README.md
8.5 KB · 247 lines · markdown Raw
1 ---
2 language: en
3 license: cc-by-4.0
4 tags:
5 - deberta
6 - deberta-v3
7 datasets:
8 - squad_v2
9 base_model: microsoft/deberta-v3-base
10 model-index:
11 - name: deepset/deberta-v3-base-squad2
12 results:
13 - task:
14 type: question-answering
15 name: Question Answering
16 dataset:
17 name: squad_v2
18 type: squad_v2
19 config: squad_v2
20 split: validation
21 metrics:
22 - type: exact_match
23 value: 83.8248
24 name: Exact Match
25 verified: true
26 verifyToken: eyJhbGciOiJFZERTQSIsInR5cCI6IkpXVCJ9.eyJoYXNoIjoiY2IyZTEyYzNlOTAwZmFlNWRiZTdiNzQzMTUyM2FmZTQ3ZWQwNWZmMzc2ZDVhYWYyMzkxOTUyMGNlMWY0M2E5MiIsInZlcnNpb24iOjF9.y8KvfefMLI977BYun0X1rAq5qudmezW_UJe9mh6sYBoiWaBosDO5TRnEGR1BHzdxmv2EgPK_PSomtZvb043jBQ
27 - type: f1
28 value: 87.41
29 name: F1
30 verified: true
31 verifyToken: eyJhbGciOiJFZERTQSIsInR5cCI6IkpXVCJ9.eyJoYXNoIjoiOWVhNjAwM2Q5N2Y3MGU4ZWY3N2Y0MmNjYWYwYmQzNTdiYWExODhkYmQ1YjIwM2I1ODEzNWIxZDI1ZWQ1YWRjNSIsInZlcnNpb24iOjF9.Jk0v1ZheLRFz6k9iNAgCMMZtPYj5eVwUCku4E76wRYc-jHPmiUuxvNiNkn6NW-jkBD8bJGMqDSjJyVpVMn9pBA
32 - task:
33 type: question-answering
34 name: Question Answering
35 dataset:
36 name: squad
37 type: squad
38 config: plain_text
39 split: validation
40 metrics:
41 - type: exact_match
42 value: 84.9678
43 name: Exact Match
44 verified: true
45 verifyToken: eyJhbGciOiJFZERTQSIsInR5cCI6IkpXVCJ9.eyJoYXNoIjoiOWUxYTg4MzU3YTdmMDRmMGM0NjFjMTcwNGM3YzljM2RkMTc1ZGNhMDQwMTgwNGI0ZDE4ZGMxZTE3YjY5YzQ0ZiIsInZlcnNpb24iOjF9.KKaJ1UtikNe2g6T8XhLoWNtL9X4dHHyl_O4VZ5LreBT9nXneGc21lI1AW3n8KXTFGemzRpRMvmCDyKVDHucdDQ
46 - type: f1
47 value: 92.2777
48 name: F1
49 verified: true
50 verifyToken: eyJhbGciOiJFZERTQSIsInR5cCI6IkpXVCJ9.eyJoYXNoIjoiNDU0ZTQwMzg4ZDY1ZWYxOGIxMzY2ODljZTBkMTNlYjA0ODBjNjcxNTg3ZDliYWU1YTdkYTM2NTIxOTg1MGM4OCIsInZlcnNpb24iOjF9.8VHg1BXx6gLw_K7MUK2QSE80Y9guiVR8n8K8nX4laGsLibxv5u_yDv9F3ahbUa1eZG_bbidl93TY2qFUiYHtAQ
51 - task:
52 type: question-answering
53 name: Question Answering
54 dataset:
55 name: adversarial_qa
56 type: adversarial_qa
57 config: adversarialQA
58 split: validation
59 metrics:
60 - type: exact_match
61 value: 30.733
62 name: Exact Match
63 - type: f1
64 value: 44.099
65 name: F1
66 - task:
67 type: question-answering
68 name: Question Answering
69 dataset:
70 name: squad_adversarial
71 type: squad_adversarial
72 config: AddOneSent
73 split: validation
74 metrics:
75 - type: exact_match
76 value: 79.295
77 name: Exact Match
78 - type: f1
79 value: 86.609
80 name: F1
81 - task:
82 type: question-answering
83 name: Question Answering
84 dataset:
85 name: squadshifts amazon
86 type: squadshifts
87 config: amazon
88 split: test
89 metrics:
90 - type: exact_match
91 value: 68.680
92 name: Exact Match
93 - type: f1
94 value: 83.832
95 name: F1
96 - task:
97 type: question-answering
98 name: Question Answering
99 dataset:
100 name: squadshifts new_wiki
101 type: squadshifts
102 config: new_wiki
103 split: test
104 metrics:
105 - type: exact_match
106 value: 80.171
107 name: Exact Match
108 - type: f1
109 value: 90.452
110 name: F1
111 - task:
112 type: question-answering
113 name: Question Answering
114 dataset:
115 name: squadshifts nyt
116 type: squadshifts
117 config: nyt
118 split: test
119 metrics:
120 - type: exact_match
121 value: 81.570
122 name: Exact Match
123 - type: f1
124 value: 90.644
125 name: F1
126 - task:
127 type: question-answering
128 name: Question Answering
129 dataset:
130 name: squadshifts reddit
131 type: squadshifts
132 config: reddit
133 split: test
134 metrics:
135 - type: exact_match
136 value: 66.990
137 name: Exact Match
138 - type: f1
139 value: 80.231
140 name: F1
141 ---
142
143 # deberta-v3-base for Extractive QA
144
145 This is the [deberta-v3-base](https://huggingface.co/microsoft/deberta-v3-base) model, fine-tuned using the [SQuAD2.0](https://huggingface.co/datasets/squad_v2) dataset. It's been trained on question-answer pairs, including unanswerable questions, for the task of Question Answering.
146
147
148 ## Overview
149 **Language model:** deberta-v3-base
150 **Language:** English
151 **Downstream-task:** Extractive QA
152 **Training data:** SQuAD 2.0
153 **Eval data:** SQuAD 2.0
154 **Code:** See [an example extractive QA pipeline built with Haystack](https://haystack.deepset.ai/tutorials/34_extractive_qa_pipeline)
155 **Infrastructure**: 1x NVIDIA A10G
156
157 ## Hyperparameters
158
159 ```
160 batch_size = 12
161 n_epochs = 4
162 base_LM_model = "deberta-v3-base"
163 max_seq_len = 512
164 learning_rate = 2e-5
165 lr_schedule = LinearWarmup
166 warmup_proportion = 0.2
167 doc_stride = 128
168 max_query_length = 64
169 ```
170
171 ## Usage
172
173 ### In Haystack
174 Haystack is an AI orchestration framework to build customizable, production-ready LLM applications. You can use this model in Haystack to do extractive question answering on documents.
175 To load and run the model with [Haystack](https://github.com/deepset-ai/haystack/):
176 ```python
177 # After running pip install haystack-ai "transformers[torch,sentencepiece]"
178
179 from haystack import Document
180 from haystack.components.readers import ExtractiveReader
181
182 docs = [
183 Document(content="Python is a popular programming language"),
184 Document(content="python ist eine beliebte Programmiersprache"),
185 ]
186
187 reader = ExtractiveReader(model="deepset/roberta-base-squad2")
188 reader.warm_up()
189
190 question = "What is a popular programming language?"
191 result = reader.run(query=question, documents=docs)
192 # {'answers': [ExtractedAnswer(query='What is a popular programming language?', score=0.5740374326705933, data='python', document=Document(id=..., content: '...'), context=None, document_offset=ExtractedAnswer.Span(start=0, end=6),...)]}
193 ```
194 For a complete example with an extractive question answering pipeline that scales over many documents, check out the [corresponding Haystack tutorial](https://haystack.deepset.ai/tutorials/34_extractive_qa_pipeline).
195
196 ### In Transformers
197 ```python
198 from transformers import AutoModelForQuestionAnswering, AutoTokenizer, pipeline
199
200 model_name = "deepset/roberta-base-squad2"
201
202 # a) Get predictions
203 nlp = pipeline('question-answering', model=model_name, tokenizer=model_name)
204 QA_input = {
205 'question': 'Why is model conversion important?',
206 'context': 'The option to convert models between FARM and transformers gives freedom to the user and let people easily switch between frameworks.'
207 }
208 res = nlp(QA_input)
209
210 # b) Load model & tokenizer
211 model = AutoModelForQuestionAnswering.from_pretrained(model_name)
212 tokenizer = AutoTokenizer.from_pretrained(model_name)
213 ```
214
215
216 ## Authors
217 **Sebastian Lee:** sebastian.lee [at] deepset.ai
218 **Timo Möller:** timo.moeller [at] deepset.ai
219 **Malte Pietsch:** malte.pietsch [at] deepset.ai
220
221 ## About us
222
223 <div class="grid lg:grid-cols-2 gap-x-4 gap-y-3">
224 <div class="w-full h-40 object-cover mb-2 rounded-lg flex items-center justify-center">
225 <img alt="" src="https://raw.githubusercontent.com/deepset-ai/.github/main/deepset-logo-colored.png" class="w-40"/>
226 </div>
227 <div class="w-full h-40 object-cover mb-2 rounded-lg flex items-center justify-center">
228 <img alt="" src="https://raw.githubusercontent.com/deepset-ai/.github/main/haystack-logo-colored.png" class="w-40"/>
229 </div>
230 </div>
231
232 [deepset](http://deepset.ai/) is the company behind the production-ready open-source AI framework [Haystack](https://haystack.deepset.ai/).
233
234 Some of our other work:
235 - [Distilled roberta-base-squad2 (aka "tinyroberta-squad2")](https://huggingface.co/deepset/tinyroberta-squad2)
236 - [German BERT](https://deepset.ai/german-bert), [GermanQuAD and GermanDPR](https://deepset.ai/germanquad), [German embedding model](https://huggingface.co/mixedbread-ai/deepset-mxbai-embed-de-large-v1)
237 - [deepset Cloud](https://www.deepset.ai/deepset-cloud-product), [deepset Studio](https://www.deepset.ai/deepset-studio)
238
239 ## Get in touch and join the Haystack community
240
241 <p>For more info on Haystack, visit our <strong><a href="https://github.com/deepset-ai/haystack">GitHub</a></strong> repo and <strong><a href="https://docs.haystack.deepset.ai">Documentation</a></strong>.
242
243 We also have a <strong><a class="h-7" href="https://haystack.deepset.ai/community">Discord community open to everyone!</a></strong></p>
244
245 [Twitter](https://twitter.com/Haystack_AI) | [LinkedIn](https://www.linkedin.com/company/deepset-ai/) | [Discord](https://haystack.deepset.ai/community) | [GitHub Discussions](https://github.com/deepset-ai/haystack/discussions) | [Website](https://haystack.deepset.ai/) | [YouTube](https://www.youtube.com/@deepset_ai)
246
247 By the way: [we're hiring!](http://www.deepset.ai/jobs)