling = config.rope_scaling
        self.is_causal = True

        if (self.head_dim * self.num_heads) != self.hidden_size:
            raise ValueError(
                f"hidden_size must be divisible by num_heads (got `hidden_size`: {self.hidden_size}"
                f" and `num_heads`: {self.num_heads})."
            )

        op_size = self.num_heads * self.head_dim + 2 * (self.num_key_value_heads * self.head_dim)
        self.o_proj = nn.Linear(self.num_heads * self.head_dim, self.hidden_size, bias=False)
        self.qkv_proj = nn.Linear(self.hidden_size, op_size, bias=False)
        self._init_rope()

    def _init_rope(self):
        """
        Initializes the RoPE (Rotary Positional Encoding) for the Phi3Attention class.

        Args:
            self: The instance of the Phi3Attention class.

        Returns:
            None

        Raises:
            ValueError: If the RoPE scaling type is unknown.

        This method initializes the RoPE based on the provided configurations. If the 'rope_scaling' attribute is None,
        it creates a Phi3RotaryEmbedding object with the specified parameters. Otherwise, it checks the type of scaling
        specified in the 'rope_scaling' attribute and creates the appropriate Phi3ScaledRotaryEmbedding object
        accordingly. The Phi3ScaledRotaryEmbedding objects provide additional scaling options for the Rotary Positional
        Encoding.

        The available scaling types are as follows:

        - 'su': Creates a Phi3SuScaledRotaryEmbedding object.
        - 'yarn': Creates a Phi3YarnScaledRotaryEmbedding object.

        Note:
            The Phi3SuScaledRotaryEmbedding and Phi3YarnScaledRotaryEmbedding classes are specific implementations of
            the Phi3RotaryEmbedding class with additional scaling capabilities.

        Example:
            ```python()
            >>> # Initialize RoPE without scaling
            >>> _init_rope()
            ...
            >>> # Initialize RoPE with 'su' scaling
            >>> self.rope_scaling = {'type': 'su'}
            >>> _init_rope()
            ...
            >>> # Initialize RoPE with 'yarn' scaling
            >>> self.rope_scaling = {'type': 'yarn'}
            >>> _init_rope()
            ```
        """
        if self.rope_scaling is None:
            self.rotary_emb = Phi3RotaryEmbedding(
                self.head_dim,
                max_position_embeddings=self.max_position_embeddings,
                base=self.rope_theta,
            )
        else:
            scaling_type = self.config.rope_scaling["type"]
            if scaling_type == "su":
                self.rotary_emb = Phi3SuScaledRotaryEmbedding(self.head_dim, self.config)
            elif scaling_type == "yarn":
                self.rotary_emb = Phi3YarnScaledRotaryEmbedding(self.head_dim, self.config)
            else:
                raise ValueError(f"Unknown RoPE scaling type {scaling_type}")

    def forward(
        self,
        hidden_states: mindspore.Tensor,
        attention_mask: Optional[mindspore.Tensor] = None,
        position_ids: Optional[mindspore.Tensor] = None,
        past_key_value: Optional[Cache] = None,
        output_attentions: bool = False,
        use_cache: bool = False,
    ) -> Tuple[mindspore.Tensor, Optional[mindspore.Tensor], Optional[Tuple[mindspore.Tensor]]]:
        """
        This method forwards the Phi3Attention mechanism.

        Args:
            self: The instance of the Phi3Attention class.
            hidden_states (mindspore.Tensor): The input hidden states tensor of shape
                (batch_size, sequence_length, hidden_size).
            attention_mask (Optional[mindspore.Tensor]): An optional mask tensor of shape
                (batch_size, 1, sequence_length, sequence_length) to mask some positions in the input.
            position_ids (Optional[mindspore.Tensor]): An optional tensor of shape (batch_size, sequence_length)
                containing the position indices.
            past_key_value (Optional[Cache]): An optional cache storing the past key and value states for efficient
                auto-regressive decoding.
            output_attentions (bool): A flag indicating whether to output the attention weights.
            use_cache (bool): A flag indicating whether to use the cache for storing key and value states.

        Returns:
            Tuple[mindspore.Tensor, Optional[mindspore.Tensor], Optional[Tuple[mindspore.Tensor]]]: A tuple containing
                the attention output tensor of shape (batch_size, sequence_length, hidden_size),
                optionally the attention weights tensor, and optionally the updated past key and value states.

        Raises:
            ValueError: Raised if the cache structure has changed, if the attention weights or mask tensors have
                incorrect shapes, or if the output tensors have unexpected shapes.
        """
        bsz, q_len, _ = hidden_states.shape

        qkv = self.qkv_proj(hidden_states)
        query_pos = self.num_heads * self.head_dim
        query_states = qkv[..., :query_pos]
        key_states = qkv[..., query_pos : query_pos + self.num_key_value_heads * self.head_dim]
        value_states = qkv[..., query_pos + self.num_key_value_heads * self.head_dim :]

        query_states = query_states.view(bsz, q_len, self.num_heads, self.head_dim).swapaxes(1, 2)
        key_states = key_states.view(bsz, q_len, self.num_key_value_heads, self.head_dim).swapaxes(1, 2)
        value_states = value_states.view(bsz, q_len, self.num_key_value_heads, self.head_dim).swapaxes(1, 2)

        kv_seq_len = key_states.shape[-2]
        if past_key_value is not None:
            if self.layer_idx is None:
                raise ValueError(
                    f"The cache structure has changed since version v4.36. If you are using {self.__class__.__name__} "
                    "for auto-regressive decoding with k/v caching, please make sure to initialize the attention class "
                    "with a layer index."
                )
            kv_seq_len += past_key_value.get_usable_length(kv_seq_len, self.layer_idx)
        cos, sin = self.rotary_emb(value_states, position_ids, seq_len=kv_seq_len)

        query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin, position_ids)

        if past_key_value is not None:
            cache_kwargs = {"sin": sin, "cos": cos}  # Specific to RoPE models
            key_states, value_states = past_key_value.update(key_states, value_states, self.layer_idx, cache_kwargs)

        # repeat k/v heads if n_kv_heads < n_heads
        key_states = repeat_kv(key_states, self.num_key_value_groups)
        value_states = repeat_kv(value_states, self.num_key_value_groups)

        attn_weights = ops.matmul(query_states, key_states.swapaxes(2, 3)) / math.sqrt(self.head_dim)

        if attn_weights.shape != (bsz, self.num_heads, q_len, kv_seq_len):
            raise ValueError(
                f"Attention weights should be of size {(bsz, self.num_heads, q_len, kv_seq_len)}, but is"
                f" {attn_weights.shape}"
            )

        if attention_mask is not None:
            if attention_mask.shape != (bsz, 1, q_len, kv_seq_len):
                raise ValueError(
                    f"Attention mask should be of size {(bsz, 1, q_len, kv_seq_len)}, but is {attention_mask.shape}"
                )
            attn_weights = attn_weights + attention_mask

        # upcast attention to fp32
        attn_weights = ops.softmax(attn_weights, axis=-1, dtype=mindspore.float32).to(value_states.dtype)
        attn_weights = ops.dropout(attn_weights, p=self.attention_dropout, training=self.training)

        attn_output = ops.matmul(attn_weights, value_states)

        if attn_output.shape != (bsz, self.num_heads, q_len, self.head_dim):
            raise ValueError(
                f"`attn_output` should be of size {(bsz, self.num_heads, q_len, self.head_dim)}, but is"
                f" {attn_output.shape}"
            )

        attn_output = attn_output.swapaxes(1, 2)
        attn_output = attn_output.reshape(bsz, q_len, self.hidden_size)

        attn_output = self.o_proj(attn_output)

        if not output_attentions:
            attn_weights = None

        return attn_output, attn_weights, past_key_value

mindnlp.transformers.models.phi3.modeling_phi3.Phi3Attention.__init__(config, layer_idx=None)

Initializes an instance of the Phi3Attention class.

PARAMETER DESCRIPTION
self

The instance of the Phi3Attention class.
config

An instance of the Phi3Config class containing the configuration settings for the attention layer.

TYPE: Phi3Config

layer_idx

The index of the layer. Defaults to None.

TYPE: Optional[int] DEFAULT: None

RETURNS DESCRIPTION

None
RAISES DESCRIPTION
ValueError

If hidden_size is not divisible by num_heads.
Notes
  • Instantiating Phi3Attention without passing a layer_idx is not recommended and may lead to errors during the forward call if caching is used. It is advised to provide a layer_idx when creating this class.

  • The Phi3Attention class expects hidden_size to be divisible by num_heads.

  • The following attributes are initialized within the __init__ method:

    • self.config: An instance of the Phi3Config class containing the configuration settings.
    • self.layer_idx: The index of the layer.
    • self.attention_dropout: The dropout rate for attention.
    • self.hidden_size: The hidden size of the layer.
    • self.num_heads: The number of attention heads.
    • self.head_dim: The dimension of each attention head.
    • self.num_key_value_heads: The number of key-value attention heads.
    • self.num_key_value_groups: The number of groups formed by key-value attention heads.
    • self.max_position_embeddings: The maximum number of position embeddings.
    • self.original_max_position_embeddings: The original maximum number of position embeddings.
    • self.rope_theta: The theta value for relative position encoding.
    • self.rope_scaling: The scaling factor for relative position encoding.
    • self.is_causal: A boolean indicating if the attention is causal.
    • self.o_proj: A fully connected layer for projecting the output.
    • self.qkv_proj: A fully connected layer for projecting the queries, keys, and values.
    • self._init_rope(): A private method for initializing the relative position encoding.
Source code in mindnlp/transformers/models/phi3/modeling_phi3.py 
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
def __init__(self, config: Phi3Config, layer_idx: Optional[int] = None):
    """
    Initializes an instance of the `Phi3Attention` class.

    Args:
        self: The instance of the `Phi3Attention` class.
        config (Phi3Config): An instance of the `Phi3Config` class containing the configuration settings
            for the attention layer.
        layer_idx (Optional[int]): The index of the layer. Defaults to None.

    Returns:
        None

    Raises:
        ValueError: If `hidden_size` is not divisible by `num_heads`.

    Notes:
        - Instantiating `Phi3Attention` without passing a `layer_idx` is not recommended and may lead to
        errors during the forward call if caching is used. It is advised to provide a `layer_idx` when
        creating this class.
        - The `Phi3Attention` class expects `hidden_size` to be divisible by `num_heads`.

        - The following attributes are initialized within the `__init__` method:

            - `self.config`: An instance of the `Phi3Config` class containing the configuration settings.
            - `self.layer_idx`: The index of the layer.
            - `self.attention_dropout`: The dropout rate for attention.
            - `self.hidden_size`: The hidden size of the layer.
            - `self.num_heads`: The number of attention heads.
            - `self.head_dim`: The dimension of each attention head.
            - `self.num_key_value_heads`: The number of key-value attention heads.
            - `self.num_key_value_groups`: The number of groups formed by key-value attention heads.
            - `self.max_position_embeddings`: The maximum number of position embeddings.
            - `self.original_max_position_embeddings`: The original maximum number of position embeddings.
            - `self.rope_theta`: The theta value for relative position encoding.
            - `self.rope_scaling`: The scaling factor for relative position encoding.
            - `self.is_causal`: A boolean indicating if the attention is causal.
            - `self.o_proj`: A fully connected layer for projecting the output.
            - `self.qkv_proj`: A fully connected layer for projecting the queries, keys, and values.
            - `self._init_rope()`: A private method for initializing the relative position encoding.

    """
    super().__init__()
    self.config = config
    self.layer_idx = layer_idx
    if layer_idx is None:
        logger.warning_once(
            f"Instantiating {self.__class__.__name__} without passing a `layer_idx` is not recommended and will "
            "lead to errors during the forward call if caching is used. Please make sure to provide a `layer_idx` "
            "when creating this class."
        )

    self.attention_dropout = config.attention_dropout
    self.hidden_size = config.hidden_size
    self.num_heads = config.num_attention_heads
    self.head_dim = self.hidden_size // self.num_heads
    self.num_key_value_heads = config.num_key_value_heads
    self.num_key_value_groups = self.num_heads // self.num_key_value_heads
    self.max_position_embeddings = config.max_position_embeddings
    self.original_max_position_embeddings = config.original_max_position_embeddings
    self.rope_theta = config.rope_theta
    self.rope_scaling = config.rope_scaling
    self.is_causal = True

    if (self.head_dim * self.num_heads) != self.hidden_size:
        raise ValueError(
            f"hidden_size must be divisible by num_heads (got `hidden_size`: {self.hidden_size}"
            f" and `num_heads`: {self.num_heads})."
        )

    op_size = self.num_heads * self.head_dim + 2 * (self.num_key_value_heads * self.head_dim)
    self.o_proj = nn.Linear(self.num_heads * self.head_dim, self.hidden_size, bias=False)
    self.qkv_proj = nn.Linear(self.hidden_size, op_size, bias=False)
    self._init_rope()

mindnlp.transformers.models.phi3.modeling_phi3.Phi3Attention.forward(hidden_states, attention_mask=None, position_ids=None, past_key_value=None, output_attentions=False, use_cache=False)

This method forwards the Phi3Attention mechanism.

PARAMETER DESCRIPTION
self

The instance of the Phi3Attention class.
hidden_states

The input hidden states tensor of shape (batch_size, sequence_length, hidden_size).

TYPE: Tensor

attention_mask

An optional mask tensor of shape (batch_size, 1, sequence_length, sequence_length) to mask some positions in the input.

TYPE: Optional[Tensor] DEFAULT: None

position_ids

An optional tensor of shape (batch_size, sequence_length) containing the position indices.

TYPE: Optional[Tensor] DEFAULT: None

past_key_value

An optional cache storing the past key and value states for efficient auto-regressive decoding.

TYPE: Optional[Cache] DEFAULT: None

output_attentions

A flag indicating whether to output the attention weights.

TYPE: bool DEFAULT: False

use_cache

A flag indicating whether to use the cache for storing key and value states.

TYPE: bool DEFAULT: False

RETURNS DESCRIPTION
Tuple[Tensor, Optional[Tensor], Optional[Tuple[Tensor]]]

Tuple[mindspore.Tensor, Optional[mindspore.Tensor], Optional[Tuple[mindspore.Tensor]]]: A tuple containing the attention output tensor of shape (batch_size, sequence_length, hidden_size), optionally the attention weights tensor, and optionally the updated past key and value states.
RAISES DESCRIPTION
ValueError

Raised if the cache structure has changed, if the attention weights or mask tensors have incorrect shapes, or if the output tensors have unexpected shapes.
Source code in mindnlp/transformers/models/phi3/modeling_phi3.py 
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
def forward(
    self,
    hidden_states: mindspore.Tensor,
    attention_mask: Optional[mindspore.Tensor] = None,
    position_ids: Optional[mindspore.Tensor] = None,
    past_key_value: Optional[Cache] = None,
    output_attentions: bool = False,
    use_cache: bool = False,
) -> Tuple[mindspore.Tensor, Optional[mindspore.Tensor], Optional[Tuple[mindspore.Tensor]]]:
    """
    This method forwards the Phi3Attention mechanism.

    Args:
        self: The instance of the Phi3Attention class.
        hidden_states (mindspore.Tensor): The input hidden states tensor of shape
            (batch_size, sequence_length, hidden_size).
        attention_mask (Optional[mindspore.Tensor]): An optional mask tensor of shape
            (batch_size, 1, sequence_length, sequence_length) to mask some positions in the input.
        position_ids (Optional[mindspore.Tensor]): An optional tensor of shape (batch_size, sequence_length)
            containing the position indices.
        past_key_value (Optional[Cache]): An optional cache storing the past key and value states for efficient
            auto-regressive decoding.
        output_attentions (bool): A flag indicating whether to output the attention weights.
        use_cache (bool): A flag indicating whether to use the cache for storing key and value states.

    Returns:
        Tuple[mindspore.Tensor, Optional[mindspore.Tensor], Optional[Tuple[mindspore.Tensor]]]: A tuple containing
            the attention output tensor of shape (batch_size, sequence_length, hidden_size),
            optionally the attention weights tensor, and optionally the updated past key and value states.

    Raises:
        ValueError: Raised if the cache structure has changed, if the attention weights or mask tensors have
            incorrect shapes, or if the output tensors have unexpected shapes.
    """
    bsz, q_len, _ = hidden_states.shape

    qkv = self.qkv_proj(hidden_states)
    query_pos = self.num_heads * self.head_dim
    query_states = qkv[..., :query_pos]
    key_states = qkv[..., query_pos : query_pos + self.num_key_value_heads * self.head_dim]
    value_states = qkv[..., query_pos + self.num_key_value_heads * self.head_dim :]

    query_states = query_states.view(bsz, q_len, self.num_heads, self.head_dim).swapaxes(1, 2)
    key_states = key_states.view(bsz, q_len, self.num_key_value_heads, self.head_dim).swapaxes(1, 2)
    value_states = value_states.view(bsz, q_len, self.num_key_value_heads, self.head_dim).swapaxes(1, 2)

    kv_seq_len = key_states.shape[-2]
    if past_key_value is not None:
        if self.layer_idx is None:
            raise ValueError(
                f"The cache structure has changed since version v4.36. If you are using {self.__class__.__name__} "
                "for auto-regressive decoding with k/v caching, please make sure to initialize the attention class "
                "with a layer index."
            )
        kv_seq_len += past_key_value.get_usable_length(kv_seq_len, self.layer_idx)
    cos, sin = self.rotary_emb(value_states, position_ids, seq_len=kv_seq_len)

    query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin, position_ids)

    if past_key_value is not None:
        cache_kwargs = {"sin": sin, "cos": cos}  # Specific to RoPE models
        key_states, value_states = past_key_value.update(key_states, value_states, self.layer_idx, cache_kwargs)

    # repeat k/v heads if n_kv_heads < n_heads
    key_states = repeat_kv(key_states, self.num_key_value_groups)
    value_states = repeat_kv(value_states, self.num_key_value_groups)

    attn_weights = ops.matmul(query_states, key_states.swapaxes(2, 3)) / math.sqrt(self.head_dim)

    if attn_weights.shape != (bsz, self.num_heads, q_len, kv_seq_len):
        raise ValueError(
            f"Attention weights should be of size {(bsz, self.num_heads, q_len, kv_seq_len)}, but is"
            f" {attn_weights.shape}"
        )

    if attention_mask is not None:
        if attention_mask.shape != (bsz, 1, q_len, kv_seq_len):
            raise ValueError(
                f"Attention mask should be of size {(bsz, 1, q_len, kv_seq_len)}, but is {attention_mask.shape}"
            )
        attn_weights = attn_weights + attention_mask

    # upcast attention to fp32
    attn_weights = ops.softmax(attn_weights, axis=-1, dtype=mindspore.float32).to(value_states.dtype)
    attn_weights = ops.dropout(attn_weights, p=self.attention_dropout, training=self.training)

    attn_output = ops.matmul(attn_weights, value_states)

    if attn_output.shape != (bsz, self.num_heads, q_len, self.head_dim):
        raise ValueError(
            f"`attn_output` should be of size {(bsz, self.num_heads, q_len, self.head_dim)}, but is"
            f" {attn_output.shape}"
        )

    attn_output = attn_output.swapaxes(1, 2)
    attn_output = attn_output.reshape(bsz, q_len, self.hidden_size)

    attn_output = self.o_proj(attn_output)

    if not output_attentions:
        attn_weights = None

    return attn_output, attn_weights, past_key_value

mindnlp.transformers.models.phi3.modeling_phi3.Phi3DecoderLayer

Bases: Module

Phi3DecoderLayer represents a single layer of the Phi3 decoder. This layer includes self-attention, residual connections, layer normalization, and a multi-layer perceptron (MLP) sublayer.

This class inherits from the nn.Module class and is designed to be used as a building block for forwarding Phi3 decoder models.

The init method initializes the Phi3DecoderLayer with the provided configuration and layer index. It sets up the self-attention mechanism, MLP, input layer normalization, and dropout layers.

The forward method processes the input hidden states through the layer. It applies input layer normalization, self-attention, residual connections, post-attention layer normalization, and the MLP sublayer. The method also handles optional arguments such as attention_mask, position_ids, past_key_value, output_attentions, and use_cache, and returns the resulting hidden states along with optional outputs based on the provided arguments.

Note

The forward method also issues a warning if the 'padding_mask' argument is used, as it is deprecated and will be removed in a future version in favor of 'attention_mask'.

PARAMETER DESCRIPTION
hidden_states

Input to the layer of shape (batch, seq_len, embed_dim).

TYPE: Tensor

attention_mask

Attention mask of size (batch, 1, tgt_len, src_len) where padding elements are indicated by very large negative values.

TYPE: mindspore.Tensor, *optional*

position_ids

Indices of positions of each input sequence tokens in the position embeddings. Selected in the range [0, config.n_positions - 1]. What are position IDs?

TYPE: mindspore.Tensor of shape `({0})`, *optional*

output_attentions

Whether or not to return the attentions tensors of all attention layers. See attentions under returned tensors for more detail.

TYPE: bool, *optional*

use_cache

If set to True, past_key_values key value states are returned and can be used to speed up decoding (see past_key_values).

TYPE: bool, *optional*

past_key_value

Cached past key and value projection states

TYPE: Tuple(mindspore.Tensor), *optional*

RETURNS DESCRIPTION

Tuple[mindspore.Tensor, Optional[Tuple[mindspore.Tensor, mindspore.Tensor]]]: The resulting hidden states, and optionally, the self-attention weights and present key-value states if requested.
Source code in mindnlp/transformers/models/phi3/modeling_phi3.py 
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
877
878
879
880
881
882
883
884
885
886
887
888
889
890
891
892
893
894
895
896
897
898
899
900
901
902
903
904
905
906
907
908
909
910
911
912
913
914
915
916
917
918
919
920
921
922
923
924
925
926
927
928
929
930
931
932
933
934
935
936
937
938
939
940
941
942
943
944
945
946
947
948
949
950
951
952
953
954
955
956
957
958
959
960
961
962
963
964
class Phi3DecoderLayer(nn.Module):

    '''
    Phi3DecoderLayer represents a single layer of the Phi3 decoder. This layer includes self-attention, residual
    connections, layer normalization, and a multi-layer perceptron (MLP) sublayer.

    This class inherits from the nn.Module class and is designed to be used as a building block for forwarding Phi3
    decoder models.

    The __init__ method initializes the Phi3DecoderLayer with the provided configuration and layer index.
    It sets up the self-attention mechanism, MLP, input layer normalization, and dropout layers.

    The forward method processes the input hidden states through the layer. It applies input layer normalization,
    self-attention, residual connections, post-attention layer normalization, and the MLP sublayer. The method also
    handles optional arguments such as attention_mask, position_ids, past_key_value, output_attentions, and use_cache,
    and returns the resulting hidden states along with optional outputs based on the provided arguments.

    Note:
        The forward method also issues a warning if the 'padding_mask' argument is used, as it is deprecated and
        will be removed in a future version in favor of 'attention_mask'.

    Args:
        hidden_states (mindspore.Tensor):
            Input to the layer of shape `(batch, seq_len, embed_dim)`.
        attention_mask (mindspore.Tensor, *optional*):
            Attention mask of size `(batch, 1, tgt_len, src_len)` where padding elements are indicated by very large
            negative values.
        position_ids (mindspore.Tensor of shape `({0})`, *optional*):
            Indices of positions of each input sequence tokens in the position embeddings. Selected in the range
            `[0, config.n_positions - 1]`. [What are position IDs?](../glossary#position-ids)
        output_attentions (bool, *optional*):
            Whether or not to return the attentions tensors of all attention layers. See `attentions` under returned
            tensors for more detail.
        use_cache (bool, *optional*):
            If set to `True`, `past_key_values` key value states are returned and can be used to speed up decoding
            (see `past_key_values`).
        past_key_value (Tuple(mindspore.Tensor), *optional*):
            Cached past key and value projection states

    Returns:
        Tuple[mindspore.Tensor, Optional[Tuple[mindspore.Tensor, mindspore.Tensor]]]:
            The resulting hidden states, and optionally, the self-attention weights and present key-value states
            if requested.
    '''
    def __init__(self, config: Phi3Config, layer_idx: int):
        """
        Initializes a new instance of the Phi3DecoderLayer class.

        Args:
            self (Phi3DecoderLayer): The current instance of the Phi3DecoderLayer class.
            config (Phi3Config): The configuration object containing parameters for the decoder layer.
            layer_idx (int): The index of the decoder layer.

        Returns:
            None.

        Raises:
            None.

        Description:
            This method initializes the Phi3DecoderLayer object with the provided configuration and layer index.
            It sets up the self-attention mechanism, multi-layer perceptron, input layer normalization, and other
            components required for the decoder layer.

            - config: The Phi3Config object that contains the configuration parameters for the decoder layer.
            This includes parameters such as hidden size, dropout rate, and RMS normalization epsilon.

            - layer_idx: An integer representing the index of the decoder layer.
            This index is used to identify the layer and is required for initializing the self-attention mechanism.

            The method does not return any value.
        """
        super().__init__()

        self.config = config
        self.self_attn = PHI3_ATTENTION_CLASSES['eager'](config, layer_idx=layer_idx)

        self.mlp = Phi3MLP(config)
        self.input_layernorm = Phi3RMSNorm(config.hidden_size, eps=config.rms_norm_eps)

        self.resid_attn_dropout = nn.Dropout(config.resid_pdrop)
        self.resid_mlp_dropout = nn.Dropout(config.resid_pdrop)
        self.post_attention_layernorm = Phi3RMSNorm(config.hidden_size, eps=config.rms_norm_eps)

    def forward(
        self,
        hidden_states: mindspore.Tensor,
        attention_mask: Optional[mindspore.Tensor] = None,
        position_ids: Optional[mindspore.Tensor] = None,
        past_key_value: Optional[Tuple[mindspore.Tensor]] = None,
        output_attentions: Optional[bool] = False,
        use_cache: Optional[bool] = False,
        **kwargs,
    ) -> Tuple[mindspore.Tensor, Optional[Tuple[mindspore.Tensor, mindspore.Tensor]]]:
        '''
        Constructs a Phi3DecoderLayer object.

        Args:
            self: The object itself.
            hidden_states (mindspore.Tensor): Input to the layer of shape `(batch, seq_len, embed_dim)`.
            attention_mask (mindspore.Tensor, optional): Attention mask of size `(batch, 1, tgt_len, src_len)`,
                where padding elements are indicated by very large negative values.
            position_ids (mindspore.Tensor, optional): Indices of positions of each input sequence tokens in the
                position embeddings. Selected in the range `[0, config.n_positions - 1]`. (default: None)
            past_key_value (Tuple[mindspore.Tensor], optional): Cached past key and value projection states.
                (default: None)
            output_attentions (bool, optional): Whether or not to return the attentions tensors of all attention layers.
                See `attentions` under returned tensors for more detail. (default: False)
            use_cache (bool, optional): If set to True, `past_key_values` key value states are returned and can be used
                to speed up decoding (see `past_key_values`). (default: False)

        Returns:
            Tuple[mindspore.Tensor, Optional[Tuple[mindspore.Tensor, mindspore.Tensor]]]: A tuple containing the
                hidden states of shape `(batch, seq_len, embed_dim)`. Optionally, the tuple may also contain the
                attentions tensors of all attention layers and the cached past key and value projection states.

        Raises:
            None.
        '''
        if "padding_mask" in kwargs:
            warnings.warn(
                "Passing `padding_mask` is deprecated and will be removed in v4.37. Please make sure use `attention_mask` instead.`"
            )
        """
        Args:
            hidden_states (`mindspore.Tensor`):
                input to the layer of shape `(batch, seq_len, embed_dim)`
            attention_mask (`mindspore.Tensor`, *optional*): attention mask of size
                `(batch, 1, tgt_len, src_len)` where padding elements are indicated by very large negative values.
            position_ids (`mindspore.Tensor` of shape `({0})`, *optional*):
                Indices of positions of each input sequence tokens in the position embeddings. Selected in the range
                `[0, config.n_positions - 1]`. [What are position IDs?](../glossary#position-ids)
            output_attentions (`bool`, *optional*):
                Whether or not to return the attentions tensors of all attention layers. See `attentions` under
                returned tensors for more detail.
            use_cache (`bool`, *optional*):
                If set to `True`, `past_key_values` key value states are returned and can be used to speed up decoding
                (see `past_key_values`).
            past_key_value (`Tuple(mindspore.Tensor)`, *optional*): cached past key and value projection states
        """

        residual = hidden_states

        hidden_states = self.input_layernorm(hidden_states)

        # Self Attention
        attn_outputs, self_attn_weights, present_key_value = self.self_attn(
            hidden_states=hidden_states,
            attention_mask=attention_mask,
            position_ids=position_ids,
            past_key_value=past_key_value,
            output_attentions=output_attentions,
            use_cache=use_cache,
        )

        hidden_states = residual + self.resid_attn_dropout(attn_outputs)

        residual = hidden_states
        hidden_states = self.post_attention_layernorm(hidden_states)
        hidden_states = self.mlp(hidden_states)
        hidden_states = residual + self.resid_mlp_dropout(hidden_states)

        outputs = (hidden_states,)

        if output_attentions:
            outputs += (self_attn_weights,)

        if use_cache:
            outputs += (present_key_value,)

        return outputs

mindnlp.transformers.models.phi3.modeling_phi3.Phi3DecoderLayer.__init__(config, layer_idx)

Initializes a new instance of the Phi3DecoderLayer class.

PARAMETER DESCRIPTION
self

The current instance of the Phi3DecoderLayer class.

TYPE: Phi3DecoderLayer

config

The configuration object containing parameters for the decoder layer.

TYPE: Phi3Config

layer_idx

The index of the decoder layer.

TYPE: int

RETURNS DESCRIPTION

None.
Description

This method initializes the Phi3DecoderLayer object with the provided configuration and layer index. It sets up the self-attention mechanism, multi-layer perceptron, input layer normalization, and other components required for the decoder layer.

  • config: The Phi3Config object that contains the configuration parameters for the decoder layer. This includes parameters such as hidden size, dropout rate, and RMS normalization epsilon.

  • layer_idx: An integer representing the index of the decoder layer. This index is used to identify the layer and is required for initializing the self-attention mechanism.

The method does not return any value.

Source code in mindnlp/transformers/models/phi3/modeling_phi3.py 
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
def __init__(self, config: Phi3Config, layer_idx: int):
    """
    Initializes a new instance of the Phi3DecoderLayer class.

    Args:
        self (Phi3DecoderLayer): The current instance of the Phi3DecoderLayer class.
        config (Phi3Config): The configuration object containing parameters for the decoder layer.
        layer_idx (int): The index of the decoder layer.

    Returns:
        None.

    Raises:
        None.

    Description:
        This method initializes the Phi3DecoderLayer object with the provided configuration and layer index.
        It sets up the self-attention mechanism, multi-layer perceptron, input layer normalization, and other
        components required for the decoder layer.

        - config: The Phi3Config object that contains the configuration parameters for the decoder layer.
        This includes parameters such as hidden size, dropout rate, and RMS normalization epsilon.

        - layer_idx: An integer representing the index of the decoder layer.
        This index is used to identify the layer and is required for initializing the self-attention mechanism.

        The method does not return any value.
    """
    super().__init__()

    self.config = config
    self.self_attn = PHI3_ATTENTION_CLASSES['eager'](config, layer_idx=layer_idx)

    self.mlp = Phi3MLP(config)
    self.input_layernorm = Phi3RMSNorm(config.hidden_size, eps=config.rms_norm_eps)

    self.resid_attn_dropout = nn.Dropout(config.resid_pdrop)
    self.resid_mlp_dropout = nn.Dropout(config.resid_pdrop)
    self.post_attention_layernorm = Phi3RMSNorm(config.hidden_size, eps=config.rms_norm_eps)

mindnlp.transformers.models.phi3.modeling_phi3.Phi3DecoderLayer.forward(hidden_states, attention_mask=None, position_ids=None, past_key_value=None, output_attentions=False, use_cache=False, **kwargs)

Constructs a Phi3DecoderLayer object.

PARAMETER DESCRIPTION
self

The object itself.
hidden_states

Input to the layer of shape (batch, seq_len, embed_dim).

TYPE: Tensor

attention_mask

Attention mask of size (batch, 1, tgt_len, src_len), where padding elements are indicated by very large negative values.

TYPE: Tensor DEFAULT: None

position_ids

Indices of positions of each input sequence tokens in the position embeddings. Selected in the range [0, config.n_positions - 1]. (default: None)

TYPE: Tensor DEFAULT: None

past_key_value

Cached past key and value projection states. (default: None)

TYPE: Tuple[Tensor] DEFAULT: None

output_attentions

Whether or not to return the attentions tensors of all attention layers. See attentions under returned tensors for more detail. (default: False)

TYPE: bool DEFAULT: False

use_cache

If set to True, past_key_values key value states are returned and can be used to speed up decoding (see past_key_values). (default: False)

TYPE: bool DEFAULT: False

RETURNS DESCRIPTION
Tuple[Tensor, Optional[Tuple[Tensor, Tensor]]]

Tuple[mindspore.Tensor, Optional[Tuple[mindspore.Tensor, mindspore.Tensor]]]: A tuple containing the hidden states of shape (batch, seq_len, embed_dim). Optionally, the tuple may also contain the attentions tensors of all attention layers and the cached past key and value projection states.
Source code in mindnlp/transformers/models/phi3/modeling_phi3.py 
878
879
880
881
882
883
884
885
886
887
888
889
890
891
892
893
894
895
896
897
898
899
900
901
902
903
904
905
906
907
908
909
910
911
912
913
914
915
916
917
918
919
920
921
922
923
924
925
926
927
928
929
930
931
932
933
934
935
936
937
938
939
940
941
942
943
944
945
946
947
948
949
950
951
952
953
954
955
956
957
958
959
960
961
962
963
964
def forward(
    self,
    hidden_states: mindspore.Tensor,
    attention_mask: Optional[mindspore.Tensor] = None,
    position_ids: Optional[mindspore.Tensor] = None,
    past_key_value: Optional[Tuple[mindspore.Tensor]] = None,
    output_attentions: Optional[bool] = False,
    use_cache: Optional[bool] = False,
    **kwargs,
) -> Tuple[mindspore.Tensor, Optional[Tuple[mindspore.Tensor, mindspore.Tensor]]]:
    '''
    Constructs a Phi3DecoderLayer object.

    Args:
        self: The object itself.
        hidden_states (mindspore.Tensor): Input to the layer of shape `(batch, seq_len, embed_dim)`.
        attention_mask (mindspore.Tensor, optional): Attention mask of size `(batch, 1, tgt_len, src_len)`,
            where padding elements are indicated by very large negative values.
        position_ids (mindspore.Tensor, optional): Indices of positions of each input sequence tokens in the
            position embeddings. Selected in the range `[0, config.n_positions - 1]`. (default: None)
        past_key_value (Tuple[mindspore.Tensor], optional): Cached past key and value projection states.
            (default: None)
        output_attentions (bool, optional): Whether or not to return the attentions tensors of all attention layers.
            See `attentions` under returned tensors for more detail. (default: False)
        use_cache (bool, optional): If set to True, `past_key_values` key value states are returned and can be used
            to speed up decoding (see `past_key_values`). (default: False)

    Returns:
        Tuple[mindspore.Tensor, Optional[Tuple[mindspore.Tensor, mindspore.Tensor]]]: A tuple containing the
            hidden states of shape `(batch, seq_len, embed_dim)`. Optionally, the tuple may also contain the
            attentions tensors of all attention layers and the cached past key and value projection states.

    Raises:
        None.
    '''
    if "padding_mask" in kwargs:
        warnings.warn(
            "Passing `padding_mask` is deprecated and will be removed in v4.37. Please make sure use `attention_mask` instead.`"
        )
    """
    Args:
        hidden_states (`mindspore.Tensor`):
            input to the layer of shape `(batch, seq_len, embed_dim)`
        attention_mask (`mindspore.Tensor`, *optional*): attention mask of size
            `(batch, 1, tgt_len, src_len)` where padding elements are indicated by very large negative values.
        position_ids (`mindspore.Tensor` of shape `({0})`, *optional*):
            Indices of positions of each input sequence tokens in the position embeddings. Selected in the range
            `[0, config.n_positions - 1]`. [What are position IDs?](../glossary#position-ids)
        output_attentions (`bool`, *optional*):
            Whether or not to return the attentions tensors of all attention layers. See `attentions` under
            returned tensors for more detail.
        use_cache (`bool`, *optional*):
            If set to `True`, `past_key_values` key value states are returned and can be used to speed up decoding
            (see `past_key_values`).
        past_key_value (`Tuple(mindspore.Tensor)`, *optional*): cached past key and value projection states
    """

    residual = hidden_states

    hidden_states = self.input_layernorm(hidden_states)

    # Self Attention
    attn_outputs, self_attn_weights, present_key_value = self.self_attn(
        hidden_states=hidden_states,
        attention_mask=attention_mask,
        position_ids=position_ids,
        past_key_value=past_key_value,
        output_attentions=output_attentions,
        use_cache=use_cache,
    )

    hidden_states = residual + self.resid_attn_dropout(attn_outputs)

    residual = hidden_states
    hidden_states = self.post_attention_layernorm(hidden_states)
    hidden_states = self.mlp(hidden_states)
    hidden_states = residual + self.resid_mlp_dropout(hidden_states)

    outputs = (hidden_states,)

    if output_attentions:
        outputs += (self_attn_weights,)

    if use_cache:
        outputs += (present_key_value,)

    return outputs

mindnlp.transformers.models.phi3.modeling_phi3.Phi3ForCausalLM

Bases: Phi3PreTrainedModel

A class representing the Phi3 model for causal language modeling.

This class extends the Phi3PreTrainedModel class and provides methods for initializing the model, setting and getting input and output embeddings, setting and getting the decoder, forwarding the model, and preparing inputs for generation.

ATTRIBUTE DESCRIPTION
model

The Phi3 model.

TYPE: Phi3Model

vocab_size

The size of the vocabulary.

TYPE: int

lm_head

The language model head.

TYPE: Linear

METHOD DESCRIPTION
__init__

Initializes the Phi3ForCausalLM instance.
get_input_embeddings

Returns the input embeddings.
set_input_embeddings

Sets the input embeddings.
get_output_embeddings

Returns the output embeddings.
set_output_embeddings

Sets the output embeddings.
set_decoder

Sets the decoder.
get_decoder

Returns the decoder.
forward

Constructs the model and returns the output.
prepare_inputs_for_generation

Prepares inputs for generation.
_reorder_cache

Reorders the cache based on the beam index.
Example 
>>> from transformers import AutoTokenizer, Phi3ForCausalLM
...
>>> model = Phi3ForCausalLM.from_pretrained("microsoft/phi-3-mini-4k-instruct")
>>> tokenizer = AutoTokenizer.from_pretrained("microsoft/phi-3-mini-4k-instruct")
...
>>> prompt = "This is an example script ."
>>> inputs = tokenizer(prompt, return_tensors="pt")
...
>>> # Generate
>>> generate_ids = model.generate(inputs.input_ids, max_length=30)
>>> tokenizer.batch_decode(generate_ids, skip_special_tokens=True, clean_up_tokenization_spaces=False)[0]
'This is an example script .\n Certainly! Below is a sample script that demonstrates a simple task, such as calculating the sum'
Source code in mindnlp/transformers/models/phi3/modeling_phi3.py 
1307
1308
1309
1310
1311
1312
1313
1314
1315
1316
1317
1318
1319
1320
1321
1322
1323
1324
1325
1326
1327
1328
1329
1330
1331
1332
1333
1334
1335
1336
1337
1338
1339
1340
1341
1342
1343
1344
1345
1346
1347
1348
1349
1350
1351
1352
1353
1354
1355
1356
1357
1358
1359
1360
1361
1362
1363
1364
1365
1366
1367
1368
1369
1370
1371
1372
1373
1374
1375
1376
1377
1378
1379
1380
1381
1382
1383
1384
1385
1386
1387
1388
1389
1390
1391
1392
1393
1394
1395
1396
1397
1398
1399
1400
1401
1402
1403
1404
1405
1406
1407
1408
1409
1410
1411
1412
1413
1414
1415
1416
1417
1418
1419
1420
1421
1422
1423
1424
1425
1426
1427
1428
1429
1430
1431
1432
1433
1434
1435
1436
1437
1438
1439
1440
1441
1442
1443
1444
1445
1446
1447
1448
1449
1450
1451
1452
1453
1454
1455
1456
1457
1458
1459
1460
1461
1462
1463
1464
1465
1466
1467
1468
1469
1470
1471
1472
1473
1474
1475
1476
1477
1478
1479
1480
1481
1482
1483
1484
1485
1486
1487
1488
1489
1490
1491
1492
1493
1494
1495
1496
1497
1498
1499
1500
1501
1502
1503
1504
1505
1506
1507
1508
1509
1510
1511
1512
1513
1514
1515
1516
1517
1518
1519
1520
1521
1522
1523
1524
1525
1526
1527
1528
1529
1530
1531
1532
1533
1534
1535
1536
1537
1538
1539
1540
1541
1542
1543
1544
1545
1546
1547
1548
1549
1550
1551
1552
1553
1554
1555
1556
1557
1558
1559
1560
1561
1562
1563
1564
1565
1566
1567
1568
1569
1570
1571
1572
1573
1574
1575
1576
1577
1578
1579
1580
1581
1582
1583
1584
1585
1586
1587
1588
1589
1590
1591
1592
1593
1594
1595
1596
1597
1598
1599
1600
1601
1602
1603
1604
1605
1606
1607
1608
1609
1610
1611
1612
1613
1614
1615
1616
1617
1618
1619
1620
1621
1622
1623
1624
1625
1626
1627
1628
1629
1630
1631
1632
1633
1634
1635
1636
1637
1638
1639
1640
1641
1642
1643
1644
1645
1646
1647
1648
1649
1650
1651
1652
1653
1654
1655
1656
1657
1658
1659
1660
1661
1662
1663
1664
1665
1666
1667
1668
1669
1670
1671
1672
1673
1674
1675
1676
class Phi3ForCausalLM(Phi3PreTrainedModel):
    r"""
    A class representing the Phi3 model for causal language modeling.

    This class extends the Phi3PreTrainedModel class and provides methods for initializing the model, setting and
    getting input and output embeddings, setting and getting the decoder, forwarding the model, and preparing inputs
    for generation.

    Attributes:
        model (Phi3Model): The Phi3 model.
        vocab_size (int): The size of the vocabulary.
        lm_head (nn.Linear): The language model head.

    Methods:
        __init__: Initializes the Phi3ForCausalLM instance.
        get_input_embeddings: Returns the input embeddings.
        set_input_embeddings: Sets the input embeddings.
        get_output_embeddings: Returns the output embeddings.
        set_output_embeddings: Sets the output embeddings.
        set_decoder: Sets the decoder.
        get_decoder: Returns the decoder.
        forward: Constructs the model and returns the output.
        prepare_inputs_for_generation: Prepares inputs for generation.
        _reorder_cache: Reorders the cache based on the beam index.

    Example:
        ```python
        >>> from transformers import AutoTokenizer, Phi3ForCausalLM
        ...
        >>> model = Phi3ForCausalLM.from_pretrained("microsoft/phi-3-mini-4k-instruct")
        >>> tokenizer = AutoTokenizer.from_pretrained("microsoft/phi-3-mini-4k-instruct")
        ...
        >>> prompt = "This is an example script ."
        >>> inputs = tokenizer(prompt, return_tensors="pt")
        ...
        >>> # Generate
        >>> generate_ids = model.generate(inputs.input_ids, max_length=30)
        >>> tokenizer.batch_decode(generate_ids, skip_special_tokens=True, clean_up_tokenization_spaces=False)[0]
        'This is an example script .\n Certainly! Below is a sample script that demonstrates a simple task, such as calculating the sum'
        ```
    """
    _tied_weights_keys = ["lm_head.weight"]

    # Copied from transformers.models.llama.modeling_llama.LlamaForCausalLM.__init__ with Llama->Phi3
    def __init__(self, config):
        """
        Initializes a new instance of the Phi3ForCausalLM class.

        Args:
            self: The object itself.
            config:
                A configuration object of type Config, containing the necessary parameters for model initialization.

                - Type: Config
                - Purpose: To provide the required parameters for model initialization.
                - Restrictions: None

        Returns:
            None

        Raises:
            None
        """
        super().__init__(config)
        self.model = Phi3Model(config)
        self.vocab_size = config.vocab_size
        self.lm_head = nn.Linear(config.hidden_size, config.vocab_size, bias=False)

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

    # Copied from transformers.models.llama.modeling_llama.LlamaForCausalLM.get_input_embeddings
    def get_input_embeddings(self):
        """
        Method to retrieve the input embeddings from the Phi3ForCausalLM model.

        Args:
            self: An instance of the Phi3ForCausalLM class.
                This parameter represents the current instance of the Phi3ForCausalLM class.
                It is used to access the model's embed_tokens attribute.

        Returns:
            None:
                This method returns None as it directly returns the embed_tokens attribute of the model.

        Raises:
            None.
        """
        return self.model.embed_tokens

    # Copied from transformers.models.llama.modeling_llama.LlamaForCausalLM.set_input_embeddings
    def set_input_embeddings(self, value):
        """
        Method to set input embeddings for the Phi3ForCausalLM model.

        Args:
            self (Phi3ForCausalLM): The instance of the Phi3ForCausalLM class.
            value (Any): The input embeddings to be set for the model.
                Should be compatible with the model's embed_tokens attribute.

        Returns:
            None.

        Raises:
            None.
        """
        self.model.embed_tokens = value

    # Copied from transformers.models.llama.modeling_llama.LlamaForCausalLM.get_output_embeddings
    def get_output_embeddings(self):
        """
        Returns the output embeddings of the Phi3ForCausalLM model.

        This method takes no additional parameters.

        Returns:
            None.

        Raises:
            None.
        """
        return self.lm_head

    # Copied from transformers.models.llama.modeling_llama.LlamaForCausalLM.set_output_embeddings
    def set_output_embeddings(self, new_embeddings):
        """
        Sets the output embeddings of the Phi3ForCausalLM model.

        Args:
            self (Phi3ForCausalLM): The instance of the Phi3ForCausalLM class.
            new_embeddings: The new embeddings to be set for the model's output. It can be of any type.

        Returns:
            None.

        Raises:
            None.
        """
        self.lm_head = new_embeddings

    # Copied from transformers.models.llama.modeling_llama.LlamaForCausalLM.set_decoder
    def set_decoder(self, decoder):
        """
        Sets the decoder for the Phi3ForCausalLM class.

        Args:
            self (Phi3ForCausalLM): The instance of Phi3ForCausalLM class.
            decoder: The decoder object to be set for the model.

        Returns:
            None.

        Raises:
            None.
        """
        self.model = decoder

    # Copied from transformers.models.llama.modeling_llama.LlamaForCausalLM.get_decoder
    def get_decoder(self):
        """
        This method returns the decoder model used in the Phi3ForCausalLM class.

        Args:
            self: The instance of the Phi3ForCausalLM class.

        Returns:
            model: The decoder model associated with the Phi3ForCausalLM instance.

        Raises:
            None.
        """
        return self.model

    def forward(
        self,
        input_ids: mindspore.Tensor = None,
        attention_mask: Optional[mindspore.Tensor] = None,
        position_ids: Optional[mindspore.Tensor] = None,
        past_key_values: Optional[List[mindspore.Tensor]] = None,
        inputs_embeds: Optional[mindspore.Tensor] = None,
        labels: Optional[mindspore.Tensor] = None,
        use_cache: Optional[bool] = None,
        output_attentions: Optional[bool] = None,
        output_hidden_states: Optional[bool] = None,
        return_dict: Optional[bool] = None,
    ) -> Union[Tuple, CausalLMOutputWithPast]:
        r"""
        Args:
            labels (`mindspore.Tensor` of shape `(batch_size, sequence_length)`, *optional*):
                Labels for computing the masked language modeling loss. Indices should either be in `[0, ...,
                config.vocab_size]` or -100 (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]`.

        Returns:
            Union[Tuple, CausalLMOutputWithPast]

        Example:
            ```python
            >>> from transformers import AutoTokenizer, Phi3ForCausalLM
            ...
            >>> model = Phi3ForCausalLM.from_pretrained("microsoft/phi-3-mini-4k-instruct")
            >>> tokenizer = AutoTokenizer.from_pretrained("microsoft/phi-3-mini-4k-instruct")
            ...
            >>> prompt = "This is an example script ."
            >>> inputs = tokenizer(prompt, return_tensors="pt")
            ...
            >>> # Generate
            >>> generate_ids = model.generate(inputs.input_ids, max_length=30)
            >>> tokenizer.batch_decode(generate_ids, skip_special_tokens=True, clean_up_tokenization_spaces=False)[0]
            'This is an example script .\n Certainly! Below is a sample script that demonstrates a simple task, such as calculating the sum'
            ```
        """
        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

        # decoder outputs consists of (dec_features, layer_state, dec_hidden, dec_attn)
        outputs = self.model(
            input_ids=input_ids,
            attention_mask=attention_mask,
            position_ids=position_ids,
            past_key_values=past_key_values,
            inputs_embeds=inputs_embeds,
            use_cache=use_cache,
            output_attentions=output_attentions,
            output_hidden_states=output_hidden_states,
            return_dict=return_dict,
        )

        hidden_states = outputs[0]
        logits = self.lm_head(hidden_states)
        logits = logits.float()

        loss = None
        if labels is not None:
            # Shift so that tokens < n predict n
            shift_logits = logits[..., :-1, :]
            shift_labels = labels[..., 1:]
            # Flatten the tokens
            shift_logits = shift_logits.view(-1, self.config.vocab_size)
            shift_labels = shift_labels.view(-1)
            # Enable model parallelism
            loss = ops.cross_entropy(shift_logits, shift_labels)

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

        return CausalLMOutputWithPast(
            loss=loss,
            logits=logits,
            past_key_values=outputs.past_key_values,
            hidden_states=outputs.hidden_states,
            attentions=outputs.attentions,
        )

    # Copied from transformers.models.persimmon.modeling_persimmon.PersimmonForCausalLM.prepare_inputs_for_generation
    def prepare_inputs_for_generation(
        self, input_ids, past_key_values=None, attention_mask=None, inputs_embeds=None, **kwargs
    ):
        """
        This method prepares inputs for generation in the Phi3ForCausalLM class.

        Args:
            self (object): The instance of Phi3ForCausalLM.
            input_ids (torch.Tensor): The input tensor containing token indices for the input sequence.
            past_key_values (Union[None, Cache, Tuple[Tensor, Tensor]]): A cache of past key values or the tuple
                of past key and value tensors. Defaults to None.
            attention_mask (Optional[torch.Tensor]): An optional tensor containing attention mask values
                for the input sequence.
            inputs_embeds (Optional[torch.Tensor]): An optional tensor containing the embedded inputs.

        Returns:
            model_inputs (Dict[str, Union[torch.Tensor, Cache]]):
                A dictionary containing the model inputs with the following keys:

                - 'inputs_embeds': The embedded inputs if 'inputs_embeds' is not None.
                - 'input_ids': The input tensor containing token indices if 'inputs_embeds' is None.
                - 'position_ids': The position indices tensor.
                - 'past_key_values': The cache of past key values.
                - 'use_cache': A boolean indicating whether to use cache.
                - 'attention_mask': The attention mask tensor.

        Raises:
            ValueError: If the dimensions of the input tensors are incompatible.
            TypeError: If the input types are invalid or incompatible.
        """
        if past_key_values is not None:
            if isinstance(past_key_values, Cache):
                cache_length = past_key_values.get_seq_length()
                past_length = past_key_values.seen_tokens
                max_cache_length = past_key_values.get_max_length()
            else:
                cache_length = past_length = past_key_values[0][0].shape[2]
                max_cache_length = None

            # Keep only the unprocessed tokens:
            # 1 - If the length of the attention_mask exceeds the length of input_ids, then we are in a setting where
            # some of the inputs are exclusively passed as part of the cache (e.g. when passing input_embeds as
            # input)
            if attention_mask is not None and attention_mask.shape[1] > input_ids.shape[1]:
                input_ids = input_ids[:, -(attention_mask.shape[1] - past_length) :]
            # 2 - If the past_length is smaller than input_ids', then input_ids holds all input tokens. We can discard
            # input_ids based on the past_length.
            elif past_length < input_ids.shape[1]:
                input_ids = input_ids[:, past_length:]
            # 3 - Otherwise (past_length >= input_ids.shape[1]), let's assume input_ids only has unprocessed tokens.

            # If we are about to go beyond the maximum cache length, we need to crop the input attention mask.
            if (
                max_cache_length is not None
                and attention_mask is not None
                and cache_length + input_ids.shape[1] > max_cache_length
            ):
                attention_mask = attention_mask[:, -max_cache_length:]

        position_ids = kwargs.get("position_ids", None)
        if attention_mask is not None and position_ids is None:
            # create position_ids on the fly for batch generation
            position_ids = attention_mask.long().cumsum(-1) - 1
            position_ids = position_ids.masked_fill(attention_mask == 0, 1)
            if past_key_values:
                position_ids = position_ids[:, -input_ids.shape[1] :]

        # if `inputs_embeds` are passed, we only want to use them in the 1st generation step
        if inputs_embeds is not None and past_key_values is None:
            model_inputs = {"inputs_embeds": inputs_embeds}
        else:
            model_inputs = {"input_ids": input_ids}

        model_inputs.update(
            {
                "position_ids": position_ids,
                "past_key_values": past_key_values,
                "use_cache": kwargs.get("use_cache"),
                "attention_mask": attention_mask,
            }
        )
        return model_inputs

    @staticmethod
    # Copied from transformers.models.llama.modeling_llama.LlamaForCausalLM._reorder_cache
    def _reorder_cache(past_key_values, beam_idx):
        """
        Reorders the cache based on the beam index for the Phi3ForCausalLM class.

        Args:
            past_key_values (tuple): A tuple containing the past key-value states for each layer. Each layer's past
                state is represented by a tensor.
            beam_idx (tensor): A tensor containing the beam index.

        Returns:
            tuple: A tuple containing the reordered past key-value states for each layer. Each layer's reordered past
                state is represented by a tensor.

        Raises:
            None.

        This static method reorders the cache based on the provided beam index. It iterates through each layer's
        past key-value states and selects the corresponding states from the past based on the beam index.
        The reordered past key-value states are then returned as a tuple.
        """
        reordered_past = ()
        for layer_past in past_key_values:
            reordered_past += (
                tuple(past_state.index_select(0, beam_idx) for past_state in layer_past),
            )
        return reordered_past

mindnlp.transformers.models.phi3.modeling_phi3.Phi3ForCausalLM.__init__(config)

Initializes a new instance of the Phi3ForCausalLM class.

PARAMETER DESCRIPTION
self

The object itself.
config

A configuration object of type Config, containing the necessary parameters for model initialization.

  • Type: Config
  • Purpose: To provide the required parameters for model initialization.
  • Restrictions: None
RETURNS DESCRIPTION

None
Source code in mindnlp/transformers/models/phi3/modeling_phi3.py 
1351
1352
1353
1354
1355
1356
1357
1358
1359
1360
1361
1362
1363
1364
1365
1366
1367
1368
1369
1370
1371
1372
1373
1374
1375
1376
def __init__(self, config):
    """
    Initializes a new instance of the Phi3ForCausalLM class.

    Args:
        self: The object itself.
        config:
            A configuration object of type Config, containing the necessary parameters for model initialization.

            - Type: Config
            - Purpose: To provide the required parameters for model initialization.
            - Restrictions: None

    Returns:
        None

    Raises:
        None
    """
    super().__init__(config)
    self.model = Phi3Model(config)
    self.vocab_size = config.vocab_size
    self.lm_head = nn.Linear(config.hidden_size, config.vocab_size, bias=False)

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

mindnlp.transformers.models.phi3.modeling_phi3.Phi3ForCausalLM.forward(input_ids=None, attention_mask=None, position_ids=None, past_key_values=None, inputs_embeds=None, labels=None, use_cache=None, output_attentions=None, output_hidden_states=None, return_dict=None)

PARAMETER DESCRIPTION
labels

Labels for computing the masked language modeling loss. Indices should either be in [0, ..., config.vocab_size] or -100 (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].

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

RETURNS DESCRIPTION
Union[Tuple, CausalLMOutputWithPast]

Union[Tuple, CausalLMOutputWithPast]
Example 
>>> from transformers import AutoTokenizer, Phi3ForCausalLM
...
>>> model = Phi3ForCausalLM.from_pretrained("microsoft/phi-3-mini-4k-instruct")
>>> tokenizer = AutoTokenizer.from_pretrained("microsoft/phi-3-mini-4k-instruct")
...
>>> prompt = "This is an example script ."
>>> inputs = tokenizer(prompt, return_tensors="pt")
...
>>> # Generate
>>> generate_ids = model.generate(inputs.input_ids, max_length=30)
>>> tokenizer.batch_decode(generate_ids, skip_special_tokens=True, clean_up_tokenization_spaces=False)[0]
'This is an example script .\n Certainly! Below is a sample script that demonstrates a simple task, such as calculating the sum'
Source code in mindnlp/transformers/models/phi3/modeling_phi3.py 
1480
1481
1482
1483
1484
1485
1486
1487
1488
1489
1490
1491
1492
1493
1494
1495
1496
1497
1498
1499
1500
1501
1502
1503
1504
1505
1506
1507
1508
1509
1510
1511
1512
1513
1514
1515
1516
1517
1518
1519
1520
1521
1522
1523
1524
1525
1526
1527
1528
1529
1530
1531
1532
1533
1534
1535
1536
1537
1538
1539
1540
1541
1542
1543
1544
1545
1546
1547
1548
1549
1550
1551
1552
1553
1554
1555
1556
1557
1558
1559
1560
1561
1562
1563
def forward(
    self,
    input_ids: mindspore.Tensor = None,
    attention_mask: Optional[mindspore.Tensor] = None,
    position_ids: Optional[mindspore.Tensor] = None,
    past_key_values: Optional[List[mindspore.Tensor]] = None,
    inputs_embeds: Optional[mindspore.Tensor] = None,
    labels: Optional[mindspore.Tensor] = None,
    use_cache: Optional[bool] = None,
    output_attentions: Optional[bool] = None,
    output_hidden_states: Optional[bool] = None,
    return_dict: Optional[bool] = None,
) -> Union[Tuple, CausalLMOutputWithPast]:
    r"""
    Args:
        labels (`mindspore.Tensor` of shape `(batch_size, sequence_length)`, *optional*):
            Labels for computing the masked language modeling loss. Indices should either be in `[0, ...,
            config.vocab_size]` or -100 (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]`.

    Returns:
        Union[Tuple, CausalLMOutputWithPast]

    Example:
        ```python
        >>> from transformers import AutoTokenizer, Phi3ForCausalLM
        ...
        >>> model = Phi3ForCausalLM.from_pretrained("microsoft/phi-3-mini-4k-instruct")
        >>> tokenizer = AutoTokenizer.from_pretrained("microsoft/phi-3-mini-4k-instruct")
        ...
        >>> prompt = "This is an example script ."
        >>> inputs = tokenizer(prompt, return_tensors="pt")
        ...
        >>> # Generate
        >>> generate_ids = model.generate(inputs.input_ids, max_length=30)
        >>> tokenizer.batch_decode(generate_ids, skip_special_tokens=True, clean_up_tokenization_spaces=False)[0]
        'This is an example script .\n Certainly! Below is a sample script that demonstrates a simple task, such as calculating the sum'
        ```
    """
    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

    # decoder outputs consists of (dec_features, layer_state, dec_hidden, dec_attn)
    outputs = self.model(
        input_ids=input_ids,
        attention_mask=attention_mask,
        position_ids=position_ids,
        past_key_values=past_key_values,
        inputs_embeds=inputs_embeds,
        use_cache=use_cache,
        output_attentions=output_attentions,
        output_hidden_states=output_hidden_states,
        return_dict=return_dict,
    )

    hidden_states = outputs[0]
    logits = self.lm_head(hidden_states)
    logits = logits.float()

    loss = None
    if labels is not None:
        # Shift so that tokens < n predict n
        shift_logits = logits[..., :-1, :]
        shift_labels = labels[..., 1:]
        # Flatten the tokens
        shift_logits = shift_logits.view(-1, self.config.vocab_size)
        shift_labels = shift_labels.view(-1)
        # Enable model parallelism
        loss = ops.cross_entropy(shift_logits, shift_labels)

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

    return CausalLMOutputWithPast(
        loss=loss,
        logits=logits,
        past_key_values=outputs.past_key_values,
        hidden_states=outputs.hidden_states,
        attentions=outputs.attentions,
    )

mindnlp.transformers.models.phi3.modeling_phi3.Phi3ForCausalLM.get_decoder()

This method returns the decoder model used in the Phi3ForCausalLM class.

PARAMETER DESCRIPTION
self

The instance of the Phi3ForCausalLM class.
RETURNS DESCRIPTION
model

The decoder model associated with the Phi3ForCausalLM instance.
Source code in mindnlp/transformers/models/phi3/modeling_phi3.py 
1465
1466
1467
1468
1469
1470
1471
1472
1473
1474
1475
1476
1477
1478
def get_decoder(self):
    """
    This method returns the decoder model used in the Phi3ForCausalLM class.

    Args:
        self: The instance of the Phi3ForCausalLM class.

    Returns:
        model: The decoder model associated with the Phi3ForCausalLM instance.

    Raises:
        None.
    """
    return self.model

mindnlp.transformers.models.phi3.modeling_phi3.Phi3ForCausalLM.get_input_embeddings()

Method to retrieve the input embeddings from the Phi3ForCausalLM model.

PARAMETER DESCRIPTION
self

An instance of the Phi3ForCausalLM class. This parameter represents the current instance of the Phi3ForCausalLM class. It is used to access the model's embed_tokens attribute.
RETURNS DESCRIPTION
None

This method returns None as it directly returns the embed_tokens attribute of the model.
Source code in mindnlp/transformers/models/phi3/modeling_phi3.py 
1379
1380
1381
1382
1383
1384
1385
1386
1387
1388
1389
1390
1391
1392
1393
1394
1395
def get_input_embeddings(self):
    """
    Method to retrieve the input embeddings from the Phi3ForCausalLM model.

    Args:
        self: An instance of the Phi3ForCausalLM class.
            This parameter represents the current instance of the Phi3ForCausalLM class.
            It is used to access the model's embed_tokens attribute.

    Returns:
        None:
            This method returns None as it directly returns the embed_tokens attribute of the model.

    Raises:
        None.
    """
    return self.model.embed_tokens

mindnlp.transformers.models.phi3.modeling_phi3.Phi3ForCausalLM.get_output_embeddings()

Returns the output embeddings of the Phi3ForCausalLM model.

This method takes no additional parameters.

RETURNS DESCRIPTION

None.
Source code in mindnlp/transformers/models/phi3/modeling_phi3.py 
1416
1417
1418
1419
1420
1421
1422
1423
1424
1425
1426
1427
1428
def get_output_embeddings(self):
    """
    Returns the output embeddings of the Phi3ForCausalLM model.

    This method takes no additional parameters.

    Returns:
        None.

    Raises:
        None.
    """
    return self.lm_head

mindnlp.transformers.models.phi3.modeling_phi3.Phi3ForCausalLM.prepare_inputs_for_generation(input_ids, past_key_values=None, attention_mask=None, inputs_embeds=None, **kwargs)

This method prepares inputs for generation in the Phi3ForCausalLM class.

PARAMETER DESCRIPTION
self

The instance of Phi3ForCausalLM.

TYPE: object

input_ids

The input tensor containing token indices for the input sequence.

TYPE: Tensor

past_key_values

A cache of past key values or the tuple of past key and value tensors. Defaults to None.

TYPE: Union[None, Cache, Tuple[Tensor, Tensor]] DEFAULT: None

attention_mask

An optional tensor containing attention mask values for the input sequence.

TYPE: Optional[Tensor] DEFAULT: None

inputs_embeds

An optional tensor containing the embedded inputs.

TYPE: Optional[Tensor] DEFAULT: None

RETURNS DESCRIPTION
model_inputs

A dictionary containing the model inputs with the following keys:

  • 'inputs_embeds': The embedded inputs if 'inputs_embeds' is not None.
  • 'input_ids': The input tensor containing token indices if 'inputs_embeds' is None.
  • 'position_ids': The position indices tensor.
  • 'past_key_values': The cache of past key values.
  • 'use_cache': A boolean indicating whether to use cache.
  • 'attention_mask': The attention mask tensor.

TYPE: Dict[str, Union[Tensor, Cache]]

RAISES DESCRIPTION
ValueError

If the dimensions of the input tensors are incompatible.
TypeError

If the input types are invalid or incompatible.
Source code in mindnlp/transformers/models/phi3/modeling_phi3.py 
1566
1567
1568
1569
1570
1571
1572
1573
1574
1575
1576
1577
1578
1579
1580
1581
1582
1583
1584
1585
1586
1587
1588
1589
1590
1591
1592
1593
1594
1595
1596
1597
1598
1599
1600
1601
1602
1603
1604
1605
1606
1607
1608
1609
1610
1611
1612
1613
1614
1615
1616
1617
1618
1619
1620
1621
1622
1623
1624
1625
1626
1627
1628
1629
1630
1631
1632
1633
1634
1635
1636
1637
1638
1639
1640
1641
1642
1643
1644
1645
1646
1647
def prepare_inputs_for_generation(
    self, input_ids, past_key_values=None, attention_mask=None, inputs_embeds=None, **kwargs
):
    """
    This method prepares inputs for generation in the Phi3ForCausalLM class.

    Args:
        self (object): The instance of Phi3ForCausalLM.
        input_ids (torch.Tensor): The input tensor containing token indices for the input sequence.
        past_key_values (Union[None, Cache, Tuple[Tensor, Tensor]]): A cache of past key values or the tuple
            of past key and value tensors. Defaults to None.
        attention_mask (Optional[torch.Tensor]): An optional tensor containing attention mask values
            for the input sequence.
        inputs_embeds (Optional[torch.Tensor]): An optional tensor containing the embedded inputs.

    Returns:
        model_inputs (Dict[str, Union[torch.Tensor, Cache]]):
            A dictionary containing the model inputs with the following keys:

            - 'inputs_embeds': The embedded inputs if 'inputs_embeds' is not None.
            - 'input_ids': The input tensor containing token indices if 'inputs_embeds' is None.
            - 'position_ids': The position indices tensor.
            - 'past_key_values': The cache of past key values.
            - 'use_cache': A boolean indicating whether to use cache.
            - 'attention_mask': The attention mask tensor.

    Raises:
        ValueError: If the dimensions of the input tensors are incompatible.
        TypeError: If the input types are invalid or incompatible.
    """
    if past_key_values is not None:
        if isinstance(past_key_values, Cache):
            cache_length = past_key_values.get_seq_length()
            past_length = past_key_values.seen_tokens
            max_cache_length = past_key_values.get_max_length()
        else:
            cache_length = past_length = past_key_values[0][0].shape[2]
            max_cache_length = None

        # Keep only the unprocessed tokens:
        # 1 - If the length of the attention_mask exceeds the length of input_ids, then we are in a setting where
        # some of the inputs are exclusively passed as part of the cache (e.g. when passing input_embeds as
        # input)
        if attention_mask is not None and attention_mask.shape[1] > input_ids.shape[1]:
            input_ids = input_ids[:, -(attention_mask.shape[1] - past_length) :]
        # 2 - If the past_length is smaller than input_ids', then input_ids holds all input tokens. We can discard
        # input_ids based on the past_length.
        elif past_length < input_ids.shape[1]:
            input_ids = input_ids[:, past_length:]
        # 3 - Otherwise (past_length >= input_ids.shape[1]), let's assume input_ids only has unprocessed tokens.

        # If we are about to go beyond the maximum cache length, we need to crop the input attention mask.
        if (
            max_cache_length is not None
            and attention_mask is not None
            and cache_length + input_ids.shape[1] > max_cache_length
        ):
            attention_mask = attention_mask[:, -max_cache_length:]

    position_ids = kwargs.get("position_ids", None)
    if attention_mask is not None and position_ids is None:
        # create position_ids on the fly for batch generation
        position_ids = attention_mask.long().cumsum(-1) - 1
        position_ids = position_ids.masked_fill(attention_mask == 0, 1)
        if past_key_values:
            position_ids = position_ids[:, -input_ids.shape[1] :]

    # if `inputs_embeds` are passed, we only want to use them in the 1st generation step
    if inputs_embeds is not None and past_key_values is None:
        model_inputs = {"inputs_embeds": inputs_embeds}
    else:
        model_inputs = {"input_ids": input_ids}

    model_inputs.update(
        {
            "position_ids": position_ids,
            "past_key_values": past_key_values,
            "use_cache": kwargs.get("use_cache"),
            "attention_mask": attention_mask,
        }
    )
    return model_inputs

mindnlp.transformers.models.phi3.modeling_phi3.Phi3ForCausalLM.set_decoder(decoder)

Sets the decoder for the Phi3ForCausalLM class.

PARAMETER DESCRIPTION
self

The instance of Phi3ForCausalLM class.

TYPE: Phi3ForCausalLM

decoder

The decoder object to be set for the model.
RETURNS DESCRIPTION

None.
Source code in mindnlp/transformers/models/phi3/modeling_phi3.py 
1448
1449
1450
1451
1452
1453
1454
1455
1456
1457
1458
1459
1460
1461
1462
def set_decoder(self, decoder):
    """
    Sets the decoder for the Phi3ForCausalLM class.

    Args:
        self (Phi3ForCausalLM): The instance of Phi3ForCausalLM class.
        decoder: The decoder object to be set for the model.

    Returns:
        None.

    Raises:
        None.
    """
    self.model = decoder

mindnlp.transformers.models.phi3.modeling_phi3.Phi3ForCausalLM.set_input_embeddings(value)

Method to set input embeddings for the Phi3ForCausalLM model.

PARAMETER DESCRIPTION
self

The instance of the Phi3ForCausalLM class.

TYPE: Phi3ForCausalLM

value

The input embeddings to be set for the model. Should be compatible with the model's embed_tokens attribute.

TYPE: Any

RETURNS DESCRIPTION

None.
Source code in mindnlp/transformers/models/phi3/modeling_phi3.py 
1398
1399
1400
1401
1402
1403
1404
1405
1406
1407
1408
1409
1410
1411
1412
1413
def set_input_embeddings(self, value):
    """
    Method to set input embeddings for the Phi3ForCausalLM model.

    Args:
        self (Phi3ForCausalLM): The instance of the Phi3ForCausalLM class.
        value (Any): The input embeddings to be set for the model.
            Should be compatible with the model's embed_tokens attribute.

    Returns:
        None.

    Raises:
        None.
    """
    self.model.embed_tokens = value

mindnlp.transformers.models.phi3.modeling_phi3.Phi3ForCausalLM.set_output_embeddings(new_embeddings)

Sets the output embeddings of the Phi3ForCausalLM model.

PARAMETER DESCRIPTION
self

The instance of the Phi3ForCausalLM class.

TYPE: Phi3ForCausalLM

new_embeddings

The new embeddings to be set for the model's output. It can be of any type.
RETURNS DESCRIPTION

None.
Source code in mindnlp/transformers/models/phi3/modeling_phi3.py 
1431
1432
1433
1434
1435
1436
1437
1438
1439
1440
1441
1442
1443
1444
1445
def set_output_embeddings(self, new_embeddings):
    """
    Sets the output embeddings of the Phi3ForCausalLM model.

    Args:
        self (Phi3ForCausalLM): The instance of the Phi3ForCausalLM class.
        new_embeddings: The new embeddings to be set for the model's output. It can be of any type.

    Returns:
        None.

    Raises:
        None.
    """
    self.lm_head = new_embeddings

mindnlp.transformers.models.phi3.modeling_phi3.Phi3ForSequenceClassification

Bases: Phi3PreTrainedModel

This class represents a Phi3 model for sequence classification. It is a subclass of the Phi3PreTrainedModel class.

ATTRIBUTE DESCRIPTION
num_labels

The number of labels for sequence classification.

TYPE: int

model

The Phi3 model for sequence classification.

TYPE: Phi3Model

score

The dense layer for scoring the hidden states.

TYPE: Linear

METHOD DESCRIPTION
__init__

Initializes a new instance of the Phi3ForSequenceClassification class.
get_input_embeddings

Retrieves the input embeddings from the Phi3 model.
set_input_embeddings

Sets the input embeddings for the Phi3 model.
forward

Constructs the Phi3 model for sequence classification.
Source code in mindnlp/transformers/models/phi3/modeling_phi3.py 
1680
1681
1682
1683
1684
1685
1686
1687
1688
1689
1690
1691
1692
1693
1694
1695
1696
1697
1698
1699
1700
1701
1702
1703
1704
1705
1706
1707
1708
1709
1710
1711
1712
1713
1714
1715
1716
1717
1718
1719
1720
1721
1722
1723
1724
1725
1726
1727
1728
1729
1730
1731
1732
1733
1734
1735
1736
1737
1738
1739
1740
1741
1742
1743
1744
1745
1746
1747
1748
1749
1750
1751
1752
1753
1754
1755
1756
1757
1758
1759
1760
1761
1762
1763
1764
1765
1766
1767
1768
1769
1770
1771
1772
1773
1774
1775
1776
1777
1778
1779
1780
1781
1782
1783
1784
1785
1786
1787
1788
1789
1790
1791
1792
1793
1794
1795
1796
1797
1798
1799
1800
1801
1802
1803
1804
1805
1806
1807
1808
1809
1810
1811
1812
1813
1814
1815
1816
1817
1818
1819
1820
1821
1822
1823
1824
1825
1826
1827
1828
1829
1830
1831
1832
1833
1834
1835
1836
1837
1838
1839
1840
1841
1842
1843
1844
1845
1846
1847
1848
1849
1850
1851
1852
class Phi3ForSequenceClassification(Phi3PreTrainedModel):

    """
    This class represents a Phi3 model for sequence classification. It is a subclass of the Phi3PreTrainedModel class.

    Attributes:
        num_labels (int): The number of labels for sequence classification.
        model (Phi3Model): The Phi3 model for sequence classification.
        score (nn.Linear): The dense layer for scoring the hidden states.

    Methods:
        __init__: Initializes a new instance of the Phi3ForSequenceClassification class.
        get_input_embeddings: Retrieves the input embeddings from the Phi3 model.
        set_input_embeddings: Sets the input embeddings for the Phi3 model.
        forward: Constructs the Phi3 model for sequence classification.

    """
    def __init__(self, config):
        """
        Initializes a new instance of the Phi3ForSequenceClassification class.

        Args:
            self (Phi3ForSequenceClassification): The current instance of the Phi3ForSequenceClassification class.
            config (object): An object containing configuration settings for the model.
                It should have the following attributes:

                - num_labels (int): The number of labels/classes for classification.

        Returns:
            None.

        Raises:
            None.
        """
        super().__init__(config)
        self.num_labels = config.num_labels
        self.model = Phi3Model(config)
        self.score = nn.Linear(config.hidden_size, self.num_labels, bias=False)

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

    def get_input_embeddings(self):
        """
        Retrieves the input embeddings from the Phi3ForSequenceClassification model.

        Args:
            self: An instance of the Phi3ForSequenceClassification class.

        Returns:
            None.

        Raises:
            None.

        Description:
            This method is used to extract the input embeddings from the Phi3ForSequenceClassification model.
            The input embeddings represent the learned representations of the input tokens in the model.

            This method takes one parameter 'self', which refers to the current instance of the
            Phi3ForSequenceClassification class. It is required to access the model and its embedded tokens.

        Example:
            ```python
            >>> model = Phi3ForSequenceClassification()
            >>> input_embeddings = model.get_input_embeddings()
            ```

        """
        return self.model.embed_tokens

    def set_input_embeddings(self, value):
        """
        Set the input embeddings for the Phi3ForSequenceClassification model.

        Args:
            self (Phi3ForSequenceClassification): The instance of Phi3ForSequenceClassification.
            value (Tensor): The input embeddings to be set for the model. Should be a tensor of shape
                (vocab_size, embedding_dim).

        Returns:
            None.

        Raises:
            None.
        """
        self.model.embed_tokens = value

    def forward(
        self,
        input_ids: mindspore.Tensor = None,
        attention_mask: Optional[mindspore.Tensor] = None,
        position_ids: Optional[mindspore.Tensor] = None,
        past_key_values: Optional[List[mindspore.Tensor]] = None,
        inputs_embeds: Optional[mindspore.Tensor] = None,
        labels: Optional[mindspore.Tensor] = None,
        use_cache: Optional[bool] = None,
        output_attentions: Optional[bool] = None,
        output_hidden_states: Optional[bool] = None,
        return_dict: Optional[bool] = None,
    ) -> Union[Tuple, SequenceClassifierOutputWithPast]:
        r"""
        Args:
            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

        model_outputs = self.model(
            input_ids,
            attention_mask=attention_mask,
            position_ids=position_ids,
            past_key_values=past_key_values,
            inputs_embeds=inputs_embeds,
            use_cache=use_cache,
            output_attentions=output_attentions,
            output_hidden_states=output_hidden_states,
            return_dict=return_dict,
        )
        hidden_states = model_outputs[0]
        logits = self.score(hidden_states)

        if input_ids is not None:
            batch_size = input_ids.shape[0]
        else:
            batch_size = inputs_embeds.shape[0]

        if self.config.pad_token_id is None and batch_size != 1:
            raise ValueError("Cannot handle batch sizes > 1 if no padding token is defined.")
        if self.config.pad_token_id is None:
            sequence_lengths = -1
        else:
            if input_ids is not None:
                # if no pad token found, use modulo instead of reverse indexing for ONNX compatibility
                sequence_lengths = ops.eq(input_ids, self.config.pad_token_id).int().argmax(-1) - 1
                sequence_lengths = sequence_lengths % input_ids.shape[-1]
            else:
                sequence_lengths = -1

        pooled_logits = logits[ops.arange(batch_size), sequence_lengths]

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

            if self.config.problem_type == "regression":
                if self.num_labels == 1:
                    loss = ops.mse_loss(pooled_logits.squeeze(), labels.squeeze())
                else:
                    loss = ops.mse_loss(pooled_logits, labels)
            elif self.config.problem_type == "single_label_classification":
                loss = ops.cross_entropy(pooled_logits.view(-1, self.num_labels), labels.view(-1))
            elif self.config.problem_type == "multi_label_classification":
                loss = ops.binary_cross_entropy_with_logits(pooled_logits, labels)
        if not return_dict:
            output = (pooled_logits,) + model_outputs[1:]
            return ((loss,) + output) if loss is not None else output

        return SequenceClassifierOutputWithPast(
            loss=loss,
            logits=pooled_logits,
            past_key_values=model_outputs.past_key_values,
            hidden_states=model_outputs.hidden_states,
            attentions=model_outputs.attentions,
        )

mindnlp.transformers.models.phi3.modeling_phi3.Phi3ForSequenceClassification.__init__(config)

Initializes a new instance of the Phi3ForSequenceClassification class.

PARAMETER DESCRIPTION
self

The current instance of the Phi3ForSequenceClassification class.

TYPE: Phi3ForSequenceClassification

config

An object containing configuration settings for the model. It should have the following attributes:

  • num_labels (int): The number of labels/classes for classification.

TYPE: object

RETURNS DESCRIPTION

None.
Source code in mindnlp/transformers/models/phi3/modeling_phi3.py 
1697
1698
1699
1700
1701
1702
1703
1704
1705
1706
1707
1708
1709
1710
1711
1712
1713
1714
1715
1716
1717
1718
1719
1720
def __init__(self, config):
    """
    Initializes a new instance of the Phi3ForSequenceClassification class.

    Args:
        self (Phi3ForSequenceClassification): The current instance of the Phi3ForSequenceClassification class.
        config (object): An object containing configuration settings for the model.
            It should have the following attributes:

            - num_labels (int): The number of labels/classes for classification.

    Returns:
        None.

    Raises:
        None.
    """
    super().__init__(config)
    self.num_labels = config.num_labels
    self.model = Phi3Model(config)
    self.score = nn.Linear(config.hidden_size, self.num_labels, bias=False)

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

mindnlp.transformers.models.phi3.modeling_phi3.Phi3ForSequenceClassification.forward(input_ids=None, attention_mask=None, position_ids=None, past_key_values=None, inputs_embeds=None, labels=None, use_cache=None, output_attentions=None, output_hidden_states=None, return_dict=None)

PARAMETER DESCRIPTION
labels

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).

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

Source code in mindnlp/transformers/models/phi3/modeling_phi3.py 
1768
1769
1770
1771
1772
1773
1774
1775
1776
1777
1778
1779
1780
1781
1782
1783
1784
1785
1786
1787
1788
1789
1790
1791
1792
1793
1794
1795
1796
1797
1798
1799
1800
1801
1802
1803
1804
1805
1806
1807
1808
1809
1810
1811
1812
1813
1814
1815
1816
1817
1818
1819
1820
1821
1822
1823
1824
1825
1826
1827
1828
1829
1830
1831
1832
1833
1834
1835
1836
1837
1838
1839
1840
1841
1842
1843
1844
1845
1846
1847
1848
1849
1850
1851
1852
def forward(
    self,
    input_ids: mindspore.Tensor = None,
    attention_mask: Optional[mindspore.Tensor] = None,
    position_ids: Optional[mindspore.Tensor] = None,
    past_key_values: Optional[List[mindspore.Tensor]] = None,
    inputs_embeds: Optional[mindspore.Tensor] = None,
    labels: Optional[mindspore.Tensor] = None,
    use_cache: Optional[bool] = None,
    output_attentions: Optional[bool] = None,
    output_hidden_states: Optional[bool] = None,
    return_dict: Optional[bool] = None,
) -> Union[Tuple, SequenceClassifierOutputWithPast]:
    r"""
    Args:
        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

    model_outputs = self.model(
        input_ids,
        attention_mask=attention_mask,
        position_ids=position_ids,
        past_key_values=past_key_values,
        inputs_embeds=inputs_embeds,
        use_cache=use_cache,
        output_attentions=output_attentions,
        output_hidden_states=output_hidden_states,
        return_dict=return_dict,
    )
    hidden_states = model_outputs[0]
    logits = self.score(hidden_states)

    if input_ids is not None:
        batch_size = input_ids.shape[0]
    else:
        batch_size = inputs_embeds.shape[0]

    if self.config.pad_token_id is None and batch_size != 1:
        raise ValueError("Cannot handle batch sizes > 1 if no padding token is defined.")
    if self.config.pad_token_id is None:
        sequence_lengths = -1
    else:
        if input_ids is not None:
            # if no pad token found, use modulo instead of reverse indexing for ONNX compatibility
            sequence_lengths = ops.eq(input_ids, self.config.pad_token_id).int().argmax(-1) - 1
            sequence_lengths = sequence_lengths % input_ids.shape[-1]
        else:
            sequence_lengths = -1

    pooled_logits = logits[ops.arange(batch_size), sequence_lengths]

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

        if self.config.problem_type == "regression":
            if self.num_labels == 1:
                loss = ops.mse_loss(pooled_logits.squeeze(), labels.squeeze())
            else:
                loss = ops.mse_loss(pooled_logits, labels)
        elif self.config.problem_type == "single_label_classification":
            loss = ops.cross_entropy(pooled_logits.view(-1, self.num_labels), labels.view(-1))
        elif self.config.problem_type == "multi_label_classification":
            loss = ops.binary_cross_entropy_with_logits(pooled_logits, labels)
    if not return_dict:
        output = (pooled_logits,) + model_outputs[1:]
        return ((loss,) + output) if loss is not None else output

    return SequenceClassifierOutputWithPast(
        loss=loss,
        logits=pooled_logits,
        past_key_values=model_outputs.past_key_values,
        hidden_states=model_outputs.hidden_states,
        attentions=model_outputs.attentions,
    )

mindnlp.transformers.models.phi3.modeling_phi3.Phi3ForSequenceClassification.get_input_embeddings()

Retrieves the input embeddings from the Phi3ForSequenceClassification model.

PARAMETER DESCRIPTION
self

An instance of the Phi3ForSequenceClassification class.
RETURNS DESCRIPTION

None.
Description

This method is used to extract the input embeddings from the Phi3ForSequenceClassification model. The input embeddings represent the learned representations of the input tokens in the model.

This method takes one parameter 'self', which refers to the current instance of the Phi3ForSequenceClassification class. It is required to access the model and its embedded tokens.

Example 
>>> model = Phi3ForSequenceClassification()
>>> input_embeddings = model.get_input_embeddings()
Source code in mindnlp/transformers/models/phi3/modeling_phi3.py 
1722
1723
1724
1725
1726
1727
1728
1729
1730
1731
1732
1733
1734
1735
1736
1737
1738
1739
1740
1741
1742
1743
1744
1745
1746
1747
1748
1749
def get_input_embeddings(self):
    """
    Retrieves the input embeddings from the Phi3ForSequenceClassification model.

    Args:
        self: An instance of the Phi3ForSequenceClassification class.

    Returns:
        None.

    Raises:
        None.

    Description:
        This method is used to extract the input embeddings from the Phi3ForSequenceClassification model.
        The input embeddings represent the learned representations of the input tokens in the model.

        This method takes one parameter 'self', which refers to the current instance of the
        Phi3ForSequenceClassification class. It is required to access the model and its embedded tokens.

    Example:
        ```python
        >>> model = Phi3ForSequenceClassification()
        >>> input_embeddings = model.get_input_embeddings()
        ```

    """
    return self.model.embed_tokens

mindnlp.transformers.models.phi3.modeling_phi3.Phi3ForSequenceClassification.set_input_embeddings(value)

Set the input embeddings for the Phi3ForSequenceClassification model.

PARAMETER DESCRIPTION
self

The instance of Phi3ForSequenceClassification.

TYPE: Phi3ForSequenceClassification

value

The input embeddings to be set for the model. Should be a tensor of shape (vocab_size, embedding_dim).

TYPE: Tensor

RETURNS DESCRIPTION

None.
Source code in mindnlp/transformers/models/phi3/modeling_phi3.py 
1751
1752
1753
1754
1755
1756
1757
1758
1759
1760
1761
1762
1763
1764
1765
1766
def set_input_embeddings(self, value):
    """
    Set the input embeddings for the Phi3ForSequenceClassification model.

    Args:
        self (Phi3ForSequenceClassification): The instance of Phi3ForSequenceClassification.
        value (Tensor): The input embeddings to be set for the model. Should be a tensor of shape
            (vocab_size, embedding_dim).

    Returns:
        None.

    Raises:
        None.
    """
    self.model.embed_tokens = value

mindnlp.transformers.models.phi3.modeling_phi3.Phi3ForTokenClassification

Bases: Phi3PreTrainedModel

Phi3ForTokenClassification is a class that represents a token classification model for Phi3, inheriting from Phi3PreTrainedModel.

The class includes an init method for initializing the model with a Phi3Config object, setting up the necessary components such as the model architecture, dropout layers, and classifier for token classification.

It also contains a forward method for performing the token classification task, taking input tensors, past key values, attention masks, and other optional arguments. It computes the classification loss using cross-entropy and returns the loss along with logits and hidden states if specified in the return_dict.

ATTRIBUTE DESCRIPTION
num_labels

The number of labels for token classification.
model

The Phi3Model instance for processing inputs.
dropout

Dropout layer for regularization.
classifier

Dense layer for classification.
METHOD DESCRIPTION
__init__

Constructor method to initialize the Phi3ForTokenClassification instance.
forward

Method for performing token classification task using the model.
Note

Ensure to set the appropriate labels for computing the loss, and handle the return_dict parameter for controlling the output format.

Source code in mindnlp/transformers/models/phi3/modeling_phi3.py 
1856
1857
1858
1859
1860
1861
1862
1863
1864
1865
1866
1867
1868
1869
1870
1871
1872
1873
1874
1875
1876
1877
1878
1879
1880
1881
1882
1883
1884
1885
1886
1887
1888
1889
1890
1891
1892
1893
1894
1895
1896
1897
1898
1899
1900
1901
1902
1903
1904
1905
1906
1907
1908
1909
1910
1911
1912
1913
1914
1915
1916
1917
1918
1919
1920
1921
1922
1923
1924
1925
1926
1927
1928
1929
1930
1931
1932
1933
1934
1935
1936
1937
1938
1939
1940
1941
1942
1943
1944
1945
1946
1947
1948
1949
1950
1951
1952
1953
1954
1955
1956
1957
1958
1959
1960
1961
1962
1963
1964
1965
1966
1967
class Phi3ForTokenClassification(Phi3PreTrainedModel):

    """
    Phi3ForTokenClassification is a class that represents a token classification model for Phi3, inheriting from
    Phi3PreTrainedModel.

    The class includes an __init__ method for initializing the model with a Phi3Config object, setting up the
    necessary components such as the model architecture, dropout layers, and classifier for token classification.

    It also contains a forward method for performing the token classification task, taking input tensors,
    past key values, attention masks, and other optional arguments. It computes the classification loss using
    cross-entropy and returns the loss along with logits and hidden states if specified in the return_dict.

    Attributes:
        num_labels: The number of labels for token classification.
        model: The Phi3Model instance for processing inputs.
        dropout: Dropout layer for regularization.
        classifier: Dense layer for classification.

    Methods:
        __init__: Constructor method to initialize the Phi3ForTokenClassification instance.
        forward: Method for performing token classification task using the model.

    Note:
        Ensure to set the appropriate labels for computing the loss, and handle the return_dict parameter for
        controlling the output format.
    """
    def __init__(self, config: Phi3Config):
        """
        Initializes an instance of Phi3ForTokenClassification with the provided configuration.

        Args:
            self: The instance of the Phi3ForTokenClassification class.
            config (Phi3Config): An instance of Phi3Config containing the configuration parameters for the model.

        Returns:
            None.

        Raises:
            TypeError: If the config parameter is not of type Phi3Config.
            AttributeError: If the config object does not have the required attributes.
        """
        super().__init__(config)
        self.num_labels = config.num_labels

        self.model = Phi3Model(config)
        if hasattr(config, "classifier_dropout") and config.classifier_dropout is not None:
            classifier_dropout = config.classifier_dropout
        elif hasattr(config, "hidden_dropout") and config.hidden_dropout is not None:
            classifier_dropout = config.hidden_dropout
        else:
            classifier_dropout = 0.1
        self.dropout = nn.Dropout(classifier_dropout)
        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,
        past_key_values: Optional[Tuple[Tuple[mindspore.Tensor, mindspore.Tensor], ...]] = None,
        attention_mask: Optional[mindspore.Tensor] = None,
        inputs_embeds: Optional[mindspore.Tensor] = None,
        labels: Optional[mindspore.Tensor] = None,
        use_cache: Optional[bool] = None,
        output_attentions: Optional[bool] = None,
        output_hidden_states: Optional[bool] = None,
        return_dict: Optional[bool] = None,
        **deprecated_arguments,
    ) -> Union[Tuple[mindspore.Tensor], TokenClassifierOutput]:
        r"""
        Args:
            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

        model_outputs = self.model(
            input_ids,
            past_key_values=past_key_values,
            attention_mask=attention_mask,
            inputs_embeds=inputs_embeds,
            use_cache=use_cache,
            output_attentions=output_attentions,
            output_hidden_states=output_hidden_states,
            return_dict=return_dict,
        )

        hidden_states = model_outputs[0]
        hidden_states = self.dropout(hidden_states)
        logits = self.classifier(hidden_states)

        loss = None
        if labels is not None:
            batch_size, seq_length = labels.shape
            loss = ops.cross_entropy(
                logits.view(batch_size * seq_length, self.num_labels), labels.view(batch_size * seq_length)
            )

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

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

mindnlp.transformers.models.phi3.modeling_phi3.Phi3ForTokenClassification.__init__(config)

Initializes an instance of Phi3ForTokenClassification with the provided configuration.

PARAMETER DESCRIPTION
self

The instance of the Phi3ForTokenClassification class.
config

An instance of Phi3Config containing the configuration parameters for the model.

TYPE: Phi3Config

RETURNS DESCRIPTION

None.
RAISES DESCRIPTION
TypeError

If the config parameter is not of type Phi3Config.
AttributeError

If the config object does not have the required attributes.
Source code in mindnlp/transformers/models/phi3/modeling_phi3.py 
1883
1884
1885
1886
1887
1888
1889
1890
1891
1892
1893
1894
1895
1896
1897
1898
1899
1900
1901
1902
1903
1904
1905
1906
1907
1908
1909
1910
1911
1912
def __init__(self, config: Phi3Config):
    """
    Initializes an instance of Phi3ForTokenClassification with the provided configuration.

    Args:
        self: The instance of the Phi3ForTokenClassification class.
        config (Phi3Config): An instance of Phi3Config containing the configuration parameters for the model.

    Returns:
        None.

    Raises:
        TypeError: If the config parameter is not of type Phi3Config.
        AttributeError: If the config object does not have the required attributes.
    """
    super().__init__(config)
    self.num_labels = config.num_labels

    self.model = Phi3Model(config)
    if hasattr(config, "classifier_dropout") and config.classifier_dropout is not None:
        classifier_dropout = config.classifier_dropout
    elif hasattr(config, "hidden_dropout") and config.hidden_dropout is not None:
        classifier_dropout = config.hidden_dropout
    else:
        classifier_dropout = 0.1
    self.dropout = nn.Dropout(classifier_dropout)
    self.classifier = nn.Linear(config.hidden_size, config.num_labels)

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

mindnlp.transformers.models.phi3.modeling_phi3.Phi3ForTokenClassification.forward(input_ids=None, past_key_values=None, attention_mask=None, inputs_embeds=None, labels=None, use_cache=None, output_attentions=None, output_hidden_states=None, return_dict=None, **deprecated_arguments)

PARAMETER DESCRIPTION
labels

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).

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

Source code in mindnlp/transformers/models/phi3/modeling_phi3.py 
1914
1915
1916
1917
1918
1919
1920
1921
1922
1923
1924
1925
1926
1927
1928
1929
1930
1931
1932
1933
1934
1935
1936
1937
1938
1939
1940
1941
1942
1943
1944
1945
1946
1947
1948
1949
1950
1951
1952
1953
1954
1955
1956
1957
1958
1959
1960
1961
1962
1963
1964
1965
1966
1967
def forward(
    self,
    input_ids: Optional[mindspore.Tensor] = None,
    past_key_values: Optional[Tuple[Tuple[mindspore.Tensor, mindspore.Tensor], ...]] = None,
    attention_mask: Optional[mindspore.Tensor] = None,
    inputs_embeds: Optional[mindspore.Tensor] = None,
    labels: Optional[mindspore.Tensor] = None,
    use_cache: Optional[bool] = None,
    output_attentions: Optional[bool] = None,
    output_hidden_states: Optional[bool] = None,
    return_dict: Optional[bool] = None,
    **deprecated_arguments,
) -> Union[Tuple[mindspore.Tensor], TokenClassifierOutput]:
    r"""
    Args:
        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

    model_outputs = self.model(
        input_ids,
        past_key_values=past_key_values,
        attention_mask=attention_mask,
        inputs_embeds=inputs_embeds,
        use_cache=use_cache,
        output_attentions=output_attentions,
        output_hidden_states=output_hidden_states,
        return_dict=return_dict,
    )

    hidden_states = model_outputs[0]
    hidden_states = self.dropout(hidden_states)
    logits = self.classifier(hidden_states)

    loss = None
    if labels is not None:
        batch_size, seq_length = labels.shape
        loss = ops.cross_entropy(
            logits.view(batch_size * seq_length, self.num_labels), labels.view(batch_size * seq_length)
        )

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

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

mindnlp.transformers.models.phi3.modeling_phi3.Phi3MLP

Bases: Module

This class represents a multi-layer perceptron (MLP) module with a Phi3 activation function. It inherits from the nn.Module class.

The Phi3MLP module is used for processing hidden states in a neural network. It consists of an up projection layer, a gate activation function, and a down projection layer.

ATTRIBUTE DESCRIPTION
config

An object containing configuration settings for the module.

TYPE: object

METHOD DESCRIPTION
__init__

Initializes a Phi3MLP instance.

Args:

  • config (object): An object containing configuration settings for the module.
forward

Constructs the Phi3MLP module.

Args:

  • hidden_states (mindspore.Tensor): The input hidden states tensor.

Returns:

  • mindspore.Tensor: The output tensor after applying the Phi3MLP module.
Source code in mindnlp/transformers/models/phi3/modeling_phi3.py 
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
class Phi3MLP(nn.Module):

    """
    This class represents a multi-layer perceptron (MLP) module with a Phi3 activation function.
    It inherits from the nn.Module class.

    The Phi3MLP module is used for processing hidden states in a neural network. It consists of an up projection layer,
    a gate activation function, and a down projection layer.

    Attributes:
        config (object): An object containing configuration settings for the module.

    Methods:
        __init__:
            Initializes a Phi3MLP instance.

            Args:

            - config (object): An object containing configuration settings for the module.

        forward:
            Constructs the Phi3MLP module.

            Args:

            - hidden_states (mindspore.Tensor): The input hidden states tensor.

            Returns:

           - mindspore.Tensor: The output tensor after applying the Phi3MLP module.
    """
    def __init__(self, config):
        """
        Initializes an instance of the Phi3MLP class.

        Args:
            self: The instance of the Phi3MLP class.
            config:
                An object containing configuration settings for the Phi3MLP model.

                - Type: Any
                - Purpose: Specifies the configuration parameters for the model.
                - Restrictions: Must be a valid configuration object.

        Returns:
            None.

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

        self.config = config
        self.gate_up_proj = nn.Linear(config.hidden_size, 2 * config.intermediate_size, bias=False)
        self.down_proj = nn.Linear(config.intermediate_size, config.hidden_size, bias=False)

        self.activation_fn = ACT2FN[config.hidden_act]

    def forward(self, hidden_states: mindspore.Tensor) -> mindspore.Tensor:
        """
        This method forwards and processes the hidden states using the Phi3MLP class.

        Args:
            self (Phi3MLP): The instance of the Phi3MLP class.
            hidden_states (mindspore.Tensor): The input tensor containing the hidden states to be processed.

        Returns:
            mindspore.Tensor: The processed tensor representing the output of the method.

        Raises:
            None
        """
        up_states = self.gate_up_proj(hidden_states)

        gate, up_states = up_states.chunk(2, axis=-1)
        up_states = up_states * self.activation_fn(gate)

        return self.down_proj(up_states)

mindnlp.transformers.models.phi3.modeling_phi3.Phi3MLP.__init__(config)

Initializes an instance of the Phi3MLP class.

PARAMETER DESCRIPTION
self

The instance of the Phi3MLP class.
config

An object containing configuration settings for the Phi3MLP model.

  • Type: Any
  • Purpose: Specifies the configuration parameters for the model.
  • Restrictions: Must be a valid configuration object.
RETURNS DESCRIPTION

None.
Source code in mindnlp/transformers/models/phi3/modeling_phi3.py 
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
def __init__(self, config):
    """
    Initializes an instance of the Phi3MLP class.

    Args:
        self: The instance of the Phi3MLP class.
        config:
            An object containing configuration settings for the Phi3MLP model.

            - Type: Any
            - Purpose: Specifies the configuration parameters for the model.
            - Restrictions: Must be a valid configuration object.

    Returns:
        None.

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

    self.config = config
    self.gate_up_proj = nn.Linear(config.hidden_size, 2 * config.intermediate_size, bias=False)
    self.down_proj = nn.Linear(config.intermediate_size, config.hidden_size, bias=False)

    self.activation_fn = ACT2FN[config.hidden_act]

mindnlp.transformers.models.phi3.modeling_phi3.Phi3MLP.forward(hidden_states)

This method forwards and processes the hidden states using the Phi3MLP class.

PARAMETER DESCRIPTION
self

The instance of the Phi3MLP class.

TYPE: Phi3MLP

hidden_states

The input tensor containing the hidden states to be processed.

TYPE: Tensor

RETURNS DESCRIPTION
Tensor

mindspore.Tensor: The processed tensor representing the output of the method.
Source code in mindnlp/transformers/models/phi3/modeling_phi3.py 
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
def forward(self, hidden_states: mindspore.Tensor) -> mindspore.Tensor:
    """
    This method forwards and processes the hidden states using the Phi3MLP class.

    Args:
        self (Phi3MLP): The instance of the Phi3MLP class.
        hidden_states (mindspore.Tensor): The input tensor containing the hidden states to be processed.

    Returns:
        mindspore.Tensor: The processed tensor representing the output of the method.

    Raises:
        None
    """
    up_states = self.gate_up_proj(hidden_states)

    gate, up_states = up_states.chunk(2, axis=-1)
    up_states = up_states * self.activation_fn(gate)

    return self.down_proj(up_states)

mindnlp.transformers.models.phi3.modeling_phi3.Phi3Model

Bases: Phi3PreTrainedModel

Transformer decoder consisting of config.num_hidden_layers layers. Each layer is a [Phi3DecoderLayer]

PARAMETER DESCRIPTION
config

Phi3Config

TYPE: Phi3Config

Source code in mindnlp/transformers/models/phi3/modeling_phi3.py 
1038
1039
1040
1041
1042
1043
1044
1045
1046
1047
1048
1049
1050
1051
1052
1053
1054
1055
1056
1057
1058
1059
1060
1061
1062
1063
1064
1065
1066
1067
1068
1069
1070
1071
1072
1073
1074
1075
1076
1077
1078
1079
1080
1081
1082
1083
1084
1085
1086
1087
1088
1089
1090
1091
1092
1093
1094
1095
1096
1097
1098
1099
1100
1101
1102
1103
1104
1105
1106
1107
1108
1109
1110
1111
1112
1113
1114
1115
1116
1117
1118
1119
1120
1121
1122
1123
1124
1125
1126
1127
1128
1129
1130
1131
1132
1133
1134
1135
1136
1137
1138
1139
1140
1141
1142
1143
1144
1145
1146
1147
1148
1149
1150
1151
1152
1153
1154
1155
1156
1157
1158
1159
1160
1161
1162
1163
1164
1165
1166
1167
1168
1169
1170
1171
1172
1173
1174
1175
1176
1177
1178
1179
1180
1181
1182
1183
1184
1185
1186
1187
1188
1189
1190
1191
1192
1193
1194
1195
1196
1197
1198
1199
1200
1201
1202
1203
1204
1205
1206
1207
1208
1209
1210
1211
1212
1213
1214
1215
1216
1217
1218
1219
1220
1221
1222
1223
1224
1225
1226
1227
1228
1229
1230
1231
1232
1233
1234
1235
1236
1237
1238
1239
1240
1241
1242
1243
1244
1245
1246
1247
1248
1249
1250
1251
1252
1253
1254
1255
1256
1257
1258
1259
1260
1261
1262
1263
1264
1265
1266
1267
1268
1269
1270
1271
1272
1273
1274
1275
1276
1277
1278
1279
1280
1281
1282
1283
1284
1285
1286
1287
1288
1289
1290
1291
1292
1293
1294
1295
1296
1297
1298
1299
1300
1301
1302
1303
1304
class Phi3Model(Phi3PreTrainedModel):
    """
    Transformer decoder consisting of *config.num_hidden_layers* layers. Each layer is a [`Phi3DecoderLayer`]

    Args:
        config: Phi3Config
    """
    def __init__(self, config: Phi3Config):
        """
        Initializes a new instance of the Phi3Model class.

        Args:
            self: The object instance.
            config (Phi3Config): The configuration object for Phi3Model.

        Returns:
            None

        Raises:
            None

        Description:
            This method initializes a new instance of the Phi3Model class. It takes in a configuration object,
            'config', which is of type Phi3Config. The 'config' parameter contains various settings and
            hyperparameters for the model.

        The method performs the following steps:

        1. Calls the __init__ method of the parent class (super().__init__(config)) to initialize the parent class
        with the provided configuration.
        2. Sets the 'padding_idx' attribute to the 'pad_token_id' value from the 'config' object.
        This value represents the padding token index in the vocabulary.
        3. Sets the 'vocab_size' attribute to the 'vocab_size' value from the 'config' object.
        This value represents the size of the vocabulary.
        4. Initializes the 'embed_tokens' attribute as an instance of the nn.Embedding class.
        It takes the 'vocab_size', 'hidden_size', and 'padding_idx' values from the 'config' object as parameters.
        This embedding layer is responsible for converting input tokens to their corresponding embeddings.
        5. Initializes the 'embed_dropout' attribute as an instance of the nn.Dropout class.
        It takes the 'embd_pdrop' value from the 'config' object as a parameter.
        This dropout layer is applied to the embeddings.
        6. Initializes the 'layers' attribute as an instance of the nn.ModuleList class.
        It contains Phi3DecoderLayer instances, one for each layer index from 0 to 'num_hidden_layers' - 1 (inclusive).
        Each Phi3DecoderLayer is initialized with the 'config' object and the corresponding layer index.
        7. Sets the '_attn_implementation' attribute to the '_attn_implementation' value from the 'config' object.
        This value represents the implementation type of the attention mechanism.
        8. Initializes the 'norm' attribute as an instance of the Phi3RMSNorm class. It takes the 'hidden_size'
        and 'eps' values from the 'config' object as parameters. This layer applies root mean square normalization to
        the hidden states.
        9. Sets the 'gradient_checkpointing' attribute to False. This attribute determines whether gradient
        checkpointing is enabled during training.
        10. Calls the 'post_init' method, which can be overridden by subclasses to perform additional initialization
        steps.

        Note:
            This method is called automatically when creating a new instance of the Phi3Model class.
        """
        super().__init__(config)
        self.padding_idx = config.pad_token_id
        self.vocab_size = config.vocab_size

        self.embed_tokens = nn.Embedding(config.vocab_size, config.hidden_size, self.padding_idx)
        self.embed_dropout = nn.Dropout(config.embd_pdrop)
        self.layers = nn.ModuleList(
            [Phi3DecoderLayer(config, layer_idx) for layer_idx in range(config.num_hidden_layers)]
        )
        self._attn_implementation = config._attn_implementation
        self.norm = Phi3RMSNorm(config.hidden_size, eps=config.rms_norm_eps)

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

    def get_input_embeddings(self):
        """
        This method retrieves the input embeddings for the Phi3Model.

        Args:
            self: The instance of the Phi3Model class.

        Returns:
            embed_tokens: The method returns the input embeddings stored in the 'embed_tokens' attribute of the
                Phi3Model instance.

        Raises:
            None.
        """
        return self.embed_tokens

    def set_input_embeddings(self, value):
        """
        Sets the input embeddings for the Phi3Model.

        Args:
            self: The instance of the Phi3Model class.
            value: A tensor representing the input embeddings. It should have a shape of
                (batch_size, sequence_length, embedding_dim).

        Returns:
            None.

        Raises:
            None.
        """
        self.embed_tokens = value

    def forward(
        self,
        input_ids: mindspore.Tensor = None,
        attention_mask: Optional[mindspore.Tensor] = None,
        position_ids: Optional[mindspore.Tensor] = None,
        past_key_values: Optional[List[mindspore.Tensor]] = None,
        inputs_embeds: Optional[mindspore.Tensor] = None,
        use_cache: Optional[bool] = None,
        output_attentions: Optional[bool] = None,
        output_hidden_states: Optional[bool] = None,
        return_dict: Optional[bool] = None,
    ) -> Union[Tuple, BaseModelOutputWithPast]:
        """
        Constructs the Phi3Model.

        Args:
            self: The object instance.
            input_ids (mindspore.Tensor, optional): The input tensor of shape (batch_size, seq_length). Defaults to None.
            attention_mask (Optional[mindspore.Tensor], optional): The attention mask tensor of shape
                (batch_size, seq_length). Defaults to None.
            position_ids (Optional[mindspore.Tensor], optional): The position ids tensor of shape
                (batch_size, seq_length). Defaults to None.
            past_key_values (Optional[List[mindspore.Tensor]], optional): List of past key value tensors.
                Defaults to None.
            inputs_embeds (Optional[mindspore.Tensor], optional): The input embeddings tensor of shape
                (batch_size, seq_length). Defaults to None.
            use_cache (Optional[bool], optional): Whether to use cache. Defaults to None.
            output_attentions (Optional[bool], optional): Whether to output attentions. Defaults to None.
            output_hidden_states (Optional[bool], optional): Whether to output hidden states. Defaults to None.
            return_dict (Optional[bool], optional): Whether to return a dictionary. Defaults to None.

        Returns:
            Union[Tuple, BaseModelOutputWithPast]: The output of the Phi3Model.

        Raises:
            ValueError: If both input_ids and inputs_embeds are specified.
            ValueError: If neither input_ids nor inputs_embeds are specified.
            ValueError: If attempting to perform batched generation with padding_side='right' in flash_attention_2
                implementation.
        """
        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
        )
        use_cache = use_cache if use_cache is not None else self.config.use_cache

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

        # retrieve input_ids and inputs_embeds
        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:
            batch_size, seq_length = input_ids.shape[:2]
        elif inputs_embeds is not None:
            batch_size, seq_length = inputs_embeds.shape[:2]
        else:
            raise ValueError("You have to specify either input_ids or inputs_embeds")

        past_key_values_length = 0

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

        if use_cache:
            use_legacy_cache = not isinstance(past_key_values, Cache)
            if use_legacy_cache:
                past_key_values = DynamicCache.from_legacy_cache(past_key_values)
            past_key_values_length = past_key_values.get_usable_length(seq_length)

        if position_ids is None:
            position_ids = ops.arange(
                past_key_values_length, seq_length + past_key_values_length, dtype=mindspore.int64
            )
            position_ids = position_ids.unsqueeze(0).view(-1, seq_length)
        else:
            position_ids = position_ids.view(-1, seq_length).long()

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

        if attention_mask is not None and self._attn_implementation == "flash_attention_2" and use_cache:
            is_padding_right = attention_mask[:, -1].sum().item() != batch_size
            if is_padding_right:
                raise ValueError(
                    "You are attempting to perform batched generation with padding_side='right'"
                    " this may lead to unexpected behaviour for Flash Attention version of Phi3. Make sure to "
                    " call `tokenizer.padding_side  = 'left'` before tokenizing the input. "
                )

        if self._attn_implementation == "flash_attention_2":
            # 2d mask is passed through the layers
            attention_mask = attention_mask if (attention_mask is not None and 0 in attention_mask) else None
        else:
            # 4d mask is passed through the layers
            attention_mask = _prepare_4d_causal_attention_mask(
                attention_mask,
                (batch_size, seq_length),
                inputs_embeds,
                past_key_values_length,
                sliding_window=self.config.sliding_window,
            )

        hidden_states = inputs_embeds

        # decoder layers
        all_hidden_states = () if output_hidden_states else None
        all_self_attns = () if output_attentions else None
        next_decoder_cache = None

        for decoder_layer in self.layers:
            if output_hidden_states:
                all_hidden_states += (hidden_states,)

            if self.gradient_checkpointing and self.training:
                layer_outputs = self._gradient_checkpointing_func(
                    decoder_layer.__call__,
                    hidden_states,
                    attention_mask,
                    position_ids,
                    past_key_values,
                    output_attentions,
                    use_cache,
                )
            else:
                layer_outputs = decoder_layer(
                    hidden_states,
                    attention_mask=attention_mask,
                    position_ids=position_ids,
                    past_key_value=past_key_values,
                    output_attentions=output_attentions,
                    use_cache=use_cache,
                )

            hidden_states = layer_outputs[0]

            if use_cache:
                next_decoder_cache = layer_outputs[2 if output_attentions else 1]

            if output_attentions:
                all_self_attns += (layer_outputs[1],)

        hidden_states = self.norm(hidden_states)

        # add hidden states from the last decoder layer
        if output_hidden_states:
            all_hidden_states += (hidden_states,)

        next_cache = None
        if use_cache:
            next_cache = next_decoder_cache.to_legacy_cache() if use_legacy_cache else next_decoder_cache
        if not return_dict:
            return tuple(v for v in [hidden_states, next_cache, all_hidden_states, all_self_attns] if v is not None)
        return BaseModelOutputWithPast(
            last_hidden_state=hidden_states,
            past_key_values=next_cache,
            hidden_states=all_hidden_states,
            attentions=all_self_attns,
        )

mindnlp.transformers.models.phi3.modeling_phi3.Phi3Model.__init__(config)

Initializes a new instance of the Phi3Model class.

PARAMETER DESCRIPTION
self

The object instance.
config

The configuration object for Phi3Model.

TYPE: Phi3Config

RETURNS DESCRIPTION

None
Description

This method initializes a new instance of the Phi3Model class. It takes in a configuration object, 'config', which is of type Phi3Config. The 'config' parameter contains various settings and hyperparameters for the model.

The method performs the following steps:

  1. Calls the init method of the parent class (super().init(config)) to initialize the parent class with the provided configuration.
  2. Sets the 'padding_idx' attribute to the 'pad_token_id' value from the 'config' object. This value represents the padding token index in the vocabulary.
  3. Sets the 'vocab_size' attribute to the 'vocab_size' value from the 'config' object. This value represents the size of the vocabulary.
  4. Initializes the 'embed_tokens' attribute as an instance of the nn.Embedding class. It takes the 'vocab_size', 'hidden_size', and 'padding_idx' values from the 'config' object as parameters. This embedding layer is responsible for converting input tokens to their corresponding embeddings.
  5. Initializes the 'embed_dropout' attribute as an instance of the nn.Dropout class. It takes the 'embd_pdrop' value from the 'config' object as a parameter. This dropout layer is applied to the embeddings.
  6. Initializes the 'layers' attribute as an instance of the nn.ModuleList class. It contains Phi3DecoderLayer instances, one for each layer index from 0 to 'num_hidden_layers' - 1 (inclusive). Each Phi3DecoderLayer is initialized with the 'config' object and the corresponding layer index.
  7. Sets the '_attn_implementation' attribute to the '_attn_implementation' value from the 'config' object. This value represents the implementation type of the attention mechanism.
  8. Initializes the 'norm' attribute as an instance of the Phi3RMSNorm class. It takes the 'hidden_size' and 'eps' values from the 'config' object as parameters. This layer applies root mean square normalization to the hidden states.
  9. Sets the 'gradient_checkpointing' attribute to False. This attribute determines whether gradient checkpointing is enabled during training.
  10. Calls the 'post_init' method, which can be overridden by subclasses to perform additional initialization steps.
Note

This method is called automatically when creating a new instance of the Phi3Model class.

Source code in mindnlp/transformers/models/phi3/modeling_phi3.py 
1045
1046
1047
1048
1049
1050
1051
1052
1053
1054
1055
1056
1057
1058
1059
1060
1061
1062
1063
1064
1065
1066
1067
1068
1069
1070
1071
1072
1073
1074
1075
1076
1077
1078
1079
1080
1081
1082
1083
1084
1085
1086
1087
1088
1089
1090
1091
1092
1093
1094
1095
1096
1097
1098
1099
1100
1101
1102
1103
1104
1105
1106
1107
1108
def __init__(self, config: Phi3Config):
    """
    Initializes a new instance of the Phi3Model class.

    Args:
        self: The object instance.
        config (Phi3Config): The configuration object for Phi3Model.

    Returns:
        None

    Raises:
        None

    Description:
        This method initializes a new instance of the Phi3Model class. It takes in a configuration object,
        'config', which is of type Phi3Config. The 'config' parameter contains various settings and
        hyperparameters for the model.

    The method performs the following steps:

    1. Calls the __init__ method of the parent class (super().__init__(config)) to initialize the parent class
    with the provided configuration.
    2. Sets the 'padding_idx' attribute to the 'pad_token_id' value from the 'config' object.
    This value represents the padding token index in the vocabulary.
    3. Sets the 'vocab_size' attribute to the 'vocab_size' value from the 'config' object.
    This value represents the size of the vocabulary.
    4. Initializes the 'embed_tokens' attribute as an instance of the nn.Embedding class.
    It takes the 'vocab_size', 'hidden_size', and 'padding_idx' values from the 'config' object as parameters.
    This embedding layer is responsible for converting input tokens to their corresponding embeddings.
    5. Initializes the 'embed_dropout' attribute as an instance of the nn.Dropout class.
    It takes the 'embd_pdrop' value from the 'config' object as a parameter.
    This dropout layer is applied to the embeddings.
    6. Initializes the 'layers' attribute as an instance of the nn.ModuleList class.
    It contains Phi3DecoderLayer instances, one for each layer index from 0 to 'num_hidden_layers' - 1 (inclusive).
    Each Phi3DecoderLayer is initialized with the 'config' object and the corresponding layer index.
    7. Sets the '_attn_implementation' attribute to the '_attn_implementation' value from the 'config' object.
    This value represents the implementation type of the attention mechanism.
    8. Initializes the 'norm' attribute as an instance of the Phi3RMSNorm class. It takes the 'hidden_size'
    and 'eps' values from the 'config' object as parameters. This layer applies root mean square normalization to
    the hidden states.
    9. Sets the 'gradient_checkpointing' attribute to False. This attribute determines whether gradient
    checkpointing is enabled during training.
    10. Calls the 'post_init' method, which can be overridden by subclasses to perform additional initialization
    steps.

    Note:
        This method is called automatically when creating a new instance of the Phi3Model class.
    """
    super().__init__(config)
    self.padding_idx = config.pad_token_id
    self.vocab_size = config.vocab_size

    self.embed_tokens = nn.Embedding(config.vocab_size, config.hidden_size, self.padding_idx)
    self.embed_dropout = nn.Dropout(config.embd_pdrop)
    self.layers = nn.ModuleList(
        [Phi3DecoderLayer(config, layer_idx) for layer_idx in range(config.num_hidden_layers)]
    )
    self._attn_implementation = config._attn_implementation
    self.norm = Phi3RMSNorm(config.hidden_size, eps=config.rms_norm_eps)

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

mindnlp.transformers.models.phi3.modeling_phi3.Phi3Model.forward(input_ids=None, attention_mask=None, position_ids=None, past_key_values=None, inputs_embeds=None, use_cache=None, output_attentions=None, output_hidden_states=None, return_dict=None)

Constructs the Phi3Model.

PARAMETER DESCRIPTION
self

The object instance.
input_ids

The input tensor of shape (batch_size, seq_length). Defaults to None.

TYPE: Tensor DEFAULT: None

attention_mask

The attention mask tensor of shape (batch_size, seq_length). Defaults to None.

TYPE: Optional[Tensor] DEFAULT: None

position_ids

The position ids tensor of shape (batch_size, seq_length). Defaults to None.

TYPE: Optional[Tensor] DEFAULT: None

past_key_values

List of past key value tensors. Defaults to None.

TYPE: Optional[List[Tensor]] DEFAULT: None

inputs_embeds

The input embeddings tensor of shape (batch_size, seq_length). Defaults to None.

TYPE: Optional[Tensor] DEFAULT: None

use_cache

Whether to use cache. Defaults to None.

TYPE: Optional[bool] DEFAULT: None

output_attentions

Whether to output attentions. Defaults to None.

TYPE: Optional[bool] DEFAULT: None

output_hidden_states

Whether to output hidden states. Defaults to None.

TYPE: Optional[bool] DEFAULT: None

return_dict

Whether to return a dictionary. Defaults to None.

TYPE: Optional[bool] DEFAULT: None

RETURNS DESCRIPTION
Union[Tuple, BaseModelOutputWithPast]

Union[Tuple, BaseModelOutputWithPast]: The output of the Phi3Model.
RAISES DESCRIPTION
ValueError

If both input_ids and inputs_embeds are specified.
ValueError

If neither input_ids nor inputs_embeds are specified.
ValueError

If attempting to perform batched generation with padding_side='right' in flash_attention_2 implementation.
Source code in mindnlp/transformers/models/phi3/modeling_phi3.py 
1143
1144
1145
1146
1147
1148
1149
1150
1151
1152
1153
1154
1155
1156
1157
1158
1159
1160
1161
1162
1163
1164
1165
1166
1167
1168
1169
1170
1171
1172
1173
1174
1175
1176
1177
1178
1179
1180
1181
1182
1183
1184
1185
1186
1187
1188
1189
1190
1191
1192
1193
1194
1195
1196
1197
1198
1199
1200
1201
1202
1203
1204
1205
1206
1207
1208
1209
1210
1211
1212
1213
1214
1215
1216
1217
1218
1219
1220
1221
1222
1223
1224
1225
1226
1227
1228
1229
1230
1231
1232
1233
1234
1235
1236
1237
1238
1239
1240
1241
1242
1243
1244
1245
1246
1247
1248
1249
1250
1251
1252
1253
1254
1255
1256
1257
1258
1259
1260
1261
1262
1263
1264
1265
1266
1267
1268
1269
1270
1271
1272
1273
1274
1275
1276
1277
1278
1279
1280
1281
1282
1283
1284
1285
1286
1287
1288
1289
1290
1291
1292
1293
1294
1295
1296
1297
1298
1299
1300
1301
1302
1303
1304
def forward(
    self,
    input_ids: mindspore.Tensor = None,
    attention_mask: Optional[mindspore.Tensor] = None,
    position_ids: Optional[mindspore.Tensor] = None,
    past_key_values: Optional[List[mindspore.Tensor]] = None,
    inputs_embeds: Optional[mindspore.Tensor] = None,
    use_cache: Optional[bool] = None,
    output_attentions: Optional[bool] = None,
    output_hidden_states: Optional[bool] = None,
    return_dict: Optional[bool] = None,
) -> Union[Tuple, BaseModelOutputWithPast]:
    """
    Constructs the Phi3Model.

    Args:
        self: The object instance.
        input_ids (mindspore.Tensor, optional): The input tensor of shape (batch_size, seq_length). Defaults to None.
        attention_mask (Optional[mindspore.Tensor], optional): The attention mask tensor of shape
            (batch_size, seq_length). Defaults to None.
        position_ids (Optional[mindspore.Tensor], optional): The position ids tensor of shape
            (batch_size, seq_length). Defaults to None.
        past_key_values (Optional[List[mindspore.Tensor]], optional): List of past key value tensors.
            Defaults to None.
        inputs_embeds (Optional[mindspore.Tensor], optional): The input embeddings tensor of shape
            (batch_size, seq_length). Defaults to None.
        use_cache (Optional[bool], optional): Whether to use cache. Defaults to None.
        output_attentions (Optional[bool], optional): Whether to output attentions. Defaults to None.
        output_hidden_states (Optional[bool], optional): Whether to output hidden states. Defaults to None.
        return_dict (Optional[bool], optional): Whether to return a dictionary. Defaults to None.

    Returns:
        Union[Tuple, BaseModelOutputWithPast]: The output of the Phi3Model.

    Raises:
        ValueError: If both input_ids and inputs_embeds are specified.
        ValueError: If neither input_ids nor inputs_embeds are specified.
        ValueError: If attempting to perform batched generation with padding_side='right' in flash_attention_2
            implementation.
    """
    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
    )
    use_cache = use_cache if use_cache is not None else self.config.use_cache

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

    # retrieve input_ids and inputs_embeds
    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:
        batch_size, seq_length = input_ids.shape[:2]
    elif inputs_embeds is not None:
        batch_size, seq_length = inputs_embeds.shape[:2]
    else:
        raise ValueError("You have to specify either input_ids or inputs_embeds")

    past_key_values_length = 0

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

    if use_cache:
        use_legacy_cache = not isinstance(past_key_values, Cache)
        if use_legacy_cache:
            past_key_values = DynamicCache.from_legacy_cache(past_key_values)
        past_key_values_length = past_key_values.get_usable_length(seq_length)

    if position_ids is None:
        position_ids = ops.arange(
            past_key_values_length, seq_length + past_key_values_length, dtype=mindspore.int64
        )
        position_ids = position_ids.unsqueeze(0).view(-1, seq_length)
    else:
        position_ids = position_ids.view(-1, seq_length).long()

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

    if attention_mask is not None and self._attn_implementation == "flash_attention_2" and use_cache:
        is_padding_right = attention_mask[:, -1].sum().item() != batch_size
        if is_padding_right:
            raise ValueError(
                "You are attempting to perform batched generation with padding_side='right'"
                " this may lead to unexpected behaviour for Flash Attention version of Phi3. Make sure to "
                " call `tokenizer.padding_side  = 'left'` before tokenizing the input. "
            )

    if self._attn_implementation == "flash_attention_2":
        # 2d mask is passed through the layers
        attention_mask = attention_mask if (attention_mask is not None and 0 in attention_mask) else None
    else:
        # 4d mask is passed through the layers
        attention_mask = _prepare_4d_causal_attention_mask(
            attention_mask,
            (batch_size, seq_length),
            inputs_embeds,
            past_key_values_length,
            sliding_window=self.config.sliding_window,
        )

    hidden_states = inputs_embeds

    # decoder layers
    all_hidden_states = () if output_hidden_states else None
    all_self_attns = () if output_attentions else None
    next_decoder_cache = None

    for decoder_layer in self.layers:
        if output_hidden_states:
            all_hidden_states += (hidden_states,)

        if self.gradient_checkpointing and self.training:
            layer_outputs = self._gradient_checkpointing_func(
                decoder_layer.__call__,
                hidden_states,
                attention_mask,
                position_ids,
                past_key_values,
                output_attentions,
                use_cache,
            )
        else:
            layer_outputs = decoder_layer(
                hidden_states,
                attention_mask=attention_mask,
                position_ids=position_ids,
                past_key_value=past_key_values,
                output_attentions=output_attentions,
                use_cache=use_cache,
            )

        hidden_states = layer_outputs[0]

        if use_cache:
            next_decoder_cache = layer_outputs[2 if output_attentions else 1]

        if output_attentions:
            all_self_attns += (layer_outputs[1],)

    hidden_states = self.norm(hidden_states)

    # add hidden states from the last decoder layer
    if output_hidden_states:
        all_hidden_states += (hidden_states,)

    next_cache = None
    if use_cache:
        next_cache = next_decoder_cache.to_legacy_cache() if use_legacy_cache else next_decoder_cache
    if not return_dict:
        return tuple(v for v in [hidden_states, next_cache, all_hidden_states, all_self_attns] if v is not None)
    return BaseModelOutputWithPast(
        last_hidden_state=hidden_states,
        past_key_values=next_cache,
        hidden_states=all_hidden_states,
        attentions=all_self_attns,
    )

mindnlp.transformers.models.phi3.modeling_phi3.Phi3Model.get_input_embeddings()

This method retrieves the input embeddings for the Phi3Model.

PARAMETER DESCRIPTION
self

The instance of the Phi3Model class.
RETURNS DESCRIPTION
embed_tokens

The method returns the input embeddings stored in the 'embed_tokens' attribute of the Phi3Model instance.
Source code in mindnlp/transformers/models/phi3/modeling_phi3.py 
1110
1111
1112
1113
1114
1115
1116
1117
1118
1119
1120
1121
1122
1123
1124
def get_input_embeddings(self):
    """
    This method retrieves the input embeddings for the Phi3Model.

    Args:
        self: The instance of the Phi3Model class.

    Returns:
        embed_tokens: The method returns the input embeddings stored in the 'embed_tokens' attribute of the
            Phi3Model instance.

    Raises:
        None.
    """
    return self.embed_tokens

mindnlp.transformers.models.phi3.modeling_phi3.Phi3Model.set_input_embeddings(value)

Sets the input embeddings for the Phi3Model.

PARAMETER DESCRIPTION
self

The instance of the Phi3Model class.
value

A tensor representing the input embeddings. It should have a shape of (batch_size, sequence_length, embedding_dim).
RETURNS DESCRIPTION

None.
Source code in mindnlp/transformers/models/phi3/modeling_phi3.py 
1126
1127
1128
1129
1130
1131
1132
1133
1134
1135
1136
1137
1138
1139
1140
1141
def set_input_embeddings(self, value):
    """
    Sets the input embeddings for the Phi3Model.

    Args:
        self: The instance of the Phi3Model class.
        value: A tensor representing the input embeddings. It should have a shape of
            (batch_size, sequence_length, embedding_dim).

    Returns:
        None.

    Raises:
        None.
    """
    self.embed_tokens = value

mindnlp.transformers.models.phi3.modeling_phi3.Phi3PreTrainedModel

Bases: PreTrainedModel

This class represents a Phi3PreTrainedModel, which is a subclass of PreTrainedModel.

Phi3PreTrainedModel inherits the following methods from PreTrainedModel:

  • forward(input_ids, attention_mask=None, token_type_ids=None, position_ids=None, head_mask=None): This method performs the forward pass for the Phi3PreTrainedModel. It takes input_ids as input and returns the model's output.

  • save_pretrained(save_directory): This method saves the model's weights and configuration to the specified directory.

  • from_pretrained(pretrained_model_name_or_path, *model_args, **kwargs): This method loads the pretrained model from the specified path or model name. Additional arguments can be passed to customize the loading process.

  • config_class: This attribute holds the configuration class of the model.

  • base_model_prefix: This attribute holds the prefix used by the model's modules.

The Phi3PreTrainedModel class introduces the following methods:

  • _init_weights: This method initializes the weights for the given module. If the module is of type nn.Linear, the weight is initialized using the Normal distribution with a standard deviation of self.config.initializer_range. If the module has a bias, it is initialized with zeros. If the module is of type nn.Embedding, the weight is randomly initialized using the Normal distribution with a standard deviation of self.config.initializer_range. If the module has a padding index, the weight at the padding index is set to zero.
Note

This class does not provide an implementation for the forward method. The implementation should be provided by subclasses that inherit from Phi3PreTrainedModel.

Source code in mindnlp/transformers/models/phi3/modeling_phi3.py 
 967
 968
 969
 970
 971
 972
 973
 974
 975
 976
 977
 978
 979
 980
 981
 982
 983
 984
 985
 986
 987
 988
 989
 990
 991
 992
 993
 994
 995
 996
 997
 998
 999
1000
1001
1002
1003
1004
1005
1006
1007
1008
1009
1010
1011
1012
1013
1014
1015
1016
1017
1018
1019
1020
1021
1022
1023
1024
1025
1026
1027
1028
1029
1030
1031
1032
1033
1034
1035
1036
class Phi3PreTrainedModel(PreTrainedModel):

    """
    This class represents a Phi3PreTrainedModel, which is a subclass of PreTrainedModel.

    Phi3PreTrainedModel inherits the following methods from PreTrainedModel:

    - forward(input_ids, attention_mask=None, token_type_ids=None, position_ids=None, head_mask=None):
    This method performs the forward pass for the Phi3PreTrainedModel. It takes input_ids as input and
    returns the model's output.

    - save_pretrained(save_directory):
    This method saves the model's weights and configuration to the specified directory.

    - from_pretrained(pretrained_model_name_or_path, *model_args, **kwargs):
    This method loads the pretrained model from the specified path or model name. Additional
    arguments can be passed to customize the loading process.

    - config_class:
    This attribute holds the configuration class of the model.

    - base_model_prefix:
    This attribute holds the prefix used by the model's modules.

    The Phi3PreTrainedModel class introduces the following methods:

    - _init_weights:
    This method initializes the weights for the given module. If the module is of type nn.Linear,
    the weight is initialized using the Normal distribution with a standard deviation of
    self.config.initializer_range. If the module has a bias, it is initialized with zeros.
    If the module is of type nn.Embedding, the weight is randomly initialized using the Normal
    distribution with a standard deviation of self.config.initializer_range. If the module has
    a padding index, the weight at the padding index is set to zero.

    Note:
        This class does not provide an implementation for the forward method. The implementation
        should be provided by subclasses that inherit from Phi3PreTrainedModel.
    """
    config_class = Phi3Config
    base_model_prefix = "model"
    supports_gradient_checkpointing = False
    _no_split_modules = ["Phi3DecoderLayer"]
    _skip_keys_device_placement = "past_key_values"
    _supports_cache_class = True

    def _init_weights(self, module):
        """
        Initializes the weights of a given module.

        Args:
            self (Phi3PreTrainedModel): The instance of the Phi3PreTrainedModel class.
            module: The module to initialize the weights for.

        Returns:
            None: This method modifies the weights of the given module in-place.

        Raises:
            None.
        """
        std = self.config.initializer_range
        if isinstance(module, nn.Linear):
            module.weight.initialize(Normal(std))
            if module.bias is not None:
                module.bias.initialize('zeros')
        elif isinstance(module, nn.Embedding):
            weight = np.random.normal(0.0, std, module.weight.shape)
            if module.padding_idx:
                weight[module.padding_idx] = 0

            module.weight.set_data(Tensor(weight, module.weight.dtype))

mindnlp.transformers.models.phi3.modeling_phi3.Phi3RMSNorm

Bases: Module

Phi3RMSNorm is a custom normalization layer that performs the Phi3 RMS normalization, equivalent to T5LayerNorm.

This class inherits from the nn.Module class in the MindSpore framework.

ATTRIBUTE DESCRIPTION
weight

The weight parameter for the normalization layer.

TYPE: Parameter

variance_epsilon

A small value added to the variance to avoid division by zero.

TYPE: float

METHOD DESCRIPTION
__init__

Initializes a new instance of the Phi3RMSNorm class.
forward

Applies Phi3 RMS normalization to the input hidden_states.
Source code in mindnlp/transformers/models/phi3/modeling_phi3.py 
 57
 58
 59
 60
 61
 62
 63
 64
 65
 66
 67
 68
 69
 70
 71
 72
 73
 74
 75
 76
 77
 78
 79
 80
 81
 82
 83
 84
 85
 86
 87
 88
 89
 90
 91
 92
 93
 94
 95
 96
 97
 98
 99
100
101
102
class Phi3RMSNorm(nn.Module):

    """ 
        Phi3RMSNorm is a custom normalization layer that performs the Phi3 RMS normalization, equivalent to T5LayerNorm.

        This class inherits from the nn.Module class in the MindSpore framework.

        Attributes:
            weight (Parameter): The weight parameter for the normalization layer.
            variance_epsilon (float): A small value added to the variance to avoid division by zero.

        Methods:
            __init__:
                Initializes a new instance of the Phi3RMSNorm class.

            forward:
                Applies Phi3 RMS normalization to the input hidden_states.
    """
    def __init__(self, hidden_size, eps=1e-6):
        """
        Phi3RMSNorm is equivalent to T5LayerNorm
        """
        super().__init__()
        self.weight = Parameter(ops.ones(hidden_size))
        self.variance_epsilon = eps

    def forward(self, hidden_states):
        """
        This method forwards Phi3RMSNorm by performing normalization on the hidden_states.

        Args:
            self: Instance of the Phi3RMSNorm class.
            hidden_states: A tensor containing the hidden states to be normalized.
                It should be of type 'Tensor' and compatible with the operations performed in the method.

        Returns:
            None.

        Raises:
            None
        """
        input_dtype = hidden_states.dtype
        hidden_states = hidden_states.to(mindspore.float32)
        variance = hidden_states.pow(2).mean(-1, keep_dims=True)
        hidden_states = hidden_states * ops.rsqrt(variance + self.variance_epsilon)
        return self.weight * hidden_states.to(input_dtype)

mindnlp.transformers.models.phi3.modeling_phi3.Phi3RMSNorm.__init__(hidden_size, eps=1e-06)

Phi3RMSNorm is equivalent to T5LayerNorm

Source code in mindnlp/transformers/models/phi3/modeling_phi3.py 
75
76
77
78
79
80
81
def __init__(self, hidden_size, eps=1e-6):
    """
    Phi3RMSNorm is equivalent to T5LayerNorm
    """
    super().__init__()
    self.weight = Parameter(ops.ones(hidden_size))
    self.variance_epsilon = eps

mindnlp.transformers.models.phi3.modeling_phi3.Phi3RMSNorm.forward(hidden_states)

This method forwards Phi3RMSNorm by performing normalization on the hidden_states.

PARAMETER DESCRIPTION
self

Instance of the Phi3RMSNorm class.
hidden_states

A tensor containing the hidden states to be normalized. It should be of type 'Tensor' and compatible with the operations performed in the method.
RETURNS DESCRIPTION

None.
Source code in mindnlp/transformers/models/phi3/modeling_phi3.py 
 83
 84
 85
 86
 87
 88
 89
 90
 91
 92
 93
 94
 95
 96
 97
 98
 99
100
101
102
def forward(self, hidden_states):
    """
    This method forwards Phi3RMSNorm by performing normalization on the hidden_states.

    Args:
        self: Instance of the Phi3RMSNorm class.
        hidden_states: A tensor containing the hidden states to be normalized.
            It should be of type 'Tensor' and compatible with the operations performed in the method.

    Returns:
        None.

    Raises:
        None
    """
    input_dtype = hidden_states.dtype
    hidden_states = hidden_states.to(mindspore.float32)
    variance = hidden_states.pow(2).mean(-1, keep_dims=True)
    hidden_states = hidden_states * ops.rsqrt(variance + self.variance_epsilon)
    return self.weight * hidden_states.to(input_dtype)

mindnlp.transformers.models.phi3.modeling_phi3.Phi3RotaryEmbedding

Bases: Module

This class represents the Phi3RotaryEmbedding, a rotary positional embedding layer used in neural network models. It is a subclass of nn.Module.

The Phi3RotaryEmbedding class provides methods for forwarding rotary embeddings based on input tensors and position IDs. It utilizes cosine and sine functions to generate embeddings with rotational properties.

ATTRIBUTE DESCRIPTION
dim

The dimension of the embeddings.

TYPE: int

max_position_embeddings

The maximum number of position embeddings.

TYPE: int

base

The base value for calculating inverse frequencies.

TYPE: int

inv_freq

The inverse frequencies calculated based on the dimension and base values.

TYPE: ndarray

METHOD DESCRIPTION
forward

Constructs rotary embeddings based on the input tensor and position IDs.

Args:

  • x (Tensor): The input tensor.
  • position_ids (Tensor): The position IDs.
  • seq_len (int, optional): The length of the sequence. Defaults to None.

Returns:

  • Tensor: The cosine and sine embeddings, converted to the same data type as the input tensor.
Source code in mindnlp/transformers/models/phi3/modeling_phi3.py 
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
class Phi3RotaryEmbedding(nn.Module):

    """
    This class represents the Phi3RotaryEmbedding, a rotary positional embedding layer used in neural network models.
    It is a subclass of nn.Module.

    The Phi3RotaryEmbedding class provides methods for forwarding rotary embeddings based on input tensors and
    position IDs. It utilizes cosine and sine functions to generate embeddings with rotational properties.

    Attributes:
        dim (int): The dimension of the embeddings.
        max_position_embeddings (int): The maximum number of position embeddings.
        base (int): The base value for calculating inverse frequencies.
        inv_freq (ndarray): The inverse frequencies calculated based on the dimension and base values.

    Methods:
        forward(x, position_ids, seq_len=None):
            Constructs rotary embeddings based on the input tensor and position IDs.

            Args:

            - x (Tensor): The input tensor.
            - position_ids (Tensor): The position IDs.
            - seq_len (int, optional): The length of the sequence. Defaults to None.

            Returns:

            - Tensor: The cosine and sine embeddings, converted to the same data type as the input tensor.
    """
    def __init__(self, dim, max_position_embeddings=2048, base=10000):
        """
        Initializes a new instance of the Phi3RotaryEmbedding class.

        Args:
            self: The instance of the class.
            dim (int): The dimension of the embedding.
            max_position_embeddings (int): The maximum number of position embeddings allowed (default is 2048).
            base (int): The base value for calculations (default is 10000).

        Returns:
            None.

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

        self.dim = dim
        self.max_position_embeddings = max_position_embeddings
        self.base = base
        self.inv_freq = None

    def forward(self, x, position_ids, seq_len=None):
        '''
        This method forwards the rotary embedding for the Phi3RotaryEmbedding class.

        Args:
            self (Phi3RotaryEmbedding): The instance of the Phi3RotaryEmbedding class.
            x (Tensor): The input tensor for which the rotary embedding is being forwarded.
            position_ids (Tensor): The tensor containing the position IDs.
            seq_len (int, optional): The length of the sequence. Defaults to None.

        Returns:
            Tuple[Tensor, Tensor]: Returns a tuple containing the cosine and sine values of the forwarded
                rotary embedding. Both tensors have the same shape as the input tensor x.

        Raises:
            ValueError: If the length of the position_ids tensor does not match the sequence length.
            TypeError: If the input parameters are not of the expected types.
        '''
        # x: [bs, num_attention_heads, seq_len, head_size]
        if self.inv_freq is None:
            self.inv_freq = 1.0 / (
                self.base ** (ops.arange(0, self.dim, 2, dtype=mindspore.int64).float() / self.dim)
            )
        inv_freq_expanded = self.inv_freq[None, :, None].float().expand(position_ids.shape[0], -1, 1)
        position_ids_expanded = position_ids[:, None, :].float()
        # Force float32 since bfloat16 loses precision on long contexts
        # See https://github.com/huggingface/transformers/pull/29285
        freqs = (inv_freq_expanded.float() @ position_ids_expanded.float()).swapaxes(1, 2)
        emb = ops.cat((freqs, freqs), axis=-1)
        cos = emb.cos()
        sin = emb.sin()
        return cos.to(dtype=x.dtype), sin.to(dtype=x.dtype)

mindnlp.transformers.models.phi3.modeling_phi3.Phi3RotaryEmbedding.__init__(dim, max_position_embeddings=2048, base=10000)

Initializes a new instance of the Phi3RotaryEmbedding class.

PARAMETER DESCRIPTION
self

The instance of the class.
dim

The dimension of the embedding.

TYPE: int

max_position_embeddings

The maximum number of position embeddings allowed (default is 2048).

TYPE: int DEFAULT: 2048

base

The base value for calculations (default is 10000).

TYPE: int DEFAULT: 10000

RETURNS DESCRIPTION

None.
Source code in mindnlp/transformers/models/phi3/modeling_phi3.py 
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
def __init__(self, dim, max_position_embeddings=2048, base=10000):
    """
    Initializes a new instance of the Phi3RotaryEmbedding class.

    Args:
        self: The instance of the class.
        dim (int): The dimension of the embedding.
        max_position_embeddings (int): The maximum number of position embeddings allowed (default is 2048).
        base (int): The base value for calculations (default is 10000).

    Returns:
        None.

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

    self.dim = dim
    self.max_position_embeddings = max_position_embeddings
    self.base = base
    self.inv_freq = None

mindnlp.transformers.models.phi3.modeling_phi3.Phi3RotaryEmbedding.forward(x, position_ids, seq_len=None)

This method forwards the rotary embedding for the Phi3RotaryEmbedding class.

PARAMETER DESCRIPTION
self

The instance of the Phi3RotaryEmbedding class.

TYPE: Phi3RotaryEmbedding

x

The input tensor for which the rotary embedding is being forwarded.

TYPE: Tensor

position_ids

The tensor containing the position IDs.

TYPE: Tensor

seq_len

The length of the sequence. Defaults to None.

TYPE: int DEFAULT: None

RETURNS DESCRIPTION

Tuple[Tensor, Tensor]: Returns a tuple containing the cosine and sine values of the forwarded rotary embedding. Both tensors have the same shape as the input tensor x.
RAISES DESCRIPTION
ValueError

If the length of the position_ids tensor does not match the sequence length.
TypeError

If the input parameters are not of the expected types.
Source code in mindnlp/transformers/models/phi3/modeling_phi3.py 
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
def forward(self, x, position_ids, seq_len=None):
    '''
    This method forwards the rotary embedding for the Phi3RotaryEmbedding class.

    Args:
        self (Phi3RotaryEmbedding): The instance of the Phi3RotaryEmbedding class.
        x (Tensor): The input tensor for which the rotary embedding is being forwarded.
        position_ids (Tensor): The tensor containing the position IDs.
        seq_len (int, optional): The length of the sequence. Defaults to None.

    Returns:
        Tuple[Tensor, Tensor]: Returns a tuple containing the cosine and sine values of the forwarded
            rotary embedding. Both tensors have the same shape as the input tensor x.

    Raises:
        ValueError: If the length of the position_ids tensor does not match the sequence length.
        TypeError: If the input parameters are not of the expected types.
    '''
    # x: [bs, num_attention_heads, seq_len, head_size]
    if self.inv_freq is None:
        self.inv_freq = 1.0 / (
            self.base ** (ops.arange(0, self.dim, 2, dtype=mindspore.int64).float() / self.dim)
        )
    inv_freq_expanded = self.inv_freq[None, :, None].float().expand(position_ids.shape[0], -1, 1)
    position_ids_expanded = position_ids[:, None, :].float()
    # Force float32 since bfloat16 loses precision on long contexts
    # See https://github.com/huggingface/transformers/pull/29285
    freqs = (inv_freq_expanded.float() @ position_ids_expanded.float()).swapaxes(1, 2)
    emb = ops.cat((freqs, freqs), axis=-1)
    cos = emb.cos()
    sin = emb.sin()
    return cos.to(dtype=x.dtype), sin.to(dtype=x.dtype)

mindnlp.transformers.models.phi3.modeling_phi3.Phi3SuScaledRotaryEmbedding

Bases: Phi3RotaryEmbedding

The Phi3SuScaledRotaryEmbedding class represents a specialized implementation of the Phi3RotaryEmbedding class, which provides functionalities for forwarding a scaled rotary embedding for a given input tensor.

ATTRIBUTE DESCRIPTION
`dim`

The dimensionality of the input tensor.

TYPE: int

`config`

The configuration object containing various parameters.

TYPE: object

`short_factor`

The scaling factor for short sequences.

TYPE: float

`long_factor`

The scaling factor for long sequences.

TYPE: float

`original_max_position_embeddings`

The maximum number of positions in the original input tensor.

TYPE: int

METHOD DESCRIPTION
`__init__`

Initializes the Phi3SuScaledRotaryEmbedding object.

Args:

  • dim (int): The dimensionality of the input tensor.
  • config (object): The configuration object containing various parameters.
`forward`

Constructs the scaled rotary embedding.

Args:

  • x (tensor): The input tensor.
  • position_ids (tensor): The tensor containing position indices.
  • seq_len (int, optional): The length of the sequence. Defaults to None.

Returns:

  • cos (tensor): The cosine component of the scaled rotary embedding.
  • sin (tensor): The sine component of the scaled rotary embedding.
Source code in mindnlp/transformers/models/phi3/modeling_phi3.py 
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
class Phi3SuScaledRotaryEmbedding(Phi3RotaryEmbedding):

    """
    The `Phi3SuScaledRotaryEmbedding` class represents a specialized implementation of the `Phi3RotaryEmbedding` class,
    which provides functionalities for forwarding a scaled rotary embedding for a given input tensor.

    Attributes:
        `dim` (int): The dimensionality of the input tensor.
        `config` (object): The configuration object containing various parameters.
        `short_factor` (float): The scaling factor for short sequences.
        `long_factor` (float): The scaling factor for long sequences.
        `original_max_position_embeddings` (int): The maximum number of positions in the original input tensor.

    Methods:
        `__init__`: Initializes the `Phi3SuScaledRotaryEmbedding` object.

            Args:

            - `dim` (int): The dimensionality of the input tensor.
            - `config` (object): The configuration object containing various parameters.

        `forward`: Constructs the scaled rotary embedding.

            Args:

            - `x` (tensor): The input tensor.
            - `position_ids` (tensor): The tensor containing position indices.
            - `seq_len` (int, optional): The length of the sequence. Defaults to None.

            Returns:

            - `cos` (tensor): The cosine component of the scaled rotary embedding.
            - `sin` (tensor): The sine component of the scaled rotary embedding.
    """
    def __init__(self, dim, config):
        """
        Initializes an instance of the Phi3SuScaledRotaryEmbedding class.

        Args:
            self (Phi3SuScaledRotaryEmbedding): The instance of the class.
            dim (int): The dimension of the embedding.
            config (object): The configuration object containing various settings.

        Returns:
            None.

        Raises:
            None.
        """
        super().__init__(dim, config.max_position_embeddings, config.rope_theta)

        self.short_factor = config.rope_scaling["short_factor"]
        self.long_factor = config.rope_scaling["long_factor"]
        self.original_max_position_embeddings = config.original_max_position_embeddings

    def forward(self, x, position_ids, seq_len=None):
        """
        Constructs the scaled rotary embedding for the Phi3SuScaledRotaryEmbedding.

        Args:
            self: The object instance.
            x (Tensor): The input tensor for which the scaled rotary embedding is forwarded.
            position_ids (Tensor): The position indices for the input tensor.
            seq_len (int, optional): The length of the sequence. Defaults to None.

        Returns:
            Tuple[Tensor, Tensor]: A tuple of tensors containing the cosine and sine of the scaled rotary embedding.

        Raises:
            ValueError: If the sequence length exceeds the original maximum position embeddings.
            TypeError: If the input tensors are not of the expected data type.
        """
        seq_len = ops.max(position_ids)[0] + 1
        if seq_len > self.original_max_position_embeddings:
            ext_factors = mindspore.Tensor(self.long_factor, dtype=mindspore.float32)
        else:
            ext_factors = mindspore.Tensor(self.short_factor, dtype=mindspore.float32)

        inv_freq_shape = ops.arange(0, self.dim, 2, dtype=mindspore.int64).float() / self.dim
        self.inv_freq = 1.0 / (ext_factors * self.base**inv_freq_shape)

        inv_freq_expanded = self.inv_freq[None, :, None].float().expand(position_ids.shape[0], -1, 1)
        position_ids_expanded = position_ids[:, None, :].float()

        # Force float32 since bfloat16 loses precision on long contexts
        # See https://github.com/huggingface/transformers/pull/29285
        freqs = (inv_freq_expanded.float() @ position_ids_expanded.float()).swapaxes(1, 2)
        emb = ops.cat((freqs, freqs), axis=-1)

        scale = self.max_position_embeddings / self.original_max_position_embeddings
        if scale <= 1.0:
            scaling_factor = 1.0
        else:
            scaling_factor = math.sqrt(1 + math.log(scale) / math.log(self.original_max_position_embeddings))

        cos = emb.cos() * scaling_factor
        sin = emb.sin() * scaling_factor
        return cos.to(dtype=x.dtype), sin.to(dtype=x.dtype)

mindnlp.transformers.models.phi3.modeling_phi3.Phi3SuScaledRotaryEmbedding.__init__(dim, config)

Initializes an instance of the Phi3SuScaledRotaryEmbedding class.

PARAMETER DESCRIPTION
self

The instance of the class.

TYPE: Phi3SuScaledRotaryEmbedding

dim

The dimension of the embedding.

TYPE: int

config

The configuration object containing various settings.

TYPE: object

RETURNS DESCRIPTION

None.
Source code in mindnlp/transformers/models/phi3/modeling_phi3.py 
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
def __init__(self, dim, config):
    """
    Initializes an instance of the Phi3SuScaledRotaryEmbedding class.

    Args:
        self (Phi3SuScaledRotaryEmbedding): The instance of the class.
        dim (int): The dimension of the embedding.
        config (object): The configuration object containing various settings.

    Returns:
        None.

    Raises:
        None.
    """
    super().__init__(dim, config.max_position_embeddings, config.rope_theta)

    self.short_factor = config.rope_scaling["short_factor"]
    self.long_factor = config.rope_scaling["long_factor"]
    self.original_max_position_embeddings = config.original_max_position_embeddings

mindnlp.transformers.models.phi3.modeling_phi3.Phi3SuScaledRotaryEmbedding.forward(x, position_ids, seq_len=None)

Constructs the scaled rotary embedding for the Phi3SuScaledRotaryEmbedding.

PARAMETER DESCRIPTION
self

The object instance.
x

The input tensor for which the scaled rotary embedding is forwarded.

TYPE: Tensor

position_ids

The position indices for the input tensor.

TYPE: Tensor

seq_len

The length of the sequence. Defaults to None.

TYPE: int DEFAULT: None

RETURNS DESCRIPTION

Tuple[Tensor, Tensor]: A tuple of tensors containing the cosine and sine of the scaled rotary embedding.
RAISES DESCRIPTION
ValueError

If the sequence length exceeds the original maximum position embeddings.
TypeError

If the input tensors are not of the expected data type.
Source code in mindnlp/transformers/models/phi3/modeling_phi3.py 
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
def forward(self, x, position_ids, seq_len=None):
    """
    Constructs the scaled rotary embedding for the Phi3SuScaledRotaryEmbedding.

    Args:
        self: The object instance.
        x (Tensor): The input tensor for which the scaled rotary embedding is forwarded.
        position_ids (Tensor): The position indices for the input tensor.
        seq_len (int, optional): The length of the sequence. Defaults to None.

    Returns:
        Tuple[Tensor, Tensor]: A tuple of tensors containing the cosine and sine of the scaled rotary embedding.

    Raises:
        ValueError: If the sequence length exceeds the original maximum position embeddings.
        TypeError: If the input tensors are not of the expected data type.
    """
    seq_len = ops.max(position_ids)[0] + 1
    if seq_len > self.original_max_position_embeddings:
        ext_factors = mindspore.Tensor(self.long_factor, dtype=mindspore.float32)
    else:
        ext_factors = mindspore.Tensor(self.short_factor, dtype=mindspore.float32)

    inv_freq_shape = ops.arange(0, self.dim, 2, dtype=mindspore.int64).float() / self.dim
    self.inv_freq = 1.0 / (ext_factors * self.base**inv_freq_shape)

    inv_freq_expanded = self.inv_freq[None, :, None].float().expand(position_ids.shape[0], -1, 1)
    position_ids_expanded = position_ids[:, None, :].float()

    # Force float32 since bfloat16 loses precision on long contexts
    # See https://github.com/huggingface/transformers/pull/29285
    freqs = (inv_freq_expanded.float() @ position_ids_expanded.float()).swapaxes(1, 2)
    emb = ops.cat((freqs, freqs), axis=-1)

    scale = self.max_position_embeddings / self.original_max_position_embeddings
    if scale <= 1.0:
        scaling_factor = 1.0
    else:
        scaling_factor = math.sqrt(1 + math.log(scale) / math.log(self.original_max_position_embeddings))

    cos = emb.cos() * scaling_factor
    sin = emb.sin() * scaling_factor
    return cos.to(dtype=x.dtype), sin.to(dtype=x.dtype)

mindnlp.transformers.models.phi3.modeling_phi3.Phi3YarnScaledRotaryEmbedding

Bases: Phi3RotaryEmbedding

This class represents the Phi3YarnScaledRotaryEmbedding, a subclass of Phi3RotaryEmbedding. It provides methods for forwarding scaled rotary embeddings for Phi3Yarn models.

ATTRIBUTE DESCRIPTION
dim

The dimension of the embeddings.

TYPE: int

config

The configuration object containing various parameters.

TYPE: object

short_factor

The scaling factor for short sequences.

TYPE: float

long_factor

The scaling factor for long sequences.

TYPE: float

original_max_position_embeddings

The original maximum position embeddings.

TYPE: int

METHOD DESCRIPTION
__init__

Initializes a Phi3YarnScaledRotaryEmbedding instance.
forward

Constructs the scaled rotary embeddings.
Source code in mindnlp/transformers/models/phi3/modeling_phi3.py 
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
class Phi3YarnScaledRotaryEmbedding(Phi3RotaryEmbedding):

    """
    This class represents the Phi3YarnScaledRotaryEmbedding, a subclass of Phi3RotaryEmbedding.
    It provides methods for forwarding scaled rotary embeddings for Phi3Yarn models.

    Attributes:
        dim (int): The dimension of the embeddings.
        config (object): The configuration object containing various parameters.
        short_factor (float): The scaling factor for short sequences.
        long_factor (float): The scaling factor for long sequences.
        original_max_position_embeddings (int): The original maximum position embeddings.

    Methods:
        __init__: Initializes a Phi3YarnScaledRotaryEmbedding instance.
        forward: Constructs the scaled rotary embeddings.

    """
    def __init__(self, dim, config):
        """
        Initializes a Phi3YarnScaledRotaryEmbedding object with the specified dimension and configuration.

        Args:
            self: The instance of the class.
            dim (int): The dimension of the embedding space.
            config (object): An object containing configuration parameters including max_position_embeddings,
                rope_theta, rope_scaling, and original_max_position_embeddings.

        Returns:
            None.

        Raises:
            KeyError: If the 'short_factor' or 'long_factor' keys are missing in the 'rope_scaling' dictionary within
                the config object.
            TypeError: If the 'max_position_embeddings' or 'original_max_position_embeddings' attributes are not present
                in the config object.
        """
        super().__init__(dim, config.max_position_embeddings, config.rope_theta)

        self.short_factor = config.rope_scaling["short_factor"]
        self.long_factor = config.rope_scaling["long_factor"]
        self.original_max_position_embeddings = config.original_max_position_embeddings

    def forward(self, x, position_ids, seq_len=None):
        """
        Constructs the Phi3YarnScaledRotaryEmbedding.

        Args:
            self: The instance of the Phi3YarnScaledRotaryEmbedding class.
            x: A tensor representing the input data.
            position_ids: A tensor containing the position IDs for each element in the input tensor.
            seq_len: An optional integer representing the length of the input sequence. If not provided, it is calculated as
                the maximum value in the position_ids tensor plus one.

        Returns:
            None

        Raises:
            None
        """
        seq_len = ops.max(position_ids) + 1
        if seq_len > self.original_max_position_embeddings:
            ext_factors = mindspore.Tensor(self.long_factor, dtype=mindspore.float32)
        else:
            ext_factors = mindspore.Tensor(self.short_factor, dtype=mindspore.float32)

        inv_freq_shape = ops.arange(0, self.dim, 2, dtype=mindspore.int64).float() / self.dim
        self.inv_freq = 1.0 / (ext_factors * self.base**inv_freq_shape)

        inv_freq_expanded = self.inv_freq[None, :, None].float().expand(position_ids.shape[0], -1, 1)
        position_ids_expanded = position_ids[:, None, :].float()

        # Force float32 since bfloat16 loses precision on long contexts
        # See https://github.com/huggingface/transformers/pull/29285
        freqs = (inv_freq_expanded.float() @ position_ids_expanded.float()).swapaxes(1, 2)
        emb = ops.cat((freqs, freqs), axis=-1)

        scale = self.max_position_embeddings / self.original_max_position_embeddings
        if scale <= 1.0:
            scaling_factor = 1.0
        else:
            scaling_factor = 0.1 * math.log(scale) + 1.0

        cos = emb.cos() * scaling_factor
        sin = emb.sin() * scaling_factor
        return cos.to(dtype=x.dtype), sin.to(dtype=x.dtype)

mindnlp.transformers.models.phi3.modeling_phi3.Phi3YarnScaledRotaryEmbedding.__init__(dim, config)

Initializes a Phi3YarnScaledRotaryEmbedding object with the specified dimension and configuration.

PARAMETER DESCRIPTION
self

The instance of the class.
dim

The dimension of the embedding space.

TYPE: int

config

An object containing configuration parameters including max_position_embeddings, rope_theta, rope_scaling, and original_max_position_embeddings.

TYPE: object

RETURNS DESCRIPTION

None.
RAISES DESCRIPTION
KeyError

If the 'short_factor' or 'long_factor' keys are missing in the 'rope_scaling' dictionary within the config object.
TypeError

If the 'max_position_embeddings' or 'original_max_position_embeddings' attributes are not present in the config object.
Source code in mindnlp/transformers/models/phi3/modeling_phi3.py 
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
def __init__(self, dim, config):
    """
    Initializes a Phi3YarnScaledRotaryEmbedding object with the specified dimension and configuration.

    Args:
        self: The instance of the class.
        dim (int): The dimension of the embedding space.
        config (object): An object containing configuration parameters including max_position_embeddings,
            rope_theta, rope_scaling, and original_max_position_embeddings.

    Returns:
        None.

    Raises:
        KeyError: If the 'short_factor' or 'long_factor' keys are missing in the 'rope_scaling' dictionary within
            the config object.
        TypeError: If the 'max_position_embeddings' or 'original_max_position_embeddings' attributes are not present
            in the config object.
    """
    super().__init__(dim, config.max_position_embeddings, config.rope_theta)

    self.short_factor = config.rope_scaling["short_factor"]
    self.long_factor = config.rope_scaling["long_factor"]
    self.original_max_position_embeddings = config.original_max_position_embeddings

mindnlp.transformers.models.phi3.modeling_phi3.Phi3YarnScaledRotaryEmbedding.forward(x, position_ids, seq_len=None)

Constructs the Phi3YarnScaledRotaryEmbedding.

PARAMETER DESCRIPTION
self

The instance of the Phi3YarnScaledRotaryEmbedding class.
x

A tensor representing the input data.
position_ids

A tensor containing the position IDs for each element in the input tensor.
seq_len

An optional integer representing the length of the input sequence. If not provided, it is calculated as the maximum value in the position_ids tensor plus one.

DEFAULT: None

RETURNS DESCRIPTION

None
Source code in mindnlp/transformers/models/phi3/modeling_phi3.py 
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
def forward(self, x, position_ids, seq_len=None):
    """
    Constructs the Phi3YarnScaledRotaryEmbedding.

    Args:
        self: The instance of the Phi3YarnScaledRotaryEmbedding class.
        x: A tensor representing the input data.
        position_ids: A tensor containing the position IDs for each element in the input tensor.
        seq_len: An optional integer representing the length of the input sequence. If not provided, it is calculated as
            the maximum value in the position_ids tensor plus one.

    Returns:
        None

    Raises:
        None
    """
    seq_len = ops.max(position_ids) + 1
    if seq_len > self.original_max_position_embeddings:
        ext_factors = mindspore.Tensor(self.long_factor, dtype=mindspore.float32)
    else:
        ext_factors = mindspore.Tensor(self.short_factor, dtype=mindspore.float32)

    inv_freq_shape = ops.arange(0, self.dim, 2, dtype=mindspore.int64).float() / self.dim
    self.inv_freq = 1.0 / (ext_factors * self.base**inv_freq_shape)

    inv_freq_expanded = self.inv_freq[None, :, None].float().expand(position_ids.shape[0], -1, 1)
    position_ids_expanded = position_ids[:, None, :].float()

    # Force float32 since bfloat16 loses precision on long contexts
    # See https://github.com/huggingface/transformers/pull/29285
    freqs = (inv_freq_expanded.float() @ position_ids_expanded.float()).swapaxes(1, 2)
    emb = ops.cat((freqs, freqs), axis=-1)

    scale = self.max_position_embeddings / self.original_max_position_embeddings
    if scale <= 1.0:
        scaling_factor = 1.0
    else:
        scaling_factor = 0.1 * math.log(scale) + 1.0

    cos = emb.cos() * scaling_factor
    sin = emb.sin() * scaling_factor
    return cos.to(dtype=x.dtype), sin.to(dtype=x.dtype)

mindnlp.transformers.models.phi3.modeling_phi3.apply_rotary_pos_emb(q, k, cos, sin, position_ids=None, unsqueeze_dim=1)

Applies Rotary Position Embedding to the query and key tensors.

PARAMETER DESCRIPTION
q

The query tensor.

TYPE: `mindspore.Tensor`

k

The key tensor.

TYPE: `mindspore.Tensor`

cos

The cosine part of the rotary embedding.

TYPE: `mindspore.Tensor`

sin

The sine part of the rotary embedding.

TYPE: `mindspore.Tensor`

position_ids

Deprecated and unused.

TYPE: `mindspore.Tensor`, *optional* DEFAULT: None

unsqueeze_dim

The 'unsqueeze_dim' argument specifies the dimension along which to unsqueeze cos[position_ids] and sin[position_ids] so that they can be properly broadcasted to the dimensions of q and k. For example, note that cos[position_ids] and sin[position_ids] have the shape [batch_size, seq_len, head_dim]. Then, if q and k have the shape [batch_size, heads, seq_len, head_dim], then setting unsqueeze_dim=1 makes cos[position_ids] and sin[position_ids] broadcastable to the shapes of q and k. Similarly, if q and k have the shape [batch_size, seq_len, heads, head_dim], then set unsqueeze_dim=2.

TYPE: `int`, *optional*, defaults to 1 DEFAULT: 1

RETURNS DESCRIPTION

tuple(mindspore.Tensor) comprising of the query and key tensors rotated using the Rotary Position Embedding.
Source code in mindnlp/transformers/models/phi3/modeling_phi3.py 
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
def apply_rotary_pos_emb(q, k, cos, sin, position_ids=None, unsqueeze_dim=1):
    """
    Applies Rotary Position Embedding to the query and key tensors.

    Args:
        q (`mindspore.Tensor`): The query tensor.
        k (`mindspore.Tensor`): The key tensor.
        cos (`mindspore.Tensor`): The cosine part of the rotary embedding.
        sin (`mindspore.Tensor`): The sine part of the rotary embedding.
        position_ids (`mindspore.Tensor`, *optional*):
            Deprecated and unused.
        unsqueeze_dim (`int`, *optional*, defaults to 1):
            The 'unsqueeze_dim' argument specifies the dimension along which to unsqueeze cos[position_ids] and
            sin[position_ids] so that they can be properly broadcasted to the dimensions of q and k. For example, note
            that cos[position_ids] and sin[position_ids] have the shape [batch_size, seq_len, head_dim]. Then, if q and
            k have the shape [batch_size, heads, seq_len, head_dim], then setting unsqueeze_dim=1 makes
            cos[position_ids] and sin[position_ids] broadcastable to the shapes of q and k. Similarly, if q and k have
            the shape [batch_size, seq_len, heads, head_dim], then set unsqueeze_dim=2.

    Returns:
        `tuple(mindspore.Tensor)` comprising of the query and key tensors rotated using the Rotary Position Embedding.
    """
    cos = cos.unsqueeze(unsqueeze_dim)
    sin = sin.unsqueeze(unsqueeze_dim)
    q_embed = (q * cos) + (rotate_half(q) * sin)
    k_embed = (k * cos) + (rotate_half(k) * sin)
    return q_embed, k_embed

mindnlp.transformers.models.phi3.modeling_phi3.repeat_kv(hidden_states, n_rep)

This is the equivalent of torch.repeat_interleave(x, dim=1, repeats=n_rep). The hidden states go from (batch, num_key_value_heads, seqlen, head_dim) to (batch, num_attention_heads, seqlen, head_dim)

Source code in mindnlp/transformers/models/phi3/modeling_phi3.py 
538
539
540
541
542
543
544
545
546
547
def repeat_kv(hidden_states: mindspore.Tensor, n_rep: int) -> mindspore.Tensor:
    """
    This is the equivalent of torch.repeat_interleave(x, dim=1, repeats=n_rep). The hidden states go from (batch,
    num_key_value_heads, seqlen, head_dim) to (batch, num_attention_heads, seqlen, head_dim)
    """
    batch, num_key_value_heads, slen, head_dim = hidden_states.shape
    if n_rep == 1:
        return hidden_states
    hidden_states = hidden_states[:, :, None, :, :].expand(batch, num_key_value_heads, n_rep, slen, head_dim)
    return hidden_states.reshape(batch, num_key_value_heads * n_rep, slen, head_dim)

mindnlp.transformers.models.phi3.modeling_phi3.rotate_half(x)

Rotates half the hidden dims of the input.

Source code in mindnlp/transformers/models/phi3/modeling_phi3.py 
419
420
421
422
423
424
def rotate_half(x):
    """Rotates half the hidden dims of the input."""
    # x1 = x[..., : x.shape[-1] // 2]
    # x2 = x[..., x.shape[-1] // 2 :]
    x1, x2 = x.tensor_split(2, -1)
    return ops.cat((-x2, x1), axis=-1)