kaanakdeniz
commited on
Commit
•
40dfc44
1
Parent(s):
5b4ff05
Upload 9 files
Browse files- config.json +112 -0
- configuration_hat.py +150 -0
- merges.txt +0 -0
- modelling_hat.py +0 -0
- pytorch_model.bin +3 -0
- special_tokens_map.json +15 -0
- tokenizer_config.json +64 -0
- training_args.bin +3 -0
- vocab.json +0 -0
config.json
ADDED
@@ -0,0 +1,112 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"_name_or_path": "kiddothe2b/hierarchical-transformer-base-4096",
|
3 |
+
"architectures": [
|
4 |
+
"HATForSequenceClassification"
|
5 |
+
],
|
6 |
+
"attention_probs_dropout_prob": 0.1,
|
7 |
+
"auto_map": {
|
8 |
+
"AutoConfig": "configuration_hat.HATConfig",
|
9 |
+
"AutoModel": "modelling_hat.HATModel",
|
10 |
+
"AutoModelForMaskedLM": "modelling_hat.HATForMaskedLM",
|
11 |
+
"AutoModelForMultipleChoice": "modelling_hat.HATForMultipleChoice",
|
12 |
+
"AutoModelForQuestionAnswering": "modelling_hat.HATForQuestionAnswering",
|
13 |
+
"AutoModelForSequenceClassification": "modelling_hat.HATForSequenceClassification",
|
14 |
+
"AutoModelForTokenClassification": "modelling_hat.HATForTokenClassification",
|
15 |
+
"AutoTokenizer": "tokenization_hat.HATTokenizer"
|
16 |
+
},
|
17 |
+
"bos_token_id": 0,
|
18 |
+
"classifier_dropout": null,
|
19 |
+
"encoder_layout": {
|
20 |
+
"0": {
|
21 |
+
"document_encoder": false,
|
22 |
+
"sentence_encoder": true
|
23 |
+
},
|
24 |
+
"1": {
|
25 |
+
"document_encoder": false,
|
26 |
+
"sentence_encoder": true
|
27 |
+
},
|
28 |
+
"10": {
|
29 |
+
"document_encoder": false,
|
30 |
+
"sentence_encoder": true
|
31 |
+
},
|
32 |
+
"11": {
|
33 |
+
"document_encoder": true,
|
34 |
+
"sentence_encoder": true
|
35 |
+
},
|
36 |
+
"2": {
|
37 |
+
"document_encoder": true,
|
38 |
+
"sentence_encoder": true
|
39 |
+
},
|
40 |
+
"3": {
|
41 |
+
"document_encoder": false,
|
42 |
+
"sentence_encoder": true
|
43 |
+
},
|
44 |
+
"4": {
|
45 |
+
"document_encoder": false,
|
46 |
+
"sentence_encoder": true
|
47 |
+
},
|
48 |
+
"5": {
|
49 |
+
"document_encoder": true,
|
50 |
+
"sentence_encoder": true
|
51 |
+
},
|
52 |
+
"6": {
|
53 |
+
"document_encoder": false,
|
54 |
+
"sentence_encoder": true
|
55 |
+
},
|
56 |
+
"7": {
|
57 |
+
"document_encoder": false,
|
58 |
+
"sentence_encoder": true
|
59 |
+
},
|
60 |
+
"8": {
|
61 |
+
"document_encoder": true,
|
62 |
+
"sentence_encoder": true
|
63 |
+
},
|
64 |
+
"9": {
|
65 |
+
"document_encoder": false,
|
66 |
+
"sentence_encoder": true
|
67 |
+
}
|
68 |
+
},
|
69 |
+
"eos_token_id": 2,
|
70 |
+
"hidden_act": "gelu",
|
71 |
+
"hidden_dropout_prob": 0.1,
|
72 |
+
"hidden_size": 768,
|
73 |
+
"id2label": {
|
74 |
+
"0": "LABEL_0",
|
75 |
+
"1": "LABEL_1",
|
76 |
+
"2": "LABEL_2",
|
77 |
+
"3": "LABEL_3",
|
78 |
+
"4": "LABEL_4",
|
79 |
+
"5": "LABEL_5",
|
80 |
+
"6": "LABEL_6"
|
81 |
+
},
|
82 |
+
"initializer_range": 0.02,
|
83 |
+
"intermediate_size": 3072,
|
84 |
+
"label2id": {
|
85 |
+
"LABEL_0": 0,
|
86 |
+
"LABEL_1": 1,
|
87 |
+
"LABEL_2": 2,
|
88 |
+
"LABEL_3": 3,
|
89 |
+
"LABEL_4": 4,
|
90 |
+
"LABEL_5": 5,
|
91 |
+
"LABEL_6": 6
|
92 |
+
},
|
93 |
+
"layer_norm_eps": 1e-12,
|
94 |
+
"max_position_embeddings": 130,
|
95 |
+
"max_sentence_length": 128,
|
96 |
+
"max_sentence_size": 128,
|
97 |
+
"max_sentences": 32,
|
98 |
+
"model_max_length": 4096,
|
99 |
+
"model_type": "hierarchical-transformer",
|
100 |
+
"num_attention_heads": 12,
|
101 |
+
"num_hidden_layers": 12,
|
102 |
+
"output_past": true,
|
103 |
+
"pad_token_id": 1,
|
104 |
+
"parameters": 136350720,
|
105 |
+
"position_embedding_type": "absolute",
|
106 |
+
"problem_type": "single_label_classification",
|
107 |
+
"torch_dtype": "float32",
|
108 |
+
"transformers_version": "4.28.1",
|
109 |
+
"type_vocab_size": 1,
|
110 |
+
"use_cache": true,
|
111 |
+
"vocab_size": 50265
|
112 |
+
}
|
configuration_hat.py
ADDED
@@ -0,0 +1,150 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
# coding=utf-8
|
2 |
+
# Licensed under the Apache License, Version 2.0 (the "License");
|
3 |
+
# you may not use this file except in compliance with the License.
|
4 |
+
# You may obtain a copy of the License at
|
5 |
+
#
|
6 |
+
# http://www.apache.org/licenses/LICENSE-2.0
|
7 |
+
#
|
8 |
+
# Unless required by applicable law or agreed to in writing, software
|
9 |
+
# distributed under the License is distributed on an "AS IS" BASIS,
|
10 |
+
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
|
11 |
+
# See the License for the specific language governing permissions and
|
12 |
+
# limitations under the License.
|
13 |
+
""" HAT configuration"""
|
14 |
+
from collections import OrderedDict
|
15 |
+
from typing import Mapping
|
16 |
+
|
17 |
+
from transformers.onnx import OnnxConfig
|
18 |
+
from transformers.utils import logging
|
19 |
+
from transformers import PretrainedConfig
|
20 |
+
|
21 |
+
|
22 |
+
logger = logging.get_logger(__name__)
|
23 |
+
|
24 |
+
HAT_PRETRAINED_CONFIG_ARCHIVE_MAP = {
|
25 |
+
"kiddothe2b/hierarchical-transformer-base-4096": "https://huggingface.co/kiddothe2b/hierarchical-transformer-base-4096/resolve/main/config.json",
|
26 |
+
"kiddothe2b/adhoc-hierarchical-transformer-base-4096": "https://huggingface.co/kiddothe2b/adhoc-hierarchical-transformer-base-4096/resolve/main/config.json",
|
27 |
+
}
|
28 |
+
|
29 |
+
|
30 |
+
class HATConfig(PretrainedConfig):
|
31 |
+
r"""
|
32 |
+
This is the configuration class to store the configuration of a :class:`~transformers.HAT`.
|
33 |
+
It is used to instantiate a HAT model according to the specified arguments,
|
34 |
+
defining the model architecture. Instantiating a configuration with the defaults will yield a similar configuration
|
35 |
+
to that of the HAT `kiddothe2b/hierarchical-transformer-base-4096
|
36 |
+
<https://huggingface.co/kiddothe2b/hierarchical-transformer-base-4096>`__ architecture.
|
37 |
+
|
38 |
+
Configuration objects inherit from :class:`~transformers.PretrainedConfig` and can be used to control the model
|
39 |
+
outputs. Read the documentation from :class:`~transformers.PretrainedConfig` for more information.
|
40 |
+
|
41 |
+
|
42 |
+
Args:
|
43 |
+
vocab_size (:obj:`int`, `optional`, defaults to 30522):
|
44 |
+
Vocabulary size of the BERT model. Defines the number of different tokens that can be represented by the
|
45 |
+
:obj:`inputs_ids` passed when calling :class:`~transformers.BertModel` or
|
46 |
+
:class:`~transformers.TFBertModel`.
|
47 |
+
max_sentences (:obj:`int`, `optional`, defaults to 64):
|
48 |
+
The maximum number of sentences that this model might ever be used with.
|
49 |
+
max_sentence_size (:obj:`int`, `optional`, defaults to 128):
|
50 |
+
The maximum sentence length that this model might ever be used with.
|
51 |
+
model_max_length (:obj:`int`, `optional`, defaults to 8192):
|
52 |
+
The maximum sequence length (max_sentences * max_sentence_size) that this model might ever be used with
|
53 |
+
encoder_layout (:obj:`Dict`):
|
54 |
+
The sentence/document encoder layout.
|
55 |
+
hidden_size (:obj:`int`, `optional`, defaults to 768):
|
56 |
+
Dimensionality of the encoder layers and the pooler layer.
|
57 |
+
num_hidden_layers (:obj:`int`, `optional`, defaults to 12):
|
58 |
+
Number of hidden layers in the Transformer encoder.
|
59 |
+
num_attention_heads (:obj:`int`, `optional`, defaults to 12):
|
60 |
+
Number of attention heads for each attention layer in the Transformer encoder.
|
61 |
+
intermediate_size (:obj:`int`, `optional`, defaults to 3072):
|
62 |
+
Dimensionality of the "intermediate" (often named feed-forward) layer in the Transformer encoder.
|
63 |
+
hidden_act (:obj:`str` or :obj:`Callable`, `optional`, defaults to :obj:`"gelu"`):
|
64 |
+
The non-linear activation function (function or string) in the encoder and pooler. If string,
|
65 |
+
:obj:`"gelu"`, :obj:`"relu"`, :obj:`"silu"` and :obj:`"gelu_new"` are supported.
|
66 |
+
hidden_dropout_prob (:obj:`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 (:obj:`float`, `optional`, defaults to 0.1):
|
69 |
+
The dropout ratio for the attention probabilities.
|
70 |
+
max_position_embeddings (:obj:`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 (:obj:`int`, `optional`, defaults to 2):
|
74 |
+
The vocabulary size of the :obj:`token_type_ids` passed when calling :class:`~transformers.BertModel` or
|
75 |
+
:class:`~transformers.TFBertModel`.
|
76 |
+
initializer_range (:obj:`float`, `optional`, defaults to 0.02):
|
77 |
+
The standard deviation of the truncated_normal_initializer for initializing all weight matrices.
|
78 |
+
layer_norm_eps (:obj:`float`, `optional`, defaults to 1e-12):
|
79 |
+
The epsilon used by the layer normalization layers.
|
80 |
+
position_embedding_type (:obj:`str`, `optional`, defaults to :obj:`"absolute"`):
|
81 |
+
Type of position embedding. Choose one of :obj:`"absolute"`, :obj:`"relative_key"`,
|
82 |
+
:obj:`"relative_key_query"`. For positional embeddings use :obj:`"absolute"`. For more information on
|
83 |
+
:obj:`"relative_key"`, please refer to `Self-Attention with Relative Position Representations (Shaw et al.)
|
84 |
+
<https://arxiv.org/abs/1803.02155>`__. For more information on :obj:`"relative_key_query"`, please refer to
|
85 |
+
`Method 4` in `Improve Transformer Models with Better Relative Position Embeddings (Huang et al.)
|
86 |
+
<https://arxiv.org/abs/2009.13658>`__.
|
87 |
+
use_cache (:obj:`bool`, `optional`, defaults to :obj:`True`):
|
88 |
+
Whether or not the model should return the last key/values attentions (not used by all models). Only
|
89 |
+
relevant if ``config.is_decoder=True``.
|
90 |
+
classifier_dropout (:obj:`float`, `optional`):
|
91 |
+
The dropout ratio for the classification head.
|
92 |
+
"""
|
93 |
+
model_type = "hierarchical-transformer"
|
94 |
+
|
95 |
+
def __init__(
|
96 |
+
self,
|
97 |
+
vocab_size=30522,
|
98 |
+
hidden_size=768,
|
99 |
+
max_sentences=64,
|
100 |
+
max_sentence_size=128,
|
101 |
+
model_max_length=8192,
|
102 |
+
num_hidden_layers=12,
|
103 |
+
num_attention_heads=12,
|
104 |
+
intermediate_size=3072,
|
105 |
+
hidden_act="gelu",
|
106 |
+
hidden_dropout_prob=0.1,
|
107 |
+
attention_probs_dropout_prob=0.1,
|
108 |
+
max_position_embeddings=512,
|
109 |
+
type_vocab_size=2,
|
110 |
+
initializer_range=0.02,
|
111 |
+
layer_norm_eps=1e-12,
|
112 |
+
pad_token_id=0,
|
113 |
+
position_embedding_type="absolute",
|
114 |
+
encoder_layout=None,
|
115 |
+
use_cache=True,
|
116 |
+
classifier_dropout=None,
|
117 |
+
**kwargs
|
118 |
+
):
|
119 |
+
super().__init__(pad_token_id=pad_token_id, **kwargs)
|
120 |
+
|
121 |
+
self.vocab_size = vocab_size
|
122 |
+
self.hidden_size = hidden_size
|
123 |
+
self.max_sentences = max_sentences
|
124 |
+
self.max_sentence_size = max_sentence_size
|
125 |
+
self.model_max_length = model_max_length
|
126 |
+
self.encoder_layout = encoder_layout
|
127 |
+
self.num_hidden_layers = num_hidden_layers
|
128 |
+
self.num_attention_heads = num_attention_heads
|
129 |
+
self.hidden_act = hidden_act
|
130 |
+
self.intermediate_size = intermediate_size
|
131 |
+
self.hidden_dropout_prob = hidden_dropout_prob
|
132 |
+
self.attention_probs_dropout_prob = attention_probs_dropout_prob
|
133 |
+
self.max_position_embeddings = max_position_embeddings
|
134 |
+
self.type_vocab_size = type_vocab_size
|
135 |
+
self.initializer_range = initializer_range
|
136 |
+
self.layer_norm_eps = layer_norm_eps
|
137 |
+
self.position_embedding_type = position_embedding_type
|
138 |
+
self.use_cache = use_cache
|
139 |
+
self.classifier_dropout = classifier_dropout
|
140 |
+
|
141 |
+
|
142 |
+
class HATOnnxConfig(OnnxConfig):
|
143 |
+
@property
|
144 |
+
def inputs(self) -> Mapping[str, Mapping[int, str]]:
|
145 |
+
return OrderedDict(
|
146 |
+
[
|
147 |
+
("input_ids", {0: "batch", 1: "sequence"}),
|
148 |
+
("attention_mask", {0: "batch", 1: "sequence"}),
|
149 |
+
]
|
150 |
+
)
|
merges.txt
ADDED
The diff for this file is too large to render.
See raw diff
|
|
modelling_hat.py
ADDED
The diff for this file is too large to render.
See raw diff
|
|
pytorch_model.bin
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:18d6129fa43bac603ebc4a24bc9da4357eb13f80e17d5cc889c27374127c60b0
|
3 |
+
size 611062969
|
special_tokens_map.json
ADDED
@@ -0,0 +1,15 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"bos_token": "<s>",
|
3 |
+
"cls_token": "<s>",
|
4 |
+
"eos_token": "</s>",
|
5 |
+
"mask_token": {
|
6 |
+
"content": "<mask>",
|
7 |
+
"lstrip": true,
|
8 |
+
"normalized": false,
|
9 |
+
"rstrip": false,
|
10 |
+
"single_word": false
|
11 |
+
},
|
12 |
+
"pad_token": "<pad>",
|
13 |
+
"sep_token": "</s>",
|
14 |
+
"unk_token": "<unk>"
|
15 |
+
}
|
tokenizer_config.json
ADDED
@@ -0,0 +1,64 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"add_prefix_space": false,
|
3 |
+
"bos_token": {
|
4 |
+
"__type": "AddedToken",
|
5 |
+
"content": "<s>",
|
6 |
+
"lstrip": false,
|
7 |
+
"normalized": true,
|
8 |
+
"rstrip": false,
|
9 |
+
"single_word": false
|
10 |
+
},
|
11 |
+
"clean_up_tokenization_spaces": true,
|
12 |
+
"cls_token": {
|
13 |
+
"__type": "AddedToken",
|
14 |
+
"content": "<s>",
|
15 |
+
"lstrip": false,
|
16 |
+
"normalized": true,
|
17 |
+
"rstrip": false,
|
18 |
+
"single_word": false
|
19 |
+
},
|
20 |
+
"eos_token": {
|
21 |
+
"__type": "AddedToken",
|
22 |
+
"content": "</s>",
|
23 |
+
"lstrip": false,
|
24 |
+
"normalized": true,
|
25 |
+
"rstrip": false,
|
26 |
+
"single_word": false
|
27 |
+
},
|
28 |
+
"errors": "replace",
|
29 |
+
"mask_token": {
|
30 |
+
"__type": "AddedToken",
|
31 |
+
"content": "<mask>",
|
32 |
+
"lstrip": true,
|
33 |
+
"normalized": true,
|
34 |
+
"rstrip": false,
|
35 |
+
"single_word": false
|
36 |
+
},
|
37 |
+
"model_max_length": 4096,
|
38 |
+
"pad_token": {
|
39 |
+
"__type": "AddedToken",
|
40 |
+
"content": "<pad>",
|
41 |
+
"lstrip": false,
|
42 |
+
"normalized": true,
|
43 |
+
"rstrip": false,
|
44 |
+
"single_word": false
|
45 |
+
},
|
46 |
+
"sep_token": {
|
47 |
+
"__type": "AddedToken",
|
48 |
+
"content": "</s>",
|
49 |
+
"lstrip": false,
|
50 |
+
"normalized": true,
|
51 |
+
"rstrip": false,
|
52 |
+
"single_word": false
|
53 |
+
},
|
54 |
+
"tokenizer_class": "RobertaTokenizer",
|
55 |
+
"trim_offsets": true,
|
56 |
+
"unk_token": {
|
57 |
+
"__type": "AddedToken",
|
58 |
+
"content": "<unk>",
|
59 |
+
"lstrip": false,
|
60 |
+
"normalized": true,
|
61 |
+
"rstrip": false,
|
62 |
+
"single_word": false
|
63 |
+
}
|
64 |
+
}
|
training_args.bin
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:b0189a090bb8630e0991931c1c3072fbd7a79d80b1f6d42d4a2c7abf433d5ceb
|
3 |
+
size 3643
|
vocab.json
ADDED
The diff for this file is too large to render.
See raw diff
|
|