Spaces:
Sleeping
Sleeping
Update requirements.txt.file
Browse files- requirements.txt.file +2 -198
requirements.txt.file
CHANGED
@@ -1,198 +1,2 @@
|
|
1 |
-
|
2 |
-
|
3 |
-
#
|
4 |
-
# Licensed under the Apache License, Version 2.0 (the "License");
|
5 |
-
# you may not use this file except in compliance with the License.
|
6 |
-
# You may obtain a copy of the License at
|
7 |
-
#
|
8 |
-
# http://www.apache.org/licenses/LICENSE-2.0
|
9 |
-
#
|
10 |
-
# Unless required by applicable law or agreed to in writing, software
|
11 |
-
# distributed under the License is distributed on an "AS IS" BASIS,
|
12 |
-
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
|
13 |
-
# See the License for the specific language governing permissions and
|
14 |
-
# limitations under the License.
|
15 |
-
""" DeBERTa model configuration"""
|
16 |
-
from collections import OrderedDict
|
17 |
-
from typing import TYPE_CHECKING, Any, Mapping, Optional, Union
|
18 |
-
|
19 |
-
from ...configuration_utils import PretrainedConfig
|
20 |
-
from ...onnx import OnnxConfig
|
21 |
-
from ...utils import logging
|
22 |
-
|
23 |
-
|
24 |
-
if TYPE_CHECKING:
|
25 |
-
from ... import FeatureExtractionMixin, PreTrainedTokenizerBase, TensorType
|
26 |
-
|
27 |
-
|
28 |
-
logger = logging.get_logger(__name__)
|
29 |
-
|
30 |
-
DEBERTA_PRETRAINED_CONFIG_ARCHIVE_MAP = {
|
31 |
-
"microsoft/deberta-base": "https://huggingface.co/microsoft/deberta-base/resolve/main/config.json",
|
32 |
-
"microsoft/deberta-large": "https://huggingface.co/microsoft/deberta-large/resolve/main/config.json",
|
33 |
-
"microsoft/deberta-xlarge": "https://huggingface.co/microsoft/deberta-xlarge/resolve/main/config.json",
|
34 |
-
"microsoft/deberta-base-mnli": "https://huggingface.co/microsoft/deberta-base-mnli/resolve/main/config.json",
|
35 |
-
"microsoft/deberta-large-mnli": "https://huggingface.co/microsoft/deberta-large-mnli/resolve/main/config.json",
|
36 |
-
"microsoft/deberta-xlarge-mnli": "https://huggingface.co/microsoft/deberta-xlarge-mnli/resolve/main/config.json",
|
37 |
-
}
|
38 |
-
|
39 |
-
|
40 |
-
class DebertaConfig(PretrainedConfig):
|
41 |
-
r"""
|
42 |
-
This is the configuration class to store the configuration of a [`DebertaModel`] or a [`TFDebertaModel`]. It is
|
43 |
-
used to instantiate a DeBERTa model according to the specified arguments, defining the model architecture.
|
44 |
-
Instantiating a configuration with the defaults will yield a similar configuration to that of the DeBERTa
|
45 |
-
[microsoft/deberta-base](https://huggingface.co/microsoft/deberta-base) architecture.
|
46 |
-
|
47 |
-
Configuration objects inherit from [`PretrainedConfig`] and can be used to control the model outputs. Read the
|
48 |
-
documentation from [`PretrainedConfig`] for more information.
|
49 |
-
|
50 |
-
Arguments:
|
51 |
-
vocab_size (`int`, *optional*, defaults to 30522):
|
52 |
-
Vocabulary size of the DeBERTa model. Defines the number of different tokens that can be represented by the
|
53 |
-
`inputs_ids` passed when calling [`DebertaModel`] or [`TFDebertaModel`].
|
54 |
-
hidden_size (`int`, *optional*, defaults to 768):
|
55 |
-
Dimensionality of the encoder layers and the pooler layer.
|
56 |
-
num_hidden_layers (`int`, *optional*, defaults to 12):
|
57 |
-
Number of hidden layers in the Transformer encoder.
|
58 |
-
num_attention_heads (`int`, *optional*, defaults to 12):
|
59 |
-
Number of attention heads for each attention layer in the Transformer encoder.
|
60 |
-
intermediate_size (`int`, *optional*, defaults to 3072):
|
61 |
-
Dimensionality of the "intermediate" (often named feed-forward) layer in the Transformer encoder.
|
62 |
-
hidden_act (`str` or `Callable`, *optional*, defaults to `"gelu"`):
|
63 |
-
The non-linear activation function (function or string) in the encoder and pooler. If string, `"gelu"`,
|
64 |
-
`"relu"`, `"silu"`, `"gelu"`, `"tanh"`, `"gelu_fast"`, `"mish"`, `"linear"`, `"sigmoid"` and `"gelu_new"`
|
65 |
-
are supported.
|
66 |
-
hidden_dropout_prob (`float`, *optional*, defaults to 0.1):
|
67 |
-
The dropout probability for all fully connected layers in the embeddings, encoder, and pooler.
|
68 |
-
attention_probs_dropout_prob (`float`, *optional*, defaults to 0.1):
|
69 |
-
The dropout ratio for the attention probabilities.
|
70 |
-
max_position_embeddings (`int`, *optional*, defaults to 512):
|
71 |
-
The maximum sequence length that this model might ever be used with. Typically set this to something large
|
72 |
-
just in case (e.g., 512 or 1024 or 2048).
|
73 |
-
type_vocab_size (`int`, *optional*, defaults to 2):
|
74 |
-
The vocabulary size of the `token_type_ids` passed when calling [`DebertaModel`] or [`TFDebertaModel`].
|
75 |
-
initializer_range (`float`, *optional*, defaults to 0.02):
|
76 |
-
The standard deviation of the truncated_normal_initializer for initializing all weight matrices.
|
77 |
-
layer_norm_eps (`float`, *optional*, defaults to 1e-12):
|
78 |
-
The epsilon used by the layer normalization layers.
|
79 |
-
relative_attention (`bool`, *optional*, defaults to `False`):
|
80 |
-
Whether use relative position encoding.
|
81 |
-
max_relative_positions (`int`, *optional*, defaults to 1):
|
82 |
-
The range of relative positions `[-max_position_embeddings, max_position_embeddings]`. Use the same value
|
83 |
-
as `max_position_embeddings`.
|
84 |
-
pad_token_id (`int`, *optional*, defaults to 0):
|
85 |
-
The value used to pad input_ids.
|
86 |
-
position_biased_input (`bool`, *optional*, defaults to `True`):
|
87 |
-
Whether add absolute position embedding to content embedding.
|
88 |
-
pos_att_type (`List[str]`, *optional*):
|
89 |
-
The type of relative position attention, it can be a combination of `["p2c", "c2p"]`, e.g. `["p2c"]`,
|
90 |
-
`["p2c", "c2p"]`.
|
91 |
-
layer_norm_eps (`float`, optional, defaults to 1e-12):
|
92 |
-
The epsilon used by the layer normalization layers.
|
93 |
-
|
94 |
-
Example:
|
95 |
-
|
96 |
-
```python
|
97 |
-
>>> from transformers import DebertaConfig, DebertaModel
|
98 |
-
|
99 |
-
>>> # Initializing a DeBERTa microsoft/deberta-base style configuration
|
100 |
-
>>> configuration = DebertaConfig()
|
101 |
-
|
102 |
-
>>> # Initializing a model (with random weights) from the microsoft/deberta-base style configuration
|
103 |
-
>>> model = DebertaModel(configuration)
|
104 |
-
|
105 |
-
>>> # Accessing the model configuration
|
106 |
-
>>> configuration = model.config
|
107 |
-
```"""
|
108 |
-
model_type = "deberta"
|
109 |
-
|
110 |
-
def __init__(
|
111 |
-
self,
|
112 |
-
vocab_size=50265,
|
113 |
-
hidden_size=768,
|
114 |
-
num_hidden_layers=12,
|
115 |
-
num_attention_heads=12,
|
116 |
-
intermediate_size=3072,
|
117 |
-
hidden_act="gelu",
|
118 |
-
hidden_dropout_prob=0.1,
|
119 |
-
attention_probs_dropout_prob=0.1,
|
120 |
-
max_position_embeddings=512,
|
121 |
-
type_vocab_size=0,
|
122 |
-
initializer_range=0.02,
|
123 |
-
layer_norm_eps=1e-7,
|
124 |
-
relative_attention=False,
|
125 |
-
max_relative_positions=-1,
|
126 |
-
pad_token_id=0,
|
127 |
-
position_biased_input=True,
|
128 |
-
pos_att_type=None,
|
129 |
-
pooler_dropout=0,
|
130 |
-
pooler_hidden_act="gelu",
|
131 |
-
**kwargs
|
132 |
-
):
|
133 |
-
super().__init__(**kwargs)
|
134 |
-
|
135 |
-
self.hidden_size = hidden_size
|
136 |
-
self.num_hidden_layers = num_hidden_layers
|
137 |
-
self.num_attention_heads = num_attention_heads
|
138 |
-
self.intermediate_size = intermediate_size
|
139 |
-
self.hidden_act = hidden_act
|
140 |
-
self.hidden_dropout_prob = hidden_dropout_prob
|
141 |
-
self.attention_probs_dropout_prob = attention_probs_dropout_prob
|
142 |
-
self.max_position_embeddings = max_position_embeddings
|
143 |
-
self.type_vocab_size = type_vocab_size
|
144 |
-
self.initializer_range = initializer_range
|
145 |
-
self.relative_attention = relative_attention
|
146 |
-
self.max_relative_positions = max_relative_positions
|
147 |
-
self.pad_token_id = pad_token_id
|
148 |
-
self.position_biased_input = position_biased_input
|
149 |
-
|
150 |
-
# Backwards compatibility
|
151 |
-
if type(pos_att_type) == str:
|
152 |
-
pos_att_type = [x.strip() for x in pos_att_type.lower().split("|")]
|
153 |
-
|
154 |
-
self.pos_att_type = pos_att_type
|
155 |
-
self.vocab_size = vocab_size
|
156 |
-
self.layer_norm_eps = layer_norm_eps
|
157 |
-
|
158 |
-
self.pooler_hidden_size = kwargs.get("pooler_hidden_size", hidden_size)
|
159 |
-
self.pooler_dropout = pooler_dropout
|
160 |
-
self.pooler_hidden_act = pooler_hidden_act
|
161 |
-
|
162 |
-
|
163 |
-
# Copied from transformers.models.deberta_v2.configuration_deberta_v2.DebertaV2OnnxConfig
|
164 |
-
class DebertaOnnxConfig(OnnxConfig):
|
165 |
-
@property
|
166 |
-
def inputs(self) -> Mapping[str, Mapping[int, str]]:
|
167 |
-
if self.task == "multiple-choice":
|
168 |
-
dynamic_axis = {0: "batch", 1: "choice", 2: "sequence"}
|
169 |
-
else:
|
170 |
-
dynamic_axis = {0: "batch", 1: "sequence"}
|
171 |
-
if self._config.type_vocab_size > 0:
|
172 |
-
return OrderedDict(
|
173 |
-
[("input_ids", dynamic_axis), ("attention_mask", dynamic_axis), ("token_type_ids", dynamic_axis)]
|
174 |
-
)
|
175 |
-
else:
|
176 |
-
return OrderedDict([("input_ids", dynamic_axis), ("attention_mask", dynamic_axis)])
|
177 |
-
|
178 |
-
@property
|
179 |
-
def default_onnx_opset(self) -> int:
|
180 |
-
return 12
|
181 |
-
|
182 |
-
def generate_dummy_inputs(
|
183 |
-
self,
|
184 |
-
preprocessor: Union["PreTrainedTokenizerBase", "FeatureExtractionMixin"],
|
185 |
-
batch_size: int = -1,
|
186 |
-
seq_length: int = -1,
|
187 |
-
num_choices: int = -1,
|
188 |
-
is_pair: bool = False,
|
189 |
-
framework: Optional["TensorType"] = None,
|
190 |
-
num_channels: int = 3,
|
191 |
-
image_width: int = 40,
|
192 |
-
image_height: int = 40,
|
193 |
-
tokenizer: "PreTrainedTokenizerBase" = None,
|
194 |
-
) -> Mapping[str, Any]:
|
195 |
-
dummy_inputs = super().generate_dummy_inputs(preprocessor=preprocessor, framework=framework)
|
196 |
-
if self._config.type_vocab_size == 0 and "token_type_ids" in dummy_inputs:
|
197 |
-
del dummy_inputs["token_type_ids"]
|
198 |
-
return dummy_inputs
|
|
|
1 |
+
transformers
|
2 |
+
torch
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|