File size: 6,478 Bytes
1ce5e18
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
# coding=utf-8
# Copyright Google AI and The HuggingFace Inc. team. All rights reserved.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
#     http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
""" CANINE model configuration"""

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


logger = logging.get_logger(__name__)

CANINE_PRETRAINED_CONFIG_ARCHIVE_MAP = {
    "google/canine-s": "https://huggingface.co/google/canine-s/resolve/main/config.json",
    # See all CANINE models at https://huggingface.co/models?filter=canine
}


class CanineConfig(PretrainedConfig):
    r"""
    This is the configuration class to store the configuration of a [`CanineModel`]. It is used to instantiate an
    CANINE 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 CANINE
    [google/canine-s](https://huggingface.co/google/canine-s) architecture.

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


    Args:
        hidden_size (`int`, *optional*, defaults to 768):
            Dimension of the encoder layers and the pooler layer.
        num_hidden_layers (`int`, *optional*, defaults to 12):
            Number of hidden layers in the deep Transformer encoder.
        num_attention_heads (`int`, *optional*, defaults to 12):
            Number of attention heads for each attention layer in the Transformer encoders.
        intermediate_size (`int`, *optional*, defaults to 3072):
            Dimension of the "intermediate" (i.e., feed-forward) layer in the Transformer encoders.
        hidden_act (`str` or `function`, *optional*, defaults to `"gelu"`):
            The non-linear activation function (function or string) in the encoder and pooler. If string, `"gelu"`,
            `"relu"`, `"selu"` and `"gelu_new"` are supported.
        hidden_dropout_prob (`float`, *optional*, defaults to 0.1):
            The dropout probabilitiy for all fully connected layers in the embeddings, encoders, and pooler.
        attention_probs_dropout_prob (`float`, *optional*, defaults to 0.1):
            The dropout ratio for the attention probabilities.
        max_position_embeddings (`int`, *optional*, defaults to 16384):
            The maximum sequence length that this model might ever be used with.
        type_vocab_size (`int`, *optional*, defaults to 16):
            The vocabulary size of the `token_type_ids` passed when calling [`CanineModel`].
        initializer_range (`float`, *optional*, defaults to 0.02):
            The standard deviation of the truncated_normal_initializer for initializing all weight matrices.
        layer_norm_eps (`float`, *optional*, defaults to 1e-12):
            The epsilon used by the layer normalization layers.
        downsampling_rate (`int`, *optional*, defaults to 4):
            The rate at which to downsample the original character sequence length before applying the deep Transformer
            encoder.
        upsampling_kernel_size (`int`, *optional*, defaults to 4):
            The kernel size (i.e. the number of characters in each window) of the convolutional projection layer when
            projecting back from `hidden_size`*2 to `hidden_size`.
        num_hash_functions (`int`, *optional*, defaults to 8):
            The number of hash functions to use. Each hash function has its own embedding matrix.
        num_hash_buckets (`int`, *optional*, defaults to 16384):
            The number of hash buckets to use.
        local_transformer_stride (`int`, *optional*, defaults to 128):
            The stride of the local attention of the first shallow Transformer encoder. Defaults to 128 for good
            TPU/XLA memory alignment.

    Example:

    ```python
    >>> from transformers import CanineConfig, CanineModel

    >>> # Initializing a CANINE google/canine-s style configuration
    >>> configuration = CanineConfig()

    >>> # Initializing a model (with random weights) from the google/canine-s style configuration
    >>> model = CanineModel(configuration)

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

    def __init__(
        self,
        hidden_size=768,
        num_hidden_layers=12,
        num_attention_heads=12,
        intermediate_size=3072,
        hidden_act="gelu",
        hidden_dropout_prob=0.1,
        attention_probs_dropout_prob=0.1,
        max_position_embeddings=16384,
        type_vocab_size=16,
        initializer_range=0.02,
        layer_norm_eps=1e-12,
        pad_token_id=0,
        bos_token_id=0xE000,
        eos_token_id=0xE001,
        downsampling_rate=4,
        upsampling_kernel_size=4,
        num_hash_functions=8,
        num_hash_buckets=16384,
        local_transformer_stride=128,  # Good TPU/XLA memory alignment.
        **kwargs,
    ):
        super().__init__(pad_token_id=pad_token_id, bos_token_id=bos_token_id, eos_token_id=eos_token_id, **kwargs)

        self.max_position_embeddings = max_position_embeddings
        self.hidden_size = hidden_size
        self.num_hidden_layers = num_hidden_layers
        self.num_attention_heads = num_attention_heads
        self.intermediate_size = intermediate_size
        self.hidden_act = hidden_act
        self.hidden_dropout_prob = hidden_dropout_prob
        self.attention_probs_dropout_prob = attention_probs_dropout_prob
        self.initializer_range = initializer_range
        self.type_vocab_size = type_vocab_size
        self.layer_norm_eps = layer_norm_eps

        # Character config:
        self.downsampling_rate = downsampling_rate
        self.upsampling_kernel_size = upsampling_kernel_size
        self.num_hash_functions = num_hash_functions
        self.num_hash_buckets = num_hash_buckets
        self.local_transformer_stride = local_transformer_stride