metadata

language: en
license: mit
tags:
  - exbert
datasets:
  - squad_v2
thumbnail: >-
  https://thumb.tildacdn.com/tild3433-3637-4830-a533-353833613061/-/resize/720x/-/format/webp/germanquad.jpg
model-index:
  - name: deepset/tinybert-6l-768d-squad2
    results:
      - task:
          type: question-answering
          name: Question Answering
        dataset:
          name: squad_v2
          type: squad_v2
          config: squad_v2
          split: validation
        metrics:
          - type: exact_match
            value: 73.8248
            name: Exact Match
            verified: true
            verifyToken: >-
              eyJhbGciOiJFZERTQSIsInR5cCI6IkpXVCJ9.eyJoYXNoIjoiMGFmZmFiN2E5ODZkOTkyMjQ1NTUzMmQwMjc0M2RlYzVlNmM4YTFlNzA4YzIwY2JkY2EyNDg2ZTY3OTdjZTVlZiIsInZlcnNpb24iOjF9.ZZ6c2OI3lzeNhuSWTh28j00zk-sPrqkTvdVBZv2wJc1D4YnR-xOj72haybT6MV_xeYqTg3-x9L8PsWSS20NaDw
          - type: f1
            value: 77.1684
            name: F1
            verified: true
            verifyToken: >-
              eyJhbGciOiJFZERTQSIsInR5cCI6IkpXVCJ9.eyJoYXNoIjoiNzAxMDk1YzI5ZjA2N2ZmMzAxNjgxYzJiNzAzYmI1ZWU5ZDRmYWY3OWJmMjlmNDcyMGE0YWY5NjNhZTk4YWY5ZSIsInZlcnNpb24iOjF9.rF3raNGUSYv5D2xzWLZztD99vwDKvWb22LG32RomrDGP6XKTbCVqZzAw5UFw93jKb0VoLApbQQ-AOGxLj3U_Cg

tinybert for Extractive QA

Overview

Language model: deepset/tinybert-6L-768D-squad2
Language: English
Training data: SQuAD 2.0 training set x 20 augmented + SQuAD 2.0 training set without augmentation
Eval data: SQuAD 2.0 dev set Code: See an example extractive QA pipeline built with Haystack
Infrastructure: 1x V100 GPU
Published: Dec 8th, 2021

Details

Haystack's intermediate layer and prediction layer distillation features were used for training (based on TinyBERT). deepset/bert-base-uncased-squad2 was used as the teacher model and huawei-noah/TinyBERT_General_6L_768D was used as the student model.

Hyperparameters

Intermediate layer distillation

batch_size = 26
n_epochs = 5
max_seq_len = 384
learning_rate = 5e-5
lr_schedule = LinearWarmup
embeds_dropout_prob = 0.1
temperature = 1

Prediction layer distillation

batch_size = 26
n_epochs = 5
max_seq_len = 384
learning_rate = 3e-5
lr_schedule = LinearWarmup
embeds_dropout_prob = 0.1
temperature = 1
distillation_loss_weight = 1.0

Usage

In Haystack

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. To load and run the model with Haystack:

# After running pip install haystack-ai "transformers[torch,sentencepiece]"

from haystack import Document
from haystack.components.readers import ExtractiveReader

docs = [
    Document(content="Python is a popular programming language"),
    Document(content="python ist eine beliebte Programmiersprache"),
]

reader = ExtractiveReader(model="deepset/tinybert-6l-768d-squad2")
reader.warm_up()

question = "What is a popular programming language?"
result = reader.run(query=question, documents=docs)
# {'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),...)]}

For a complete example with an extractive question answering pipeline that scales over many documents, check out the corresponding Haystack tutorial.

In Transformers

from transformers import AutoModelForQuestionAnswering, AutoTokenizer, pipeline

model_name = "deepset/tinybert-6l-768d-squad2"

# a) Get predictions
nlp = pipeline('question-answering', model=model_name, tokenizer=model_name)
QA_input = {
    'question': 'Why is model conversion important?',
    'context': 'The option to convert models between FARM and transformers gives freedom to the user and let people easily switch between frameworks.'
}
res = nlp(QA_input)

# b) Load model & tokenizer
model = AutoModelForQuestionAnswering.from_pretrained(model_name)
tokenizer = AutoTokenizer.from_pretrained(model_name)

Performance

"exact": 71.87736882001179
"f1": 76.36111895973675

Authors

Timo Möller: timo.moeller [at] deepset.ai
Julian Risch: julian.risch [at] deepset.ai
Malte Pietsch: malte.pietsch [at] deepset.ai
Michel Bartels: michel.bartels [at] deepset.ai

About us

deepset is the company behind the production-ready open-source AI framework Haystack.

Some of our other work:

Get in touch and join the Haystack community

For more info on Haystack, visit our GitHub repo and Documentation.

We also have a Discord community open to everyone!

By the way: we're hiring!