File size: 6,992 Bytes
bd4b02d
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
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
""" CrystalCoder configuration"""

from transformers.configuration_utils import PretrainedConfig
from transformers.utils import logging


logger = logging.get_logger(__name__)


class CrystalCoderConfig(PretrainedConfig):
    """
    This is the configuration class to store the configuration of a [`CrystalCoderModel`]. It is used to instantiate a CrystalCoder
    model according to the specified arguments, defining the model architecture.

    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 50257):
            Vocabulary size of the CrystalCoder model. Defines the number of different tokens that can be represented by the
            `inputs_ids` passed when calling [`CrystalCoderModel`].
        n_positions (`int`, *optional*, defaults to 1024):
            The maximum sequence length that this model might ever be used with. Typically set this to something large
            just in case (e.g., 512 or 1024 or 2048).
        n_embd (`int`, *optional*, defaults to 768):
            Dimensionality of the embeddings and hidden states.
        n_layer (`int`, *optional*, defaults to 12):
            Number of hidden layers in the Transformer encoder.
        n_head (`int`, *optional*, defaults to 12):
            Number of attention heads for each attention layer in the Transformer encoder.
        n_inner (`int`, *optional*, defaults to None):
            Dimensionality of the inner feed-forward layers. `None` will set it to 4 times n_embd
        activation_function (`str`, *optional*, defaults to `"gelu"`):
            Activation function, to be selected in the list `["relu", "silu", "gelu", "tanh", "gelu_new", "swiglu"]`.
        resid_pdrop (`float`, *optional*, defaults to 0.1):
            The dropout probability for all fully connected layers in the embeddings, encoder, and pooler.
        embd_pdrop (`float`, *optional*, defaults to 0.1):
            The dropout ratio for the embeddings.
        attn_pdrop (`float`, *optional*, defaults to 0.1):
            The dropout ratio for the attention.
        layer_norm_epsilon (`float`, *optional*, defaults to 1e-5):
            The epsilon to use in the layer normalization layers.
        initializer_range (`float`, *optional*, defaults to 0.02):
            The standard deviation of the truncated_normal_initializer for initializing all weight matrices.
        scale_attn_weights (`bool`, *optional*, defaults to `True`):
            Scale attention weights by dividing by sqrt(hidden_size)..
        use_cache (`bool`, *optional*, defaults to `True`):
            Whether or not the model should return the last key/values attentions (not used by all models).
        scale_attn_by_inverse_layer_idx (`bool`, *optional*, defaults to `False`):
            Whether to additionally scale attention weights by `1 / layer_idx + 1`.
        reorder_and_upcast_attn (`bool`, *optional*, defaults to `False`):
            Whether to scale keys (K) prior to computing attention (dot-product) and upcast attention
            dot-product/softmax to float() when training with mixed precision.
        position_embedding_type (`str`, *optional*, defaults to `"learned"`):
            Positional embedding can be either `"alibi"`, `"learned"`, or `"learned"`.
        rotary_dim (`int`, *optional*, defaults to `n_embd / n_head`):
            The dimension along which to apply rope.
        mup_width_scale (`float`, *optional*, defaults to 1.0):
            muP parameter to scale learning rate and initializers. Calculated as (`d_model,0 / d_model`), where
            `d_model` is the model's width and `d_model,0` is the proxy model's width.
        mup_embeddings_scale (`float`, *optional*, defaults to 1.0):
            muP parameter to scale token and position embeddings.
        mup_output_alpha (`float`, *optional*, defaults to 1.0):
            muP parameter to scale output logits (`output_logits_scale = mup_output_alpha * mup_width_scale`).
        mup_scale_qk_dot_by_d (`bool`, *optional*, defaults to `False`):
            Scale attention weights by dividing by hidden_size instead of sqrt(hidden_size). Need to set
            scale_attn_weights to `True` as well.

    Example:

    ```python
    >>> from transformers import CrystalCoderConfig, CrystalCoderModel

    >>> # Initializing a CrystalCoder configuration
    >>> configuration = CrystalCoderConfig()

    >>> # Initializing a model (with random weights) from the configuration
    >>> model = CrystalCoderModel(configuration)

    >>> # Accessing the model configuration
    >>> configuration = model.config
    ```"""

    model_type = "crystalcoder"
    keys_to_ignore_at_inference = ["past_key_values"]
    attribute_map = {
        "hidden_size": "n_embd",
        "max_position_embeddings": "n_positions",
        "num_attention_heads": "n_head",
        "num_hidden_layers": "n_layer",
    }

    def __init__(
        self,
        vocab_size=32032,
        n_positions=2048,
        n_embd=4096,
        n_layer=32,
        n_head=32,
        n_inner=None,
        activation_function="swiglu",
        resid_pdrop=0.1,
        embd_pdrop=0.1,
        attn_pdrop=0.1,
        layer_norm_epsilon=1e-5,
        initializer_range=0.02,
        scale_attn_weights=True,
        use_cache=True,
        bos_token_id=1,
        eos_token_id=2,
        scale_attn_by_inverse_layer_idx=False,
        reorder_and_upcast_attn=False,
        position_embedding_type="rotary",
        rotary_dim=None,
        mup_width_scale=1.0,
        mup_embeddings_scale=1.0,
        mup_output_alpha=1.0,
        mup_scale_qk_dot_by_d=False,
        **kwargs,
    ):
        self.vocab_size = vocab_size
        self.n_positions = n_positions
        self.n_embd = n_embd
        self.n_layer = n_layer
        self.n_head = n_head
        self.n_inner = n_inner
        self.activation_function = activation_function
        self.resid_pdrop = resid_pdrop
        self.embd_pdrop = embd_pdrop
        self.attn_pdrop = attn_pdrop
        self.layer_norm_epsilon = layer_norm_epsilon
        self.initializer_range = initializer_range
        self.scale_attn_weights = scale_attn_weights
        self.use_cache = use_cache
        self.scale_attn_by_inverse_layer_idx = scale_attn_by_inverse_layer_idx
        self.reorder_and_upcast_attn = reorder_and_upcast_attn

        self.bos_token_id = bos_token_id
        self.eos_token_id = eos_token_id

        self.position_embedding_type = position_embedding_type
        self.rotary_dim = rotary_dim
        self.mup_width_scale = mup_width_scale
        self.mup_embeddings_scale = mup_embeddings_scale
        self.mup_output_alpha = mup_output_alpha
        self.mup_scale_qk_dot_by_d = mup_scale_qk_dot_by_d

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