README.md

metadata

license: mit
datasets:
  - squad_v2
language:
  - en
library_name: transformers
pipeline_tag: question-answering
tags:
  - deberta
  - deberta-v3
  - question-answering
model-index:
  - name: sjrhuschlee/deberta-v3-base-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: 85.648
            name: Exact Match
          - type: f1
            value: 88.728
            name: F1
      - task:
          type: question-answering
          name: Question Answering
        dataset:
          name: squad
          type: squad
          config: plain_text
          split: validation
        metrics:
          - type: exact_match
            value: 87.862
            name: Exact Match
          - type: f1
            value: 93.924
            name: F1

deberta-v3-base for QA

This is the deberta-v3-base model, fine-tuned using the SQuAD2.0 dataset. It's been trained on question-answer pairs, including unanswerable questions, for the task of Extractive Question Answering.

Overview

Language model: deberta-v3-base
Language: English
Downstream-task: Extractive QA
Training data: SQuAD 2.0
Eval data: SQuAD 2.0
Infrastructure: 1x NVIDIA 3070

Model Usage

from transformers import AutoModelForQuestionAnswering, AutoTokenizer, pipeline
model_name = "sjrhuschlee/deberta-v3-base-squad2"

# a) Using pipelines
nlp = pipeline('question-answering', model=model_name, tokenizer=model_name)
qa_input = {
'question': 'Where do I live?',
'context': 'My name is Sarah and I live in London'
}
res = nlp(qa_input)

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

Metrics

# Squad v2
{
    "eval_HasAns_exact": 82.72604588394061,
    "eval_HasAns_f1": 88.89430905100325,
    "eval_HasAns_total": 5928,
    "eval_NoAns_exact": 88.56181665264928,
    "eval_NoAns_f1": 88.56181665264928,
    "eval_NoAns_total": 5945,
    "eval_best_exact": 85.64810915522614,
    "eval_best_exact_thresh": 0.0,
    "eval_best_f1": 88.72782481717712,
    "eval_best_f1_thresh": 0.0,
    "eval_exact": 85.64810915522614,
    "eval_f1": 88.72782481717726,
    "eval_runtime": 219.6226,
    "eval_samples": 11951,
    "eval_samples_per_second": 54.416,
    "eval_steps_per_second": 2.268,
    "eval_total": 11873
}

# Squad
{
    "eval_exact_match": 87.86187322611164,
    "eval_f1": 93.92373735474943,
    "eval_runtime": 195.2115,
    "eval_samples": 10618,
    "eval_samples_per_second": 54.392,
    "eval_steps_per_second": 2.269
}

Training procedure

Training hyperparameters

The following hyperparameters were used during training:

• learning_rate: 5e-06
• train_batch_size: 8
• eval_batch_size: 8
• seed: 42
• gradient_accumulation_steps: 8
• total_train_batch_size: 64
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_ratio: 0.1
• num_epochs: 4.0

Framework versions

• Transformers 4.30.0.dev0
• Pytorch 2.0.1+cu117
• Datasets 2.12.0
• Tokenizers 0.13.3