tuan.ljn commited on
Commit
946bef4
1 Parent(s): ebe9ebd

Add: add model

Browse files
1_Pooling/config.json ADDED
@@ -0,0 +1,10 @@
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "word_embedding_dimension": 768,
3
+ "pooling_mode_cls_token": true,
4
+ "pooling_mode_mean_tokens": false,
5
+ "pooling_mode_max_tokens": false,
6
+ "pooling_mode_mean_sqrt_len_tokens": false,
7
+ "pooling_mode_weightedmean_tokens": false,
8
+ "pooling_mode_lasttoken": false,
9
+ "include_prompt": true
10
+ }
README.md ADDED
@@ -0,0 +1,134 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ library_name: sentence-transformers
3
+ pipeline_tag: sentence-similarity
4
+ tags:
5
+ - sentence-transformers
6
+ - feature-extraction
7
+ - sentence-similarity
8
+ - transformers
9
+ - phobert
10
+ - vietnamese
11
+ - sentence-embedding
12
+ license: apache-2.0
13
+ language:
14
+ - vi
15
+ metrics:
16
+ - pearsonr
17
+ - spearmanr
18
+ ---
19
+ ## Model Description:
20
+ [**vietnamese-embedding-LongContext**](https://huggingface.co/dangvantuan/vietnamese-embedding-LongContext) is the Embedding Model for Vietnamese language with context length up to 8096 tokens. This model is a specialized sentence-embedding trained specifically for the Vietnamese language, which is built upon [gte-multilingual](Alibaba-NLP/gte-multilingual-base) and trained using the Multi-Negative Ranking Loss, Matryoshka2dLoss and SimilarityLoss.
21
+
22
+ ## Full Model Architecture
23
+ ```
24
+ SentenceTransformer(
25
+ (0): Transformer({'max_seq_length': 8192, 'do_lower_case': False}) with Transformer model: VietnameseModel
26
+ (1): Pooling({'word_embedding_dimension': 768, 'pooling_mode_cls_token': True, 'pooling_mode_mean_tokens': False, 'pooling_mode_max_tokens': False, 'pooling_mode_mean_sqrt_len_tokens': False, 'pooling_mode_weightedmean_tokens': False, 'pooling_mode_lasttoken': False, 'include_prompt': True})
27
+ (2): Normalize()
28
+ )
29
+ ```
30
+ ## Training and Fine-tuning process
31
+ The model underwent a rigorous four-stage training and fine-tuning process, each tailored to enhance its ability to generate precise and contextually relevant sentence embeddings for the Vietnamese language. Below is an outline of these stages:
32
+ #### Stage 1: Training NLI on dataset XNLI:
33
+ - Dataset: [XNLI-vn ](https://huggingface.co/datasets/xnli/viewer/vi)
34
+ - Method: Training using Multi-Negative Ranking Loss and Matryoshka2dLoss. This stage focused on improving the model's ability to discern and rank nuanced differences in sentence semantics.
35
+ ### Stage 2: Fine-tuning for Semantic Textual Similarity on STS Benchmark
36
+ - Dataset: [STSB-vn](https://huggingface.co/datasets/doanhieung/vi-stsbenchmark)
37
+ - Method: Fine-tuning specifically for the semantic textual similarity benchmark using Siamese BERT-Networks configured with the 'sentence-transformers' library. This stage honed the model's precision in capturing semantic similarity across various types of Vietnamese texts.
38
+
39
+
40
+ ## Usage:
41
+
42
+ Using this model becomes easy when you have [sentence-transformers](https://www.SBERT.net) installed:
43
+
44
+ ```
45
+ pip install -U sentence-transformers
46
+ ```
47
+
48
+ Then you can use the model like this:
49
+
50
+ ```python
51
+ from sentence_transformers import SentenceTransformer
52
+ sentences = ["Hà Nội là thủ đô của Việt Nam", "Đà Nẵng là thành phố du lịch"]
53
+
54
+
55
+ model = SentenceTransformer('dangvantuan/vietnamese-embedding-LongContext')
56
+ embeddings = model.encode(sentences)
57
+ print(embeddings)
58
+
59
+ ```
60
+
61
+
62
+ ## Evaluation
63
+ The model can be evaluated as follows on the [Vienamese data of stsb](https://huggingface.co/datasets/doanhieung/vi-stsbenchmark).
64
+
65
+ ```python
66
+ from sentence_transformers import SentenceTransformer
67
+ from sentence_transformers.readers import InputExample
68
+ from datasets import load_dataset
69
+ def convert_dataset(dataset):
70
+ dataset_samples=[]
71
+ for df in dataset:
72
+ score = float(df['score'])/5.0 # Normalize score to range 0 ... 1
73
+ inp_example = InputExample(texts=[df['sentence1'], df['sentence2']], label=score)
74
+ dataset_samples.append(inp_example)
75
+ return dataset_samples
76
+
77
+ # Loading the dataset for evaluation
78
+ vi_sts = load_dataset("doanhieung/vi-stsbenchmark")["train"]
79
+ df_dev = vi_sts.filter(lambda example: example['split'] == 'dev')
80
+ df_test = vi_sts.filter(lambda example: example['split'] == 'test')
81
+
82
+ # Convert the dataset for evaluation
83
+
84
+ # For Dev set:
85
+ dev_samples = convert_dataset(df_dev)
86
+ val_evaluator = EmbeddingSimilarityEvaluator.from_input_examples(dev_samples, name='sts-dev')
87
+ val_evaluator(model, output_path="./")
88
+
89
+ # For Test set:
90
+ test_samples = convert_dataset(df_test)
91
+ test_evaluator = EmbeddingSimilarityEvaluator.from_input_examples(test_samples, name='sts-test')
92
+ test_evaluator(model, output_path="./")
93
+ ```
94
+
95
+
96
+
97
+
98
+ ### Metric for all dataset of [Semantic Textual Similarity on STS Benchmark](https://huggingface.co/datasets/anti-ai/ViSTS)
99
+
100
+ **Spearman score**
101
+ | Model | [STSB] | [STS12]| [STS13] | [STS14] | [STS15] | [STS16] | [SICK] | Mean |
102
+ |-----------------------------------------------------------|---------|----------|----------|----------|----------|----------|---------|--------|
103
+ | [dangvantuan/vietnamese-embedding](https://huggingface.co/dangvantuan/vietnamese-embedding) |**84.84**| **79.04**| **85.30**| **81.38**| **87.06**| **79.95**| **79.58**| **82.45**|
104
+ | [dangvantuan/vietnamese-embedding-LongContext](https://huggingface.co/dangvantuan/vietnamese-embedding-LongContext) |85.25| 75.77| 83.82| 81.69| 88.48| 81.5| 78.2| 82.10|
105
+
106
+ ## Citation
107
+
108
+
109
+ @article{reimers2019sentence,
110
+ title={Sentence-BERT: Sentence Embeddings using Siamese BERT-Networks},
111
+ author={Nils Reimers, Iryna Gurevych},
112
+ journal={https://arxiv.org/abs/1908.10084},
113
+ year={2019}
114
+ }
115
+
116
+
117
+ @article{martin2020camembert,
118
+ title={CamemBERT: a Tasty French Language Mode},
119
+ author={Martin, Louis and Muller, Benjamin and Su{\'a}rez, Pedro Javier Ortiz and Dupont, Yoann and Romary, Laurent and de la Clergerie, {\'E}ric Villemonte and Seddah, Djam{\'e} and Sagot, Beno{\^\i}t},
120
+ journal={Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics},
121
+ year={2020}
122
+ }
123
+ @article{thakur2020augmented,
124
+ title={Augmented SBERT: Data Augmentation Method for Improving Bi-Encoders for Pairwise Sentence Scoring Tasks},
125
+ author={Thakur, Nandan and Reimers, Nils and Daxenberger, Johannes and Gurevych, Iryna},
126
+ journal={arXiv e-prints},
127
+ pages={arXiv--2010},
128
+ year={2020}
129
+ @article{li20242d,
130
+ title={2d matryoshka sentence embeddings},
131
+ author={Li, Xianming and Li, Zongxi and Li, Jing and Xie, Haoran and Li, Qing},
132
+ journal={arXiv preprint arXiv:2402.14776},
133
+ year={2024}
134
+ }
config.json ADDED
@@ -0,0 +1,50 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "_name_or_path": "dangvantuan/Vietnamese_impl",
3
+ "architectures": [
4
+ "VietnameseModel"
5
+ ],
6
+ "attention_probs_dropout_prob": 0.0,
7
+ "auto_map": {
8
+ "AutoConfig": "dangvantuan/Vietnamese_impl--configuration.VietnameseConfig",
9
+ "AutoModel": "dangvantuan/Vietnamese_impl--modeling.VietnameseModel",
10
+ "AutoModelForMaskedLM": "dangvantuan/Vietnamese_impl--modeling.VietnameseForMaskedLM",
11
+ "AutoModelForMultipleChoice": "dangvantuan/Vietnamese_impl--modeling.VietnameseForMultipleChoice",
12
+ "AutoModelForQuestionAnswering": "dangvantuan/Vietnamese_impl--modeling.VietnameseForQuestionAnswering",
13
+ "AutoModelForSequenceClassification": "dangvantuan/Vietnamese_impl--modeling.VietnameseForSequenceClassification",
14
+ "AutoModelForTokenClassification": "dangvantuan/Vietnamese_impl--modeling.VietnameseForTokenClassification"
15
+ },
16
+ "classifier_dropout": 0.0,
17
+ "hidden_act": "gelu",
18
+ "hidden_dropout_prob": 0.1,
19
+ "hidden_size": 768,
20
+ "id2label": {
21
+ "0": "LABEL_0"
22
+ },
23
+ "initializer_range": 0.02,
24
+ "intermediate_size": 3072,
25
+ "label2id": {
26
+ "LABEL_0": 0
27
+ },
28
+ "layer_norm_eps": 1e-12,
29
+ "layer_norm_type": "layer_norm",
30
+ "logn_attention_clip1": false,
31
+ "logn_attention_scale": false,
32
+ "max_position_embeddings": 8192,
33
+ "model_type": "Vietnamese",
34
+ "num_attention_heads": 12,
35
+ "num_hidden_layers": 12,
36
+ "pack_qkv": true,
37
+ "pad_token_id": 1,
38
+ "position_embedding_type": "rope",
39
+ "rope_scaling": {
40
+ "factor": 8.0,
41
+ "type": "ntk"
42
+ },
43
+ "rope_theta": 20000,
44
+ "torch_dtype": "float32",
45
+ "transformers_version": "4.42.3",
46
+ "type_vocab_size": 1,
47
+ "unpad_inputs": false,
48
+ "use_memory_efficient_attention": false,
49
+ "vocab_size": 250048
50
+ }
config_sentence_transformers.json ADDED
@@ -0,0 +1,9 @@
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "__version__": {
3
+ "sentence_transformers": "2.7.0",
4
+ "transformers": "4.42.3",
5
+ "pytorch": "2.2.1+cu121"
6
+ },
7
+ "prompts": {},
8
+ "default_prompt_name": null
9
+ }
configuration.py ADDED
@@ -0,0 +1,145 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ # coding=utf-8
2
+ # Copyright 2024 The GTE Team Authors and Alibaba Group.
3
+ # Copyright (c) 2018, NVIDIA CORPORATION. All rights reserved.
4
+ #
5
+ # Licensed under the Apache License, Version 2.0 (the "License");
6
+ # you may not use this file except in compliance with the License.
7
+ # You may obtain a copy of the License at
8
+ #
9
+ # http://www.apache.org/licenses/LICENSE-2.0
10
+ #
11
+ # Unless required by applicable law or agreed to in writing, software
12
+ # distributed under the License is distributed on an "AS IS" BASIS,
13
+ # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
14
+ # See the License for the specific language governing permissions and
15
+ # limitations under the License.
16
+ """ NEW model configuration"""
17
+ from transformers.configuration_utils import PretrainedConfig
18
+ from transformers.utils import logging
19
+
20
+ logger = logging.get_logger(__name__)
21
+
22
+
23
+ class NewConfig(PretrainedConfig):
24
+ r"""
25
+ This is the configuration class to store the configuration of a [`NewModel`] or a [`TFNewModel`]. It is used to
26
+ instantiate a NEW model according to the specified arguments, defining the model architecture. Instantiating a
27
+ configuration with the defaults will yield a similar configuration to that of the NEW
28
+ [izhx/new-base-en](https://huggingface.co/izhx/new-base-en) architecture.
29
+
30
+ Configuration objects inherit from [`PretrainedConfig`] and can be used to control the model outputs. Read the
31
+ documentation from [`PretrainedConfig`] for more information.
32
+
33
+
34
+ Args:
35
+ vocab_size (`int`, *optional*, defaults to 30522):
36
+ Vocabulary size of the NEW model. Defines the number of different tokens that can be represented by the
37
+ `inputs_ids` passed when calling [`NewModel`] or [`TFNewModel`].
38
+ hidden_size (`int`, *optional*, defaults to 768):
39
+ Dimensionality of the encoder layers and the pooler layer.
40
+ num_hidden_layers (`int`, *optional*, defaults to 12):
41
+ Number of hidden layers in the Transformer encoder.
42
+ num_attention_heads (`int`, *optional*, defaults to 12):
43
+ Number of attention heads for each attention layer in the Transformer encoder.
44
+ intermediate_size (`int`, *optional*, defaults to 3072):
45
+ Dimensionality of the "intermediate" (often named feed-forward) layer in the Transformer encoder.
46
+ hidden_act (`str` or `Callable`, *optional*, defaults to `"gelu"`):
47
+ The non-linear activation function (function or string) in the encoder and pooler. If string, `"gelu"`,
48
+ `"relu"`, `"silu"` and `"gelu_new"` are supported.
49
+ hidden_dropout_prob (`float`, *optional*, defaults to 0.1):
50
+ The dropout probability for all fully connected layers in the embeddings, encoder, and pooler.
51
+ attention_probs_dropout_prob (`float`, *optional*, defaults to 0.1):
52
+ The dropout ratio for the attention probabilities.
53
+ max_position_embeddings (`int`, *optional*, defaults to 512):
54
+ The maximum sequence length that this model might ever be used with. Typically set this to something large
55
+ just in case (e.g., 512 or 1024 or 2048).
56
+ type_vocab_size (`int`, *optional*, defaults to 2):
57
+ The vocabulary size of the `token_type_ids` passed when calling [`NewModel`] or [`TFNewModel`].
58
+ initializer_range (`float`, *optional*, defaults to 0.02):
59
+ The standard deviation of the truncated_normal_initializer for initializing all weight matrices.
60
+ layer_norm_eps (`float`, *optional*, defaults to 1e-12):
61
+ The epsilon used by the layer normalization layers.
62
+ position_embedding_type (`str`, *optional*, defaults to `"rope"`):
63
+ Type of position embedding. Choose one of `"absolute"`, `"rope"`.
64
+ rope_theta (`float`, *optional*, defaults to 10000.0):
65
+ The base period of the RoPE embeddings.
66
+ rope_scaling (`Dict`, *optional*):
67
+ Dictionary containing the scaling configuration for the RoPE embeddings. Currently supports two scaling
68
+ strategies: linear and dynamic. Their scaling factor must be a float greater than 1. The expected format is
69
+ `{"type": strategy name, "factor": scaling factor}`. When using this flag, don't update
70
+ `max_position_embeddings` to the expected new maximum. See the following thread for more information on how
71
+ these scaling strategies behave:
72
+ https://www.reddit.com/r/LocalLLaMA/comments/14mrgpr/dynamically_scaled_rope_further_increases/. This is an
73
+ experimental feature, subject to breaking API changes in future versions.
74
+ classifier_dropout (`float`, *optional*):
75
+ The dropout ratio for the classification head.
76
+
77
+ Examples:
78
+
79
+ ```python
80
+ >>> from transformers import NewConfig, NewModel
81
+
82
+ >>> # Initializing a NEW izhx/new-base-en style configuration
83
+ >>> configuration = NewConfig()
84
+
85
+ >>> # Initializing a model (with random weights) from the izhx/new-base-en style configuration
86
+ >>> model = NewModel(configuration)
87
+
88
+ >>> # Accessing the model configuration
89
+ >>> configuration = model.config
90
+ ```"""
91
+
92
+ model_type = "new"
93
+
94
+ def __init__(
95
+ self,
96
+ vocab_size=30528,
97
+ hidden_size=768,
98
+ num_hidden_layers=12,
99
+ num_attention_heads=12,
100
+ intermediate_size=3072,
101
+ hidden_act="gelu",
102
+ hidden_dropout_prob=0.1,
103
+ attention_probs_dropout_prob=0.0,
104
+ max_position_embeddings=2048,
105
+ type_vocab_size=1,
106
+ initializer_range=0.02,
107
+ layer_norm_type='layer_norm',
108
+ layer_norm_eps=1e-12,
109
+ # pad_token_id=0,
110
+ position_embedding_type="rope",
111
+ rope_theta=10000.0,
112
+ rope_scaling=None,
113
+ classifier_dropout=None,
114
+ pack_qkv=True,
115
+ unpad_inputs=False,
116
+ use_memory_efficient_attention=False,
117
+ logn_attention_scale=False,
118
+ logn_attention_clip1=False,
119
+ **kwargs,
120
+ ):
121
+ super().__init__(**kwargs)
122
+
123
+ self.vocab_size = vocab_size
124
+ self.hidden_size = hidden_size
125
+ self.num_hidden_layers = num_hidden_layers
126
+ self.num_attention_heads = num_attention_heads
127
+ self.hidden_act = hidden_act
128
+ self.intermediate_size = intermediate_size
129
+ self.hidden_dropout_prob = hidden_dropout_prob
130
+ self.attention_probs_dropout_prob = attention_probs_dropout_prob
131
+ self.max_position_embeddings = max_position_embeddings
132
+ self.type_vocab_size = type_vocab_size
133
+ self.initializer_range = initializer_range
134
+ self.layer_norm_type = layer_norm_type
135
+ self.layer_norm_eps = layer_norm_eps
136
+ self.position_embedding_type = position_embedding_type
137
+ self.rope_theta = rope_theta
138
+ self.rope_scaling = rope_scaling
139
+ self.classifier_dropout = classifier_dropout
140
+
141
+ self.pack_qkv = pack_qkv
142
+ self.unpad_inputs = unpad_inputs
143
+ self.use_memory_efficient_attention = use_memory_efficient_attention
144
+ self.logn_attention_scale = logn_attention_scale
145
+ self.logn_attention_clip1 = logn_attention_clip1
model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:065a1836422f393b06c40cc2aabb25c36c401692cdb38239e3fe4eb8b59fce23
3
+ size 1221487872
modeling.py ADDED
@@ -0,0 +1,1418 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ # coding=utf-8
2
+ # Copyright 2024 The GTE Team Authors and Alibaba Group.
3
+ # Copyright (c) 2018, NVIDIA CORPORATION. All rights reserved.
4
+ #
5
+ # Licensed under the Apache License, Version 2.0 (the "License");
6
+ # you may not use this file except in compliance with the License.
7
+ # You may obtain a copy of the License at
8
+ #
9
+ # http://www.apache.org/licenses/LICENSE-2.0
10
+ #
11
+ # Unless required by applicable law or agreed to in writing, software
12
+ # distributed under the License is distributed on an "AS IS" BASIS,
13
+ # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
14
+ # See the License for the specific language governing permissions and
15
+ # limitations under the License.
16
+ """PyTorch NEW model."""
17
+
18
+ import math
19
+ from dataclasses import dataclass
20
+ from typing import List, Optional, Tuple, Union
21
+
22
+ import torch
23
+ import torch.utils.checkpoint
24
+ from torch import nn
25
+
26
+ from transformers.activations import ACT2FN
27
+ from transformers.modeling_outputs import (
28
+ BaseModelOutput,
29
+ BaseModelOutputWithPooling,
30
+ MaskedLMOutput,
31
+ MultipleChoiceModelOutput,
32
+ QuestionAnsweringModelOutput,
33
+ SequenceClassifierOutput,
34
+ ModelOutput,
35
+ )
36
+ from transformers.modeling_utils import PreTrainedModel
37
+ from transformers.utils import logging
38
+
39
+ try:
40
+ import xformers.ops as xops
41
+ except ImportError as e:
42
+ xops = None
43
+
44
+ from .configuration import NewConfig
45
+
46
+
47
+ logger = logging.get_logger(__name__)
48
+
49
+
50
+ # Adapted from https://github.com/HazyResearch/flash-attention/blob/main/flash_attn/bert_padding.py
51
+ # Which was adapted from https://github.com/mlcommons/training_results_v1.1/blob/main/NVIDIA/benchmarks/bert/implementations/pytorch/padding.py
52
+ class IndexFirstAxis(torch.autograd.Function):
53
+ @staticmethod
54
+ def forward(ctx, input, indices):
55
+ ctx.save_for_backward(indices)
56
+ assert input.ndim >= 2
57
+ ctx.first_axis_dim, other_shape = input.shape[0], input.shape[1:]
58
+ second_dim = other_shape.numel()
59
+ # TD [2022-03-04] For some reason torch.gather is a bit faster than indexing.
60
+ # return input[indices]
61
+ # return torch.gather(
62
+ # rearrange(input, "b ... -> b (...)"), 0, repeat(indices, "z -> z d", d=second_dim)
63
+ # ).reshape(-1, *other_shape)
64
+ return torch.gather(
65
+ input.view(ctx.first_axis_dim, second_dim),
66
+ 0,
67
+ indices.unsqueeze(-1).expand(indices.size(0), second_dim)
68
+ ).reshape(-1, *other_shape)
69
+
70
+ @staticmethod
71
+ def backward(ctx, grad_output):
72
+ (indices,) = ctx.saved_tensors
73
+ assert grad_output.ndim >= 2
74
+ other_shape = grad_output.shape[1:]
75
+ # grad_output = rearrange(grad_output, "b ... -> b (...)")
76
+ grad_output = grad_output.view(grad_output.size(0), other_shape.numel())
77
+ grad_input = torch.zeros(
78
+ [ctx.first_axis_dim, grad_output.shape[1]],
79
+ device=grad_output.device,
80
+ dtype=grad_output.dtype,
81
+ )
82
+ # TD [2022-03-04] For some reason torch.scatter is a bit faster than indexing.
83
+ # grad_input[indices] = grad_output
84
+ # grad_input.scatter_(0, repeat(indices, "z -> z d", d=grad_output.shape[1]), grad_output)
85
+ grad_input.scatter_(
86
+ 0, indices.unsqueeze(-1).expand(indices.size(0), grad_output.size(1)), grad_output
87
+ )
88
+ return grad_input.reshape(ctx.first_axis_dim, *other_shape), None
89
+
90
+
91
+ index_first_axis = IndexFirstAxis.apply
92
+
93
+
94
+ def unpad_input(hidden_states, attention_mask=None, indices=None):
95
+ """
96
+ Arguments:
97
+ hidden_states: (batch, seqlen, ...)
98
+ attention_mask: (batch, seqlen), bool / int, 1 means valid and 0 means not valid.
99
+ indices: (total_nnz), the indices of non-masked tokens from the flattened input sequence.
100
+ Return:
101
+ hidden_states: (total_nnz, ...), where total_nnz = number of tokens in selected in attention_mask.
102
+ """
103
+ if indices is None:
104
+ assert attention_mask is not None
105
+ indices = torch.nonzero(attention_mask.flatten(), as_tuple=False).flatten()
106
+
107
+ # TD [2022-03-04] We don't want to index with a bool mask, because Pytorch will expand the
108
+ # bool mask, then call nonzero to get the indices, then index with those. The indices is @dim
109
+ # times larger than it needs to be, wasting memory. It's faster and more memory-efficient to
110
+ # index with integer indices. Moreover, torch's index is a bit slower than it needs to be,
111
+ # so we write custom forward and backward to make it a bit faster.
112
+ hidden_states = hidden_states.view(-1, *hidden_states.shape[2:])
113
+ return index_first_axis(hidden_states, indices)
114
+
115
+
116
+ class IndexPutFirstAxis(torch.autograd.Function):
117
+ @staticmethod
118
+ def forward(
119
+ ctx,
120
+ values: torch.Tensor,
121
+ indices: torch.Tensor,
122
+ first_axis_dim
123
+ ) -> torch.Tensor:
124
+ ctx.save_for_backward(indices)
125
+ assert indices.ndim == 1
126
+ assert values.ndim >= 2
127
+ output = torch.zeros(
128
+ first_axis_dim, *values.shape[1:], device=values.device, dtype=values.dtype
129
+ )
130
+ output[indices] = values
131
+ return output
132
+
133
+ @staticmethod
134
+ def backward(ctx, grad_output: torch.Tensor) -> Tuple[torch.Tensor, None, None]:
135
+ indices, = ctx.saved_tensors
136
+ grad_values = grad_output[indices]
137
+ return grad_values, None, None
138
+
139
+
140
+ index_put_first_axis = IndexPutFirstAxis.apply
141
+
142
+
143
+ def pad_input(inputs: torch.Tensor, indices: torch.Tensor, batch: int, seqlen: int) -> torch.Tensor:
144
+ """Add padding to sequences.
145
+
146
+ Arguments:
147
+ inputs: (total_nnz, ...), where total_nnz = number of tokens in selected in attention_mask.
148
+ indices: (total_nnz), `indices = torch.nonzero(attention_mask.flatten(), as_tuple=False).flatten()`
149
+ batch: int batch_size
150
+ seqlen: int max sequence length
151
+
152
+ Returns:
153
+ inputs: (batch, seqlen, ...)
154
+ """
155
+ output = index_put_first_axis(inputs, indices, batch * seqlen)
156
+ return output.view(batch, seqlen, *inputs.shape[1:])
157
+
158
+
159
+ def rotate_half(x):
160
+ """Rotates half the hidden dims of the input."""
161
+ x1 = x[..., : x.shape[-1] // 2]
162
+ x2 = x[..., x.shape[-1] // 2 :]
163
+ return torch.cat((-x2, x1), dim=-1)
164
+
165
+
166
+ def apply_rotary_pos_emb(q, k, cos, sin):
167
+ """Applies Rotary Position Embedding to the query and key tensors.
168
+
169
+ Args:
170
+ q (`torch.Tensor`): The query tensor.
171
+ k (`torch.Tensor`): The key tensor.
172
+ cos (`torch.Tensor`): The cosine part of the rotary embedding.
173
+ sin (`torch.Tensor`): The sine part of the rotary embedding.
174
+ Returns:
175
+ `tuple(torch.Tensor)` comprising of the query and key tensors rotated using the Rotary Position Embedding.
176
+ """
177
+ cos, sin = cos.to(q.dtype), sin.to(q.dtype)
178
+ q_embed = (q * cos) + (rotate_half(q) * sin)
179
+ k_embed = (k * cos) + (rotate_half(k) * sin)
180
+ return q_embed, k_embed
181
+
182
+
183
+ class RotaryEmbedding(torch.nn.Module):
184
+ def __init__(self, dim, max_position_embeddings=512, base=10000.0, device=None):
185
+ super().__init__()
186
+
187
+ self.dim = dim
188
+ self.max_position_embeddings = max_position_embeddings
189
+ self.base = base
190
+ inv_freq = 1.0 / (self.base ** (torch.arange(0, self.dim, 2).float().to(device) / self.dim))
191
+ self.register_buffer("inv_freq", inv_freq, persistent=False)
192
+
193
+ # Build here to make `torch.jit.trace` work.
194
+ self._set_cos_sin_cache(
195
+ seq_len=max_position_embeddings, device=self.inv_freq.device, dtype=torch.get_default_dtype()
196
+ )
197
+
198
+ def _set_cos_sin_cache(self, seq_len, device, dtype):
199
+ self.max_seq_len_cached = seq_len
200
+ t = torch.arange(self.max_seq_len_cached, device=device, dtype=torch.float32)
201
+
202
+ freqs = torch.einsum("i,j->ij", t, self.inv_freq)
203
+ # Different from paper, but it uses a different permutation in order to obtain the same calculation
204
+ emb = torch.cat((freqs, freqs), dim=-1)
205
+ self.register_buffer("cos_cached", emb.cos().to(dtype), persistent=False)
206
+ self.register_buffer("sin_cached", emb.sin().to(dtype), persistent=False)
207
+
208
+ def forward(self, x, seq_len=None):
209
+ # x: [bs, num_attention_heads, seq_len, head_size]
210
+ if seq_len > self.max_seq_len_cached:
211
+ self._set_cos_sin_cache(seq_len=seq_len, device=x.device, dtype=x.dtype)
212
+
213
+ return (
214
+ self.cos_cached[:seq_len, ...].to(dtype=x.dtype),
215
+ self.sin_cached[:seq_len, ...].to(dtype=x.dtype),
216
+ )
217
+
218
+
219
+ class NTKScalingRotaryEmbedding(RotaryEmbedding):
220
+ """RotaryEmbedding extended with fixed and mixed NTK scaling. https://kexue.fm/archives/9706 """
221
+
222
+ def __init__(self, dim, max_position_embeddings=512, base=10000, device=None, scaling_factor=1.0, mixed_b=None):
223
+ self.scaling_factor = scaling_factor
224
+ self.mixed_b = mixed_b
225
+ super().__init__(dim, max_position_embeddings, base, device)
226
+ max_position_embeddings = max_position_embeddings * self.scaling_factor
227
+ self._set_cos_sin_cache(max_position_embeddings, self.inv_freq.device, torch.get_default_dtype())
228
+
229
+ def _set_cos_sin_cache(self, seq_len, device, dtype):
230
+ self.max_seq_len_cached = seq_len
231
+
232
+ if seq_len > self.max_position_embeddings:
233
+ base = self.base * (self.scaling_factor if self.mixed_b is None else 1)
234
+ inv_freq = 1.0 / (base ** (torch.arange(0, self.dim, 2).float().to(device) / self.dim))
235
+
236
+ if self.mixed_b is None:
237
+ inv_freq = inv_freq / self.scaling_factor ** (2 / self.dim) # (6)
238
+ else:
239
+ a = torch.tensor(self.scaling_factor).log() / (self.dim / 2) ** self.mixed_b # (13)
240
+ lambda_1_m = (a * torch.arange(1, self.dim // 2 + 1).float().to(device) ** self.mixed_b).exp() # (12)
241
+ inv_freq = inv_freq / lambda_1_m # (10)
242
+
243
+ self.register_buffer("inv_freq", inv_freq, persistent=False)
244
+
245
+ t = torch.arange(self.max_seq_len_cached, device=device, dtype=torch.float32)
246
+
247
+ freqs = torch.einsum("i,j->ij", t, self.inv_freq)
248
+ # Different from paper, but it uses a different permutation in order to obtain the same calculation
249
+ emb = torch.cat((freqs, freqs), dim=-1)
250
+ self.register_buffer("cos_cached", emb.cos().to(dtype), persistent=False)
251
+ self.register_buffer("sin_cached", emb.sin().to(dtype), persistent=False)
252
+
253
+
254
+ class RMSNorm(nn.Module):
255
+ def __init__(self, hidden_size, eps=1e-6):
256
+ """
257
+ RMSNorm is equivalent to T5LayerNorm
258
+ """
259
+ super().__init__()
260
+ self.weight = nn.Parameter(torch.ones(hidden_size))
261
+ self.variance_epsilon = eps
262
+
263
+ def forward(self, hidden_states):
264
+ input_dtype = hidden_states.dtype
265
+ hidden_states = hidden_states.to(torch.float32)
266
+ variance = hidden_states.pow(2).mean(-1, keepdim=True)
267
+ hidden_states = hidden_states * torch.rsqrt(variance + self.variance_epsilon)
268
+ return self.weight * hidden_states.to(input_dtype)
269
+
270
+
271
+ LAYER_NORM = {
272
+ 'layer_norm': nn.LayerNorm,
273
+ 'rms_norm': RMSNorm
274
+ }
275
+
276
+
277
+ class NewEmbeddings(nn.Module):
278
+ """
279
+ Embedding and Unpadding.
280
+ """
281
+
282
+ def __init__(self, config: NewConfig):
283
+ super().__init__()
284
+ self.padding_idx = config.pad_token_id
285
+ self.word_embeddings = nn.Embedding(
286
+ config.vocab_size, config.hidden_size, padding_idx=self.padding_idx
287
+ )
288
+
289
+ self.position_embedding_type = config.position_embedding_type
290
+ if self.position_embedding_type == 'absolute':
291
+ self.position_embeddings = nn.Embedding(
292
+ config.max_position_embeddings, config.hidden_size, padding_idx=self.padding_idx
293
+ )
294
+ elif self.position_embedding_type == 'rope':
295
+ self._init_rope(config)
296
+ else:
297
+ raise ValueError
298
+
299
+ self.type_vocab_size = config.type_vocab_size
300
+ if self.type_vocab_size > 0:
301
+ self.token_type_embeddings = nn.Embedding(config.type_vocab_size, config.hidden_size)
302
+
303
+ # self.LayerNorm is not snake-cased to stick with TensorFlow model variable name and be able to load
304
+ # any TensorFlow checkpoint file
305
+ self.LayerNorm = nn.LayerNorm(config.hidden_size, eps=config.layer_norm_eps)
306
+ self.dropout = nn.Dropout(config.hidden_dropout_prob)
307
+ # position_ids is contiguous in memory and excluded when serialized
308
+ self.register_buffer(
309
+ "position_ids", torch.arange(config.max_position_embeddings), persistent=False
310
+ )
311
+
312
+ def _init_rope(self, config):
313
+ kwargs = dict(
314
+ dim=int(config.hidden_size / config.num_attention_heads),
315
+ max_position_embeddings=config.max_position_embeddings,
316
+ base=config.rope_theta
317
+ )
318
+ if config.rope_scaling is None:
319
+ self.rotary_emb = RotaryEmbedding(**kwargs)
320
+ else:
321
+ kwargs.update(scaling_factor=config.rope_scaling["factor"])
322
+ scaling_type = config.rope_scaling["type"]
323
+ if scaling_type == 'ntk':
324
+ kwargs.update(mixed_b=config.rope_scaling.get('mixed_b', None))
325
+ self.rotary_emb = NTKScalingRotaryEmbedding(**kwargs)
326
+ # elif scaling_type == "linear":
327
+ # self.rotary_emb = LinearScalingRotaryEmbedding(**kwargs)
328
+ # elif scaling_type == "dynamic":
329
+ # self.rotary_emb = DynamicNTKScalingRotaryEmbedding(**kwargs)
330
+ else:
331
+ raise ValueError(f"Unknown RoPE scaling type {scaling_type}")
332
+
333
+ def forward(
334
+ self,
335
+ unpad_inputs: bool,
336
+ input_ids: Optional[torch.Tensor] = None,
337
+ attention_mask: Optional[torch.Tensor] = None,
338
+ length: Optional[List[int]] = None,
339
+ token_type_ids: Optional[torch.Tensor] = None,
340
+ position_ids: Optional[torch.Tensor] = None,
341
+ inputs_embeds: Optional[torch.Tensor] = None,
342
+ ) -> Tuple[torch.Tensor, torch.Tensor, Optional[Tuple], Optional[List[int]]]:
343
+ """
344
+ """
345
+ if inputs_embeds is None:
346
+ device, input_shape = input_ids.device, input_ids.shape
347
+ else:
348
+ device, input_shape = inputs_embeds.device, inputs_embeds.shape[:2]
349
+ batch_size, seq_length = input_shape
350
+
351
+ # Set attention_mask if it's None
352
+ if attention_mask is None:
353
+ attention_mask = torch.ones(input_shape, device=device)
354
+ if length is not None:
355
+ for i, l in enumerate(length):
356
+ attention_mask[i, l:] = 0
357
+
358
+ # Set attention_mask_bool for unpadding
359
+ if unpad_inputs:
360
+ attention_mask_bool = attention_mask.bool()
361
+ if length is None:
362
+ length = attention_mask.sum(-1).tolist()
363
+
364
+ # Get word embeddings
365
+ if inputs_embeds is None:
366
+ if unpad_inputs:
367
+ input_ids = input_ids[attention_mask_bool].unsqueeze(0)
368
+ inputs_embeds = self.word_embeddings(input_ids)
369
+ else:
370
+ if unpad_inputs:
371
+ inputs_embeds = inputs_embeds[attention_mask_bool].unsqueeze(0)
372
+ embeddings = inputs_embeds
373
+
374
+ # Set and unpad position_ids
375
+ if position_ids is None:
376
+ if seq_length > self.position_ids.size(0):
377
+ self.register_buffer(
378
+ "position_ids", torch.arange(seq_length, device=embeddings.device), persistent=False
379
+ )
380
+ if unpad_inputs:
381
+ # [1, cumsum_seq_len]
382
+ position_ids = torch.cat([self.position_ids[:l] for l in length]).unsqueeze(0)
383
+ else:
384
+ # [bs, seq_len]
385
+ position_ids = self.position_ids[:seq_length].expand(batch_size, -1)
386
+ elif unpad_inputs:
387
+ position_ids = position_ids[attention_mask_bool].unsqueeze(0) # [1, cumsum_seq_len]
388
+
389
+ # Compute rotary embedding
390
+ if self.position_embedding_type == 'rope':
391
+ rope_cos, rope_sin = self.rotary_emb(inputs_embeds, seq_len=seq_length)
392
+ rope_cos = rope_cos[position_ids].unsqueeze(2) # [bs, seq_len, 1, dim]
393
+ rope_sin = rope_sin[position_ids].unsqueeze(2) # [bs, seq_len, 1, dim]
394
+ rope_embeds = rope_cos, rope_sin
395
+ else:
396
+ rope_embeds = None
397
+
398
+ if self.type_vocab_size > 0:
399
+ if token_type_ids is None:
400
+ token_type_ids = position_ids.mul(0)
401
+ else:
402
+ if self.type_vocab_size < 2:
403
+ token_type_ids.mul_(0)
404
+ if unpad_inputs:
405
+ token_type_ids = token_type_ids[attention_mask_bool].unsqueeze(0)
406
+
407
+ token_type_embeddings = self.token_type_embeddings(token_type_ids)
408
+ embeddings = embeddings + token_type_embeddings
409
+
410
+ # BERT position
411
+ if self.position_embedding_type == "absolute":
412
+ position_embeddings = self.position_embeddings(position_ids)
413
+ embeddings = embeddings + position_embeddings
414
+
415
+ embeddings = self.LayerNorm(embeddings)
416
+ embeddings = self.dropout(embeddings)
417
+
418
+ return embeddings, attention_mask, rope_embeds, length
419
+
420
+
421
+ class NewAttention(nn.Module):
422
+ def __init__(self, config: NewConfig, pack_qkv=None, use_memory_efficient_attention=None):
423
+ super().__init__()
424
+ self.config = config
425
+ if config.hidden_size % config.num_attention_heads != 0 and not hasattr(config, "embedding_size"):
426
+ raise ValueError(
427
+ f"The hidden size ({config.hidden_size}) is not a multiple of the number of attention "
428
+ f"heads ({config.num_attention_heads})"
429
+ )
430
+
431
+ self.hidden_size = config.hidden_size
432
+ self.num_attention_heads = config.num_attention_heads
433
+ self.attention_head_size = int(config.hidden_size / config.num_attention_heads)
434
+ self.all_head_size = self.num_attention_heads * self.attention_head_size
435
+
436
+ if pack_qkv is None:
437
+ pack_qkv = config.pack_qkv
438
+ self.pack_qkv = pack_qkv
439
+
440
+ if self.pack_qkv:
441
+ self.qkv_proj = nn.Linear(config.hidden_size, self.all_head_size * 3, bias=True)
442
+ else:
443
+ self.q_proj = nn.Linear(config.hidden_size, self.all_head_size, bias=True)
444
+ self.k_proj = nn.Linear(config.hidden_size, self.all_head_size, bias=True)
445
+ self.v_proj = nn.Linear(config.hidden_size, self.all_head_size, bias=True)
446
+
447
+ self.dropout = nn.Dropout(config.attention_probs_dropout_prob)
448
+ self.o_proj = nn.Linear(config.hidden_size, config.hidden_size, bias=True)
449
+
450
+ if use_memory_efficient_attention is None:
451
+ use_memory_efficient_attention = self.config.use_memory_efficient_attention
452
+ self.use_memory_efficient_attention = use_memory_efficient_attention
453
+ self.memory_efficient_attention = None if xops is None else xops.memory_efficient_attention
454
+ if self.use_memory_efficient_attention:
455
+ assert self.memory_efficient_attention is not None, 'please install xformers'
456
+
457
+ def forward(
458
+ self,
459
+ hidden_states: torch.Tensor,
460
+ attention_bias: torch.FloatTensor,
461
+ rope_embeds: Optional[Tuple[torch.FloatTensor, torch.FloatTensor]] = None,
462
+ padding_inputs: Optional[Tuple] = None, # indices, batch, seqlen
463
+ attention_scale: Optional[torch.FloatTensor] = None,
464
+ head_mask: Optional[torch.FloatTensor] = None,
465
+ output_attentions: Optional[bool] = False,
466
+ qkv_inputs: Optional[Tuple] = None, # For RetroMAE
467
+ ) -> Tuple[torch.Tensor, ...]:
468
+ shape_hd = (self.num_attention_heads, self.attention_head_size)
469
+ # qkv
470
+ if self.pack_qkv and qkv_inputs is None:
471
+ qkv_pack = self.qkv_proj(hidden_states).split(self.all_head_size, dim=-1)
472
+ else:
473
+ if qkv_inputs is None:
474
+ qkv_inputs = (hidden_states, hidden_states, hidden_states)
475
+ qkv_pack = [
476
+ getattr(self, n + '_proj')(s) for s, n in zip(qkv_inputs, 'qkv')
477
+ ]
478
+ query_states, key_states, value_states = [t.view(t.shape[:-1] + shape_hd) for t in qkv_pack]
479
+
480
+ if self.config.position_embedding_type == 'rope':
481
+ query_states, key_states = apply_rotary_pos_emb(query_states, key_states, *rope_embeds)
482
+
483
+ dtype = query_states.dtype
484
+
485
+ if self.config.logn_attention_scale and attention_scale is not None:
486
+ # https://kexue.fm/archives/8823
487
+ query_states = query_states * attention_scale.to(dtype)
488
+
489
+ if padding_inputs is not None:
490
+ query_states = pad_input(query_states.squeeze(), *padding_inputs)
491
+ key_states = pad_input(key_states.squeeze(), *padding_inputs)
492
+ value_states = pad_input(value_states.squeeze(), *padding_inputs)
493
+
494
+ if self.use_memory_efficient_attention:
495
+ assert self.memory_efficient_attention is not None, "xformers is not loaded"
496
+ assert output_attentions is False, "memory_efficient_attention do not output attentions"
497
+ assert head_mask is None, "Not support yet"
498
+ attention_probs = None
499
+ if torch.is_tensor(attention_bias):
500
+ attention_bias = attention_bias.to(dtype)
501
+ context_layer = self.memory_efficient_attention(
502
+ query_states,
503
+ key_states,
504
+ value_states,
505
+ attn_bias=attention_bias,
506
+ p=self.dropout.p
507
+ )
508
+ else:
509
+ if output_attentions and isinstance(self, NewSdpaAttention):
510
+ raise RuntimeError("SDPA do not output attentions")
511
+ context_layer, attention_probs = self._attention(
512
+ query_states, key_states, value_states, attention_bias, head_mask
513
+ )
514
+
515
+ if padding_inputs is not None:
516
+ context_layer = unpad_input(context_layer, indices=padding_inputs[0])
517
+
518
+ new_context_layer_shape = context_layer.size()[:-2] + (self.all_head_size,)
519
+ context_layer = context_layer.view(new_context_layer_shape)
520
+
521
+ # output proj
522
+ attn_output = self.o_proj(context_layer)
523
+
524
+ # add attentions if we output them
525
+ outputs = (attn_output, attention_probs) if output_attentions else (attn_output,)
526
+ return outputs
527
+
528
+ def _attention(self, query_states, key_states, value_states, attention_bias, head_mask):
529
+ """
530
+ Args:
531
+ q/k/v: (B, L, n_head, head_dim),
532
+ Returns:
533
+ attn_output: (B L, n_head, head_dim)
534
+ """
535
+ query_states = query_states.transpose(1, 2)
536
+ key_states = key_states.transpose(1, 2)
537
+ value_states = value_states.transpose(1, 2)
538
+ # Take the dot product between "query" and "key" to get the raw attention scores.
539
+ attention_scores = torch.matmul(query_states, key_states.transpose(-1, -2))
540
+
541
+ attention_scores = attention_scores / math.sqrt(self.attention_head_size)
542
+ if attention_bias is not None:
543
+ # Apply the attention mask is (precomputed for all layers in BertModel forward() function)
544
+ attention_scores = attention_scores + attention_bias
545
+
546
+ # Normalize the attention scores to probabilities.
547
+ attention_probs = nn.functional.softmax(attention_scores, dim=-1)
548
+
549
+ # This is actually dropping out entire tokens to attend to, which might
550
+ # seem a bit unusual, but is taken from the original Transformer paper.
551
+ if self.dropout.p > 0:
552
+ attention_probs = self.dropout(attention_probs)
553
+
554
+ # Mask heads if we want to
555
+ if head_mask is not None:
556
+ attention_probs = attention_probs * head_mask
557
+
558
+ context_layer = torch.matmul(attention_probs, value_states)
559
+
560
+ context_layer = context_layer.permute(0, 2, 1, 3).contiguous()
561
+ return context_layer, attention_probs
562
+
563
+
564
+ class NewSdpaAttention(NewAttention):
565
+ """
566
+ New attention module using torch.nn.functional.scaled_dot_product_attention. This module inherits from
567
+ `NewAttention` as the weights of the module stays untouched. The only changes are on the forward pass to adapt to
568
+ SDPA API.
569
+ """
570
+ def __init__(self, config: NewConfig, **kwargs):
571
+ super().__init__(config, **kwargs)
572
+ # torch.backends.cuda.enable_mem_efficient_sdp(False)
573
+ # logger.warning(
574
+ # "Disable memory efficient attention kernel for `NewSdpaAttention`, you can set "
575
+ # "`use_memory_efficient_attention=True` if it expected to use."
576
+ # )
577
+
578
+ def _attention(self, query_states, key_states, value_states, attention_bias, head_mask):
579
+ attn_output = torch.nn.functional.scaled_dot_product_attention(
580
+ query_states.transpose(1, 2),
581
+ key_states.transpose(1, 2),
582
+ value_states.transpose(1, 2),
583
+ attn_mask=attention_bias,
584
+ dropout_p=self.dropout.p if self.training else 0.0,
585
+ )
586
+ attn_output = attn_output.permute(0, 2, 1, 3).contiguous()
587
+ return attn_output, None
588
+
589
+
590
+ NEW_ATTENTION_CLASSES = {
591
+ "eager": NewAttention,
592
+ # "flash_attention_2": , # TODO
593
+ "sdpa": NewSdpaAttention,
594
+ }
595
+
596
+
597
+ class NewGatedMLP(nn.Module):
598
+ """
599
+ GLU Variants Improve Transformer.
600
+ """
601
+
602
+ def __init__(self, config: NewConfig):
603
+ super().__init__()
604
+ self.intermediate_size = config.intermediate_size
605
+ self.up_gate_proj = nn.Linear(config.hidden_size, self.intermediate_size * 2, bias=False)
606
+ self.down_proj = nn.Linear(self.intermediate_size, config.hidden_size, bias=True)
607
+ self.act_fn = ACT2FN[config.hidden_act]
608
+ if config.hidden_dropout_prob > 0:
609
+ self.hidden_dropout = nn.Dropout(config.hidden_dropout_prob)
610
+ else:
611
+ self.hidden_dropout = None
612
+
613
+ def forward(self, hidden_states):
614
+ up_gate = self.up_gate_proj(hidden_states)
615
+ up_states, gate = torch.split(up_gate, self.intermediate_size, dim=-1)
616
+ gate = self.act_fn(gate)
617
+ gated_states = gate * up_states
618
+ if self.hidden_dropout is not None:
619
+ gated_states = self.hidden_dropout(gated_states)
620
+ down_states = self.down_proj(gated_states)
621
+ return down_states
622
+
623
+
624
+ class NewLayer(nn.Module):
625
+ def __init__(
626
+ self,
627
+ config: NewConfig,
628
+ pack_qkv=None,
629
+ use_memory_efficient_attention=None,
630
+ attn_implementation=None
631
+ ):
632
+ super().__init__()
633
+ if attn_implementation is None:
634
+ attn_implementation = config._attn_implementation
635
+ if use_memory_efficient_attention is None:
636
+ use_memory_efficient_attention = config.use_memory_efficient_attention
637
+ if use_memory_efficient_attention:
638
+ if attn_implementation != 'eager':
639
+ logger.warning_once(f"Override {attn_implementation=} to 'eager' as {use_memory_efficient_attention=}")
640
+ attn_implementation = 'eager' # Since it will be SDPA by default for torch>=2.1.1
641
+ self.attention = NEW_ATTENTION_CLASSES[attn_implementation](
642
+ config, pack_qkv=pack_qkv, use_memory_efficient_attention=use_memory_efficient_attention
643
+ )
644
+ self.mlp = NewGatedMLP(config)
645
+
646
+ ln_class = LAYER_NORM[config.layer_norm_type]
647
+ self.attn_ln = ln_class(config.hidden_size, eps=config.layer_norm_eps)
648
+ self.mlp_ln = ln_class(config.hidden_size, eps=config.layer_norm_eps)
649
+
650
+ if config.hidden_dropout_prob > 0:
651
+ self.hidden_dropout = nn.Dropout(config.hidden_dropout_prob)
652
+ else:
653
+ self.hidden_dropout = None
654
+
655
+ def forward(
656
+ self,
657
+ hidden_states: torch.Tensor,
658
+ attention_bias: torch.FloatTensor,
659
+ rope_embeds: Optional[Tuple[torch.FloatTensor, torch.FloatTensor]] = None,
660
+ padding_inputs: Optional[Tuple] = None, # indices, batch, seqlen
661
+ attention_scale: Optional[torch.FloatTensor] = None,
662
+ subset_indices: Optional[torch.LongTensor] = None,
663
+ head_mask: Optional[torch.FloatTensor] = None,
664
+ output_attentions: Optional[bool] = False,
665
+ qkv_inputs: Optional[Tuple] = None, # For RetroMAE
666
+ ) -> Tuple[torch.Tensor, ...]:
667
+ # Multi head self attention
668
+ residual = hidden_states if qkv_inputs is None else qkv_inputs[0]
669
+ attention_outputs = self.attention(
670
+ hidden_states,
671
+ attention_bias,
672
+ rope_embeds,
673
+ padding_inputs,
674
+ attention_scale,
675
+ head_mask,
676
+ output_attentions=output_attentions,
677
+ qkv_inputs=qkv_inputs,
678
+ )
679
+ hidden_states = attention_outputs[0]
680
+ if self.hidden_dropout is not None:
681
+ hidden_states = self.hidden_dropout(hidden_states)
682
+ hidden_states = residual + hidden_states
683
+
684
+ # In pretraining, after the attention of last layer, we only need the masked tokens.
685
+ if subset_indices is not None:
686
+ hidden_states = hidden_states[subset_indices]
687
+
688
+ hidden_states = self.attn_ln(hidden_states)
689
+
690
+ # Fully Connected
691
+ residual = hidden_states
692
+ hidden_states = self.mlp(hidden_states)
693
+ if self.hidden_dropout is not None:
694
+ hidden_states = self.hidden_dropout(hidden_states)
695
+ hidden_states = residual + hidden_states
696
+ hidden_states = self.mlp_ln(hidden_states)
697
+
698
+ # add self attentions if we output attention weights
699
+ outputs = (hidden_states,) + attention_outputs[1:]
700
+ return outputs
701
+
702
+
703
+ class NewEncoder(nn.Module):
704
+ def __init__(self, config):
705
+ super().__init__()
706
+ self.config = config
707
+ self.layer = nn.ModuleList([NewLayer(config) for _ in range(config.num_hidden_layers)])
708
+ self.gradient_checkpointing = False
709
+
710
+ def forward(
711
+ self,
712
+ hidden_states: torch.Tensor,
713
+ attention_bias: Optional[torch.FloatTensor] = None,
714
+ rope_embeds: Optional[Tuple[torch.FloatTensor, torch.FloatTensor]] = None,
715
+ padding_inputs: Optional[Tuple] = None, # indices, batch, seqlen
716
+ attention_scale: Optional[torch.FloatTensor] = None,
717
+ subset_indices: Optional[torch.LongTensor] = None,
718
+ head_mask: Optional[torch.FloatTensor] = None,
719
+ output_attentions: Optional[bool] = False,
720
+ output_hidden_states: Optional[bool] = False,
721
+ return_dict: Optional[bool] = True,
722
+ ) -> Union[Tuple[torch.Tensor], BaseModelOutput]:
723
+ all_hidden_states = () if output_hidden_states else None
724
+ all_self_attentions = () if output_attentions else None
725
+
726
+ for i, layer_module in enumerate(self.layer):
727
+ if output_hidden_states:
728
+ all_hidden_states = all_hidden_states + (hidden_states,)
729
+
730
+ if i >= len(self.layer) - 1:
731
+ layer_subset_indices = subset_indices
732
+ else:
733
+ layer_subset_indices = None
734
+
735
+ layer_head_mask = head_mask[i] if head_mask is not None else None
736
+
737
+ if self.gradient_checkpointing and self.training:
738
+ layer_outputs = self._gradient_checkpointing_func(
739
+ layer_module.__call__,
740
+ hidden_states,
741
+ attention_bias,
742
+ rope_embeds,
743
+ padding_inputs,
744
+ attention_scale,
745
+ layer_subset_indices,
746
+ layer_head_mask,
747
+ )
748
+ else:
749
+ layer_outputs = layer_module(
750
+ hidden_states,
751
+ attention_bias,
752
+ rope_embeds,
753
+ padding_inputs,
754
+ attention_scale,
755
+ layer_subset_indices,
756
+ layer_head_mask,
757
+ output_attentions,
758
+ )
759
+
760
+ hidden_states = layer_outputs[0]
761
+ if output_attentions:
762
+ all_self_attentions = all_self_attentions + (layer_outputs[1],)
763
+
764
+ if output_hidden_states:
765
+ all_hidden_states = all_hidden_states + (hidden_states,)
766
+
767
+ if not return_dict:
768
+ return tuple(
769
+ v
770
+ for v in [
771
+ hidden_states,
772
+ all_hidden_states,
773
+ all_self_attentions,
774
+ ]
775
+ if v is not None
776
+ )
777
+ return BaseModelOutput(
778
+ last_hidden_state=hidden_states,
779
+ hidden_states=all_hidden_states,
780
+ attentions=all_self_attentions,
781
+ )
782
+
783
+
784
+ # Copied from transformers.models.bert.modeling_bert.BertPooler with Bert->New
785
+ class NewPooler(nn.Module):
786
+ def __init__(self, config):
787
+ super().__init__()
788
+ self.dense = nn.Linear(config.hidden_size, config.hidden_size)
789
+ self.activation = nn.Tanh()
790
+
791
+ def forward(self, hidden_states: torch.Tensor) -> torch.Tensor:
792
+ # We "pool" the model by simply taking the hidden state corresponding
793
+ # to the first token.
794
+ first_token_tensor = hidden_states[:, 0]
795
+ pooled_output = self.dense(first_token_tensor)
796
+ pooled_output = self.activation(pooled_output)
797
+ return pooled_output
798
+
799
+
800
+ class NewPreTrainedModel(PreTrainedModel):
801
+ """
802
+ An abstract class to handle weights initialization and a simple interface for downloading and loading pretrained
803
+ models.
804
+ """
805
+
806
+ config_class = NewConfig
807
+ base_model_prefix = "new"
808
+ supports_gradient_checkpointing = True
809
+ _supports_sdpa = True
810
+
811
+ def _init_weights(self, module):
812
+ """Initialize the weights"""
813
+ if isinstance(module, nn.Linear):
814
+ # Slightly different from the TF version which uses truncated_normal for initialization
815
+ # cf https://github.com/pytorch/pytorch/pull/5617
816
+ module.weight.data.normal_(mean=0.0, std=self.config.initializer_range)
817
+ if module.bias is not None:
818
+ module.bias.data.zero_()
819
+ elif isinstance(module, nn.Embedding):
820
+ module.weight.data.normal_(mean=0.0, std=self.config.initializer_range)
821
+ if module.padding_idx is not None:
822
+ module.weight.data[module.padding_idx].zero_()
823
+ elif isinstance(module, nn.LayerNorm):
824
+ module.bias.data.zero_()
825
+ module.weight.data.fill_(1.0)
826
+
827
+
828
+ class NewModel(NewPreTrainedModel):
829
+ """
830
+ The bare New Model transformer outputting raw hidden-states without any specific head on top.
831
+ """
832
+
833
+ def __init__(self, config: NewConfig, add_pooling_layer=False):
834
+ super().__init__(config)
835
+ self.config = config
836
+
837
+ self.embeddings = NewEmbeddings(config)
838
+ self.encoder = NewEncoder(config)
839
+
840
+ self.pooler = NewPooler(config) if add_pooling_layer else None
841
+
842
+ # Initialize weights and apply final processing
843
+ self.post_init()
844
+
845
+ def get_input_embeddings(self):
846
+ return self.embeddings.word_embeddings
847
+
848
+ def set_input_embeddings(self, value):
849
+ self.embeddings.word_embeddings = value
850
+
851
+ def forward(
852
+ self,
853
+ input_ids: Optional[torch.Tensor] = None,
854
+ attention_mask: Optional[torch.Tensor] = None,
855
+ length: Optional[List[int]] = None,
856
+ subset_indices: Optional[torch.LongTensor] = None,
857
+ token_type_ids: Optional[torch.Tensor] = None,
858
+ position_ids: Optional[torch.Tensor] = None,
859
+ head_mask: Optional[torch.Tensor] = None,
860
+ inputs_embeds: Optional[torch.Tensor] = None,
861
+ output_attentions: Optional[bool] = None,
862
+ output_hidden_states: Optional[bool] = None,
863
+ return_dict: Optional[bool] = None,
864
+ unpad_inputs: Optional[bool] = None,
865
+ ) -> Union[Tuple[torch.Tensor], BaseModelOutputWithPooling]:
866
+ r"""
867
+ length (`list` of length `batch_size`, *optional*):
868
+ If is `None`, return padded `last_hidden_state`.
869
+ subset_indices ():
870
+ pass
871
+ unpad_inputs (`bool`, *optional*):
872
+ pass
873
+ """
874
+ output_attentions = output_attentions if output_attentions is not None else self.config.output_attentions
875
+ output_hidden_states = (
876
+ output_hidden_states if output_hidden_states is not None else self.config.output_hidden_states
877
+ )
878
+ return_dict = return_dict if return_dict is not None else self.config.use_return_dict
879
+ unpad_inputs = unpad_inputs if unpad_inputs is not None else self.config.unpad_inputs
880
+ output_padded = length is None
881
+
882
+ if input_ids is not None and inputs_embeds is not None:
883
+ raise ValueError("You cannot specify both input_ids and inputs_embeds at the same time")
884
+ elif input_ids is not None:
885
+ self.warn_if_padding_and_no_attention_mask(input_ids, attention_mask)
886
+ input_shape = input_ids.size()
887
+ elif inputs_embeds is not None:
888
+ input_shape = inputs_embeds.size()[:-1]
889
+ else:
890
+ raise ValueError("You have to specify either input_ids or inputs_embeds")
891
+
892
+ # TODO: not used
893
+ # # Prepare head mask if needed
894
+ # # 1.0 in head_mask indicate we keep the head
895
+ # # attention_probs has shape bsz x n_heads x N x N
896
+ # # input head_mask has shape [num_heads] or [num_hidden_layers x num_heads]
897
+ # # and head_mask is converted to shape [num_hidden_layers x batch x num_heads x seq_length x seq_length]
898
+ # head_mask = self.get_head_mask(head_mask, self.config.num_hidden_layers)
899
+
900
+ # Get embeddings, may unpad them
901
+ (embedding_output, attention_mask, rope_embeds, length) = self.embeddings(
902
+ unpad_inputs,
903
+ input_ids=input_ids,
904
+ attention_mask=attention_mask,
905
+ length=length,
906
+ token_type_ids=token_type_ids,
907
+ position_ids=position_ids,
908
+ inputs_embeds=inputs_embeds
909
+ )
910
+
911
+ batch_size, seq_length = input_shape
912
+ if unpad_inputs and self.config.use_memory_efficient_attention:
913
+ attention_bias = xops.fmha.attn_bias.BlockDiagonalMask.from_seqlens(length)
914
+ else:
915
+ # We can provide a self-attention mask of dimensions [batch_size, from_seq_length, to_seq_length]
916
+ # ourselves in which case we just need to make it broadcastable to all heads.
917
+ attention_bias = self.get_extended_attention_mask(attention_mask, input_shape)
918
+ if self.config.use_memory_efficient_attention:
919
+ # Invalid shape for attention bias: torch.Size([48, 1, 1, 512]) (expected (48, 12, 512, 512))
920
+ attention_bias = attention_bias.expand(-1, self.config.num_attention_heads, seq_length, -1)
921
+
922
+ padding_inputs = None
923
+ if unpad_inputs and (output_padded or not self.config.use_memory_efficient_attention):
924
+ indices = torch.nonzero(attention_mask.flatten(), as_tuple=False).flatten()
925
+ if not self.config.use_memory_efficient_attention:
926
+ padding_inputs = (indices, *input_shape)
927
+
928
+ attention_scale = None
929
+ if self.config.logn_attention_scale:
930
+ logger.warning_once("TODO: logn_attention_scale")
931
+ # # attention scale log_512(input_len)
932
+ # attention_scale = attention_mask.sum(1).log() / torch.tensor(self.config.max_position_embeddings).log()
933
+ # # inference-time logn scale need clip 1
934
+ # if self.config.logn_attention_clip1:
935
+ # attention_scale.clip_(1)
936
+ # attention_scale = attention_scale[:, None, None, None]
937
+ # else:
938
+ # attention_scale = None
939
+
940
+ encoder_outputs = self.encoder(
941
+ embedding_output,
942
+ attention_bias=attention_bias,
943
+ rope_embeds=rope_embeds,
944
+ padding_inputs=padding_inputs,
945
+ attention_scale=attention_scale,
946
+ subset_indices=subset_indices,
947
+ head_mask=head_mask,
948
+ output_attentions=output_attentions,
949
+ output_hidden_states=output_hidden_states,
950
+ return_dict=return_dict,
951
+ )
952
+ sequence_output = encoder_outputs[0]
953
+ if unpad_inputs and output_padded:
954
+ sequence_output = pad_input(
955
+ sequence_output.squeeze(), indices, batch_size, seq_length
956
+ )
957
+
958
+ pooled_output = self.pooler(sequence_output) if self.pooler is not None else None
959
+
960
+ if not return_dict:
961
+ return (sequence_output, pooled_output) + encoder_outputs[1:]
962
+
963
+ return BaseModelOutputWithPooling(
964
+ last_hidden_state=sequence_output,
965
+ pooler_output=pooled_output,
966
+ hidden_states=encoder_outputs.hidden_states,
967
+ attentions=encoder_outputs.attentions,
968
+ )
969
+
970
+
971
+ class NewLMPredictionHead(nn.Module):
972
+ def __init__(self, config):
973
+ super().__init__()
974
+ self.dense = nn.Linear(config.hidden_size, config.hidden_size)
975
+ self.transform_act_fn = ACT2FN[config.hidden_act]
976
+ self.norm = nn.LayerNorm(config.hidden_size, eps=config.layer_norm_eps)
977
+
978
+ # The output weights are the same as the input embeddings, but there is
979
+ # an output-only bias for each token.
980
+ self.decoder = nn.Linear(config.hidden_size, config.vocab_size)
981
+
982
+ def forward(self, hidden_states):
983
+ hidden_states = self.dense(hidden_states)
984
+ hidden_states = self.transform_act_fn(hidden_states)
985
+ hidden_states = self.norm(hidden_states)
986
+ hidden_states = self.decoder(hidden_states)
987
+ return hidden_states
988
+
989
+
990
+ class NewForMaskedLM(NewPreTrainedModel):
991
+ _tied_weights_keys = ["lm_head.decoder.bias", "lm_head.decoder.weight"]
992
+
993
+ def __init__(self, config: NewConfig):
994
+ super().__init__(config)
995
+ self.new = NewModel(config, add_pooling_layer=False)
996
+ self.lm_head = NewLMPredictionHead(config)
997
+ self.loss_fct = nn.CrossEntropyLoss()
998
+
999
+ # Initialize weights and apply final processing
1000
+ self.post_init()
1001
+
1002
+ def get_output_embeddings(self):
1003
+ return self.lm_head.decoder
1004
+
1005
+ def set_output_embeddings(self, new_embeddings):
1006
+ self.lm_head.decoder = new_embeddings
1007
+
1008
+ def forward(
1009
+ self,
1010
+ input_ids: Optional[torch.Tensor] = None,
1011
+ attention_mask: Optional[torch.Tensor] = None,
1012
+ token_type_ids: Optional[torch.Tensor] = None,
1013
+ position_ids: Optional[torch.Tensor] = None,
1014
+ head_mask: Optional[torch.Tensor] = None,
1015
+ inputs_embeds: Optional[torch.Tensor] = None,
1016
+ labels: Optional[torch.Tensor] = None,
1017
+ output_attentions: Optional[bool] = None,
1018
+ output_hidden_states: Optional[bool] = None,
1019
+ return_dict: Optional[bool] = None,
1020
+ unpad_inputs: Optional[bool] = None,
1021
+ ) -> Union[Tuple[torch.Tensor], MaskedLMOutput]:
1022
+ r"""
1023
+ labels (`torch.LongTensor` of shape `(batch_size, sequence_length)`, *optional*):
1024
+ Labels for computing the masked language modeling loss. Indices should be in `[-100, 0, ...,
1025
+ config.vocab_size]` (see `input_ids` docstring) Tokens with indices set to `-100` are ignored (masked), the
1026
+ loss is only computed for the tokens with labels in `[0, ..., config.vocab_size]`
1027
+ """
1028
+
1029
+ return_dict = return_dict if return_dict is not None else self.config.use_return_dict
1030
+
1031
+ if labels is None or not self.new.config.unpad_inputs:
1032
+ length = None
1033
+ subset_indices = None
1034
+ else:
1035
+ length = attention_mask.sum(-1).tolist()
1036
+ labels = labels[attention_mask.bool()].unsqueeze(0)
1037
+ subset_indices = labels > -100
1038
+
1039
+ outputs = self.new(
1040
+ input_ids,
1041
+ attention_mask=attention_mask,
1042
+ length=length,
1043
+ subset_indices=subset_indices,
1044
+ token_type_ids=token_type_ids,
1045
+ position_ids=position_ids,
1046
+ head_mask=head_mask,
1047
+ inputs_embeds=inputs_embeds,
1048
+ output_attentions=output_attentions,
1049
+ output_hidden_states=output_hidden_states,
1050
+ return_dict=return_dict,
1051
+ unpad_inputs=unpad_inputs,
1052
+ )
1053
+
1054
+ sequence_output = outputs[0]
1055
+ prediction_scores = self.lm_head(sequence_output)
1056
+
1057
+ masked_lm_loss = None
1058
+ if labels is not None:
1059
+ if subset_indices is None:
1060
+ mask = attention_mask.bool()
1061
+ prediction_scores = prediction_scores[mask]
1062
+ labels = labels[mask]
1063
+ else:
1064
+ labels = labels[subset_indices]
1065
+ masked_lm_loss = self.loss_fct(prediction_scores, labels)
1066
+
1067
+ if not return_dict:
1068
+ output = (prediction_scores,) + outputs[2:]
1069
+ return ((masked_lm_loss,) + output) if masked_lm_loss is not None else output
1070
+
1071
+ return MaskedLMOutput(
1072
+ loss=masked_lm_loss,
1073
+ logits=prediction_scores,
1074
+ hidden_states=outputs.hidden_states,
1075
+ attentions=outputs.attentions,
1076
+ )
1077
+
1078
+
1079
+ class NewForSequenceClassification(NewPreTrainedModel):
1080
+ def __init__(self, config):
1081
+ super().__init__(config)
1082
+ self.num_labels = config.num_labels
1083
+ self.config = config
1084
+
1085
+ self.new = NewModel(config, add_pooling_layer=True)
1086
+ classifier_dropout = (
1087
+ config.classifier_dropout if config.classifier_dropout is not None else config.hidden_dropout_prob
1088
+ )
1089
+ self.dropout = nn.Dropout(classifier_dropout)
1090
+ self.classifier = nn.Linear(config.hidden_size, config.num_labels)
1091
+
1092
+ # Initialize weights and apply final processing
1093
+ self.post_init()
1094
+
1095
+ def forward(
1096
+ self,
1097
+ input_ids: Optional[torch.Tensor] = None,
1098
+ attention_mask: Optional[torch.Tensor] = None,
1099
+ token_type_ids: Optional[torch.Tensor] = None,
1100
+ position_ids: Optional[torch.Tensor] = None,
1101
+ head_mask: Optional[torch.Tensor] = None,
1102
+ inputs_embeds: Optional[torch.Tensor] = None,
1103
+ labels: Optional[torch.Tensor] = None,
1104
+ output_attentions: Optional[bool] = None,
1105
+ output_hidden_states: Optional[bool] = None,
1106
+ return_dict: Optional[bool] = None,
1107
+ unpad_inputs: Optional[bool] = None,
1108
+ ) -> Union[Tuple[torch.Tensor], SequenceClassifierOutput]:
1109
+ r"""
1110
+ labels (`torch.LongTensor` of shape `(batch_size,)`, *optional*):
1111
+ Labels for computing the sequence classification/regression loss. Indices should be in `[0, ...,
1112
+ config.num_labels - 1]`. If `config.num_labels == 1` a regression loss is computed (Mean-Square loss), If
1113
+ `config.num_labels > 1` a classification loss is computed (Cross-Entropy).
1114
+ """
1115
+ return_dict = return_dict if return_dict is not None else self.config.use_return_dict
1116
+
1117
+ outputs = self.new(
1118
+ input_ids,
1119
+ attention_mask=attention_mask,
1120
+ token_type_ids=token_type_ids,
1121
+ position_ids=position_ids,
1122
+ head_mask=head_mask,
1123
+ inputs_embeds=inputs_embeds,
1124
+ output_attentions=output_attentions,
1125
+ output_hidden_states=output_hidden_states,
1126
+ return_dict=return_dict,
1127
+ unpad_inputs=unpad_inputs,
1128
+ )
1129
+
1130
+ pooled_output = outputs[1]
1131
+
1132
+ pooled_output = self.dropout(pooled_output)
1133
+ logits = self.classifier(pooled_output)
1134
+
1135
+ loss = None
1136
+ if labels is not None:
1137
+ if self.config.problem_type is None:
1138
+ if self.num_labels == 1:
1139
+ self.config.problem_type = "regression"
1140
+ elif self.num_labels > 1 and (labels.dtype == torch.long or labels.dtype == torch.int):
1141
+ self.config.problem_type = "single_label_classification"
1142
+ else:
1143
+ self.config.problem_type = "multi_label_classification"
1144
+
1145
+ if self.config.problem_type == "regression":
1146
+ loss_fct = nn.MSELoss()
1147
+ if self.num_labels == 1:
1148
+ loss = loss_fct(logits.squeeze(), labels.squeeze())
1149
+ else:
1150
+ loss = loss_fct(logits, labels)
1151
+ elif self.config.problem_type == "single_label_classification":
1152
+ loss_fct = nn.CrossEntropyLoss()
1153
+ loss = loss_fct(logits.view(-1, self.num_labels), labels.view(-1))
1154
+ elif self.config.problem_type == "multi_label_classification":
1155
+ loss_fct = nn.BCEWithLogitsLoss()
1156
+ loss = loss_fct(logits, labels)
1157
+
1158
+ if not return_dict:
1159
+ output = (logits,) + outputs[2:]
1160
+ return ((loss,) + output) if loss is not None else output
1161
+
1162
+ return SequenceClassifierOutput(
1163
+ loss=loss,
1164
+ logits=logits,
1165
+ hidden_states=outputs.hidden_states,
1166
+ attentions=outputs.attentions,
1167
+ )
1168
+
1169
+
1170
+ class NewForMultipleChoice(NewPreTrainedModel):
1171
+ def __init__(self, config):
1172
+ super().__init__(config)
1173
+
1174
+ self.new = NewModel(config, add_pooling_layer=True)
1175
+ classifier_dropout = (
1176
+ config.classifier_dropout if config.classifier_dropout is not None else config.hidden_dropout_prob
1177
+ )
1178
+ self.dropout = nn.Dropout(classifier_dropout)
1179
+ self.classifier = nn.Linear(config.hidden_size, 1)
1180
+
1181
+ # Initialize weights and apply final processing
1182
+ self.post_init()
1183
+
1184
+ def forward(
1185
+ self,
1186
+ input_ids: Optional[torch.Tensor] = None,
1187
+ attention_mask: Optional[torch.Tensor] = None,
1188
+ token_type_ids: Optional[torch.Tensor] = None,
1189
+ position_ids: Optional[torch.Tensor] = None,
1190
+ head_mask: Optional[torch.Tensor] = None,
1191
+ inputs_embeds: Optional[torch.Tensor] = None,
1192
+ labels: Optional[torch.Tensor] = None,
1193
+ output_attentions: Optional[bool] = None,
1194
+ output_hidden_states: Optional[bool] = None,
1195
+ return_dict: Optional[bool] = None,
1196
+ unpad_inputs: Optional[bool] = None,
1197
+ ) -> Union[Tuple[torch.Tensor], MultipleChoiceModelOutput]:
1198
+ r"""
1199
+ labels (`torch.LongTensor` of shape `(batch_size,)`, *optional*):
1200
+ Labels for computing the multiple choice classification loss. Indices should be in `[0, ...,
1201
+ num_choices-1]` where `num_choices` is the size of the second dimension of the input tensors. (See
1202
+ `input_ids` above)
1203
+ """
1204
+ return_dict = return_dict if return_dict is not None else self.config.use_return_dict
1205
+ num_choices = input_ids.shape[1] if input_ids is not None else inputs_embeds.shape[1]
1206
+
1207
+ input_ids = input_ids.view(-1, input_ids.size(-1)) if input_ids is not None else None
1208
+ attention_mask = attention_mask.view(-1, attention_mask.size(-1)) if attention_mask is not None else None
1209
+ token_type_ids = token_type_ids.view(-1, token_type_ids.size(-1)) if token_type_ids is not None else None
1210
+ position_ids = position_ids.view(-1, position_ids.size(-1)) if position_ids is not None else None
1211
+ inputs_embeds = (
1212
+ inputs_embeds.view(-1, inputs_embeds.size(-2), inputs_embeds.size(-1))
1213
+ if inputs_embeds is not None
1214
+ else None
1215
+ )
1216
+
1217
+ outputs = self.new(
1218
+ input_ids,
1219
+ attention_mask=attention_mask,
1220
+ token_type_ids=token_type_ids,
1221
+ position_ids=position_ids,
1222
+ head_mask=head_mask,
1223
+ inputs_embeds=inputs_embeds,
1224
+ output_attentions=output_attentions,
1225
+ output_hidden_states=output_hidden_states,
1226
+ return_dict=return_dict,
1227
+ unpad_inputs=unpad_inputs,
1228
+ )
1229
+
1230
+ pooled_output = outputs[1]
1231
+
1232
+ pooled_output = self.dropout(pooled_output)
1233
+ logits = self.classifier(pooled_output)
1234
+ reshaped_logits = logits.view(-1, num_choices)
1235
+
1236
+ loss = None
1237
+ if labels is not None:
1238
+ loss_fct = nn.CrossEntropyLoss()
1239
+ loss = loss_fct(reshaped_logits, labels)
1240
+
1241
+ if not return_dict:
1242
+ output = (reshaped_logits,) + outputs[2:]
1243
+ return ((loss,) + output) if loss is not None else output
1244
+
1245
+ return MultipleChoiceModelOutput(
1246
+ loss=loss,
1247
+ logits=reshaped_logits,
1248
+ hidden_states=outputs.hidden_states,
1249
+ attentions=outputs.attentions,
1250
+ )
1251
+
1252
+
1253
+ @dataclass
1254
+ class NewTokenClassifierOutput(ModelOutput):
1255
+ loss: Optional[torch.FloatTensor] = None
1256
+ logits: torch.FloatTensor = None
1257
+ last_hidden_state: torch.FloatTensor = None
1258
+ hidden_states: Optional[Tuple[torch.FloatTensor, ...]] = None
1259
+ attentions: Optional[Tuple[torch.FloatTensor, ...]] = None
1260
+
1261
+
1262
+ class NewForTokenClassification(NewPreTrainedModel):
1263
+ def __init__(self, config):
1264
+ super().__init__(config)
1265
+ self.num_labels = config.num_labels
1266
+
1267
+ self.new = NewModel(config, add_pooling_layer=False)
1268
+ classifier_dropout = (
1269
+ config.classifier_dropout if config.classifier_dropout is not None else config.hidden_dropout_prob
1270
+ )
1271
+ self.dropout = nn.Dropout(classifier_dropout)
1272
+ self.classifier = nn.Linear(config.hidden_size, config.num_labels)
1273
+
1274
+ # Initialize weights and apply final processing
1275
+ self.post_init()
1276
+
1277
+ def forward(
1278
+ self,
1279
+ input_ids: Optional[torch.Tensor] = None,
1280
+ attention_mask: Optional[torch.Tensor] = None,
1281
+ token_type_ids: Optional[torch.Tensor] = None,
1282
+ position_ids: Optional[torch.Tensor] = None,
1283
+ head_mask: Optional[torch.Tensor] = None,
1284
+ inputs_embeds: Optional[torch.Tensor] = None,
1285
+ labels: Optional[torch.Tensor] = None,
1286
+ output_attentions: Optional[bool] = None,
1287
+ output_hidden_states: Optional[bool] = None,
1288
+ return_dict: Optional[bool] = None,
1289
+ unpad_inputs: Optional[bool] = None,
1290
+ ) -> Union[Tuple[torch.Tensor], NewTokenClassifierOutput]:
1291
+ r"""
1292
+ labels (`torch.LongTensor` of shape `(batch_size, sequence_length)`, *optional*):
1293
+ Labels for computing the token classification loss. Indices should be in `[0, ..., config.num_labels - 1]`.
1294
+ """
1295
+ return_dict = return_dict if return_dict is not None else self.config.use_return_dict
1296
+
1297
+ outputs = self.new(
1298
+ input_ids,
1299
+ attention_mask=attention_mask,
1300
+ token_type_ids=token_type_ids,
1301
+ position_ids=position_ids,
1302
+ head_mask=head_mask,
1303
+ inputs_embeds=inputs_embeds,
1304
+ output_attentions=output_attentions,
1305
+ output_hidden_states=output_hidden_states,
1306
+ return_dict=return_dict,
1307
+ unpad_inputs=unpad_inputs,
1308
+ )
1309
+
1310
+ sequence_output = outputs[0]
1311
+
1312
+ sequence_output = self.dropout(sequence_output)
1313
+ logits = self.classifier(sequence_output)
1314
+
1315
+ loss = None
1316
+ if labels is not None:
1317
+ loss_fct = nn.CrossEntropyLoss()
1318
+ loss = loss_fct(logits.view(-1, self.num_labels), labels.view(-1))
1319
+
1320
+ if not return_dict:
1321
+ output = (logits,) + outputs[2:]
1322
+ return ((loss,) + output) if loss is not None else output
1323
+
1324
+ return NewTokenClassifierOutput(
1325
+ loss=loss,
1326
+ logits=logits,
1327
+ last_hidden_state=sequence_output,
1328
+ hidden_states=outputs.hidden_states,
1329
+ attentions=outputs.attentions,
1330
+ )
1331
+
1332
+
1333
+ class NewForQuestionAnswering(NewPreTrainedModel):
1334
+ def __init__(self, config):
1335
+ super().__init__(config)
1336
+ self.num_labels = config.num_labels
1337
+
1338
+ self.new = NewModel(config, add_pooling_layer=False)
1339
+ self.qa_outputs = nn.Linear(config.hidden_size, config.num_labels)
1340
+
1341
+ # Initialize weights and apply final processing
1342
+ self.post_init()
1343
+
1344
+ def forward(
1345
+ self,
1346
+ input_ids: Optional[torch.Tensor] = None,
1347
+ attention_mask: Optional[torch.Tensor] = None,
1348
+ token_type_ids: Optional[torch.Tensor] = None,
1349
+ position_ids: Optional[torch.Tensor] = None,
1350
+ head_mask: Optional[torch.Tensor] = None,
1351
+ inputs_embeds: Optional[torch.Tensor] = None,
1352
+ start_positions: Optional[torch.Tensor] = None,
1353
+ end_positions: Optional[torch.Tensor] = None,
1354
+ output_attentions: Optional[bool] = None,
1355
+ output_hidden_states: Optional[bool] = None,
1356
+ return_dict: Optional[bool] = None,
1357
+ unpad_inputs: Optional[bool] = None,
1358
+ ) -> Union[Tuple[torch.Tensor], QuestionAnsweringModelOutput]:
1359
+ r"""
1360
+ start_positions (`torch.LongTensor` of shape `(batch_size,)`, *optional*):
1361
+ Labels for position (index) of the start of the labelled span for computing the token classification loss.
1362
+ Positions are clamped to the length of the sequence (`sequence_length`). Position outside of the sequence
1363
+ are not taken into account for computing the loss.
1364
+ end_positions (`torch.LongTensor` of shape `(batch_size,)`, *optional*):
1365
+ Labels for position (index) of the end of the labelled span for computing the token classification loss.
1366
+ Positions are clamped to the length of the sequence (`sequence_length`). Position outside of the sequence
1367
+ are not taken into account for computing the loss.
1368
+ """
1369
+ return_dict = return_dict if return_dict is not None else self.config.use_return_dict
1370
+
1371
+ outputs = self.new(
1372
+ input_ids,
1373
+ attention_mask=attention_mask,
1374
+ token_type_ids=token_type_ids,
1375
+ position_ids=position_ids,
1376
+ head_mask=head_mask,
1377
+ inputs_embeds=inputs_embeds,
1378
+ output_attentions=output_attentions,
1379
+ output_hidden_states=output_hidden_states,
1380
+ return_dict=return_dict,
1381
+ unpad_inputs=unpad_inputs,
1382
+ )
1383
+
1384
+ sequence_output = outputs[0]
1385
+
1386
+ logits = self.qa_outputs(sequence_output)
1387
+ start_logits, end_logits = logits.split(1, dim=-1)
1388
+ start_logits = start_logits.squeeze(-1).contiguous()
1389
+ end_logits = end_logits.squeeze(-1).contiguous()
1390
+
1391
+ total_loss = None
1392
+ if start_positions is not None and end_positions is not None:
1393
+ # If we are on multi-GPU, split add a dimension
1394
+ if len(start_positions.size()) > 1:
1395
+ start_positions = start_positions.squeeze(-1)
1396
+ if len(end_positions.size()) > 1:
1397
+ end_positions = end_positions.squeeze(-1)
1398
+ # sometimes the start/end positions are outside our model inputs, we ignore these terms
1399
+ ignored_index = start_logits.size(1)
1400
+ start_positions = start_positions.clamp(0, ignored_index)
1401
+ end_positions = end_positions.clamp(0, ignored_index)
1402
+
1403
+ loss_fct = nn.CrossEntropyLoss(ignore_index=ignored_index)
1404
+ start_loss = loss_fct(start_logits, start_positions)
1405
+ end_loss = loss_fct(end_logits, end_positions)
1406
+ total_loss = (start_loss + end_loss) / 2
1407
+
1408
+ if not return_dict:
1409
+ output = (start_logits, end_logits) + outputs[2:]
1410
+ return ((total_loss,) + output) if total_loss is not None else output
1411
+
1412
+ return QuestionAnsweringModelOutput(
1413
+ loss=total_loss,
1414
+ start_logits=start_logits,
1415
+ end_logits=end_logits,
1416
+ hidden_states=outputs.hidden_states,
1417
+ attentions=outputs.attentions,
1418
+ )
modules.json ADDED
@@ -0,0 +1,20 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ [
2
+ {
3
+ "idx": 0,
4
+ "name": "0",
5
+ "path": "",
6
+ "type": "sentence_transformers.models.Transformer"
7
+ },
8
+ {
9
+ "idx": 1,
10
+ "name": "1",
11
+ "path": "1_Pooling",
12
+ "type": "sentence_transformers.models.Pooling"
13
+ },
14
+ {
15
+ "idx": 2,
16
+ "name": "2",
17
+ "path": "2_Normalize",
18
+ "type": "sentence_transformers.models.Normalize"
19
+ }
20
+ ]
sentence_bert_config.json ADDED
@@ -0,0 +1,4 @@
 
 
 
 
 
1
+ {
2
+ "max_seq_length": 8192,
3
+ "do_lower_case": false
4
+ }
special_tokens_map.json ADDED
@@ -0,0 +1,51 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "bos_token": {
3
+ "content": "<s>",
4
+ "lstrip": false,
5
+ "normalized": false,
6
+ "rstrip": false,
7
+ "single_word": false
8
+ },
9
+ "cls_token": {
10
+ "content": "<s>",
11
+ "lstrip": false,
12
+ "normalized": false,
13
+ "rstrip": false,
14
+ "single_word": false
15
+ },
16
+ "eos_token": {
17
+ "content": "</s>",
18
+ "lstrip": false,
19
+ "normalized": false,
20
+ "rstrip": false,
21
+ "single_word": false
22
+ },
23
+ "mask_token": {
24
+ "content": "<mask>",
25
+ "lstrip": true,
26
+ "normalized": false,
27
+ "rstrip": false,
28
+ "single_word": false
29
+ },
30
+ "pad_token": {
31
+ "content": "<pad>",
32
+ "lstrip": false,
33
+ "normalized": false,
34
+ "rstrip": false,
35
+ "single_word": false
36
+ },
37
+ "sep_token": {
38
+ "content": "</s>",
39
+ "lstrip": false,
40
+ "normalized": false,
41
+ "rstrip": false,
42
+ "single_word": false
43
+ },
44
+ "unk_token": {
45
+ "content": "<unk>",
46
+ "lstrip": false,
47
+ "normalized": false,
48
+ "rstrip": false,
49
+ "single_word": false
50
+ }
51
+ }
tokenizer_config.json ADDED
@@ -0,0 +1,61 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "added_tokens_decoder": {
3
+ "0": {
4
+ "content": "<s>",
5
+ "lstrip": false,
6
+ "normalized": false,
7
+ "rstrip": false,
8
+ "single_word": false,
9
+ "special": true
10
+ },
11
+ "1": {
12
+ "content": "<pad>",
13
+ "lstrip": false,
14
+ "normalized": false,
15
+ "rstrip": false,
16
+ "single_word": false,
17
+ "special": true
18
+ },
19
+ "2": {
20
+ "content": "</s>",
21
+ "lstrip": false,
22
+ "normalized": false,
23
+ "rstrip": false,
24
+ "single_word": false,
25
+ "special": true
26
+ },
27
+ "3": {
28
+ "content": "<unk>",
29
+ "lstrip": false,
30
+ "normalized": false,
31
+ "rstrip": false,
32
+ "single_word": false,
33
+ "special": true
34
+ },
35
+ "250001": {
36
+ "content": "<mask>",
37
+ "lstrip": true,
38
+ "normalized": false,
39
+ "rstrip": false,
40
+ "single_word": false,
41
+ "special": true
42
+ }
43
+ },
44
+ "bos_token": "<s>",
45
+ "clean_up_tokenization_spaces": true,
46
+ "cls_token": "<s>",
47
+ "eos_token": "</s>",
48
+ "mask_token": "<mask>",
49
+ "max_length": 8192,
50
+ "model_max_length": 32768,
51
+ "pad_to_multiple_of": null,
52
+ "pad_token": "<pad>",
53
+ "pad_token_type_id": 0,
54
+ "padding_side": "right",
55
+ "sep_token": "</s>",
56
+ "stride": 0,
57
+ "tokenizer_class": "XLMRobertaTokenizer",
58
+ "truncation_side": "right",
59
+ "truncation_strategy": "longest_first",
60
+ "unk_token": "<unk>"
61
+ }