metadata

license: mit
datasets:
  - squad_v2
  - squad
language:
  - en
library_name: transformers
pipeline_tag: question-answering
tags:
  - question-answering
  - squad
  - squad_v2
  - t5
model-index:
  - name: sjrhuschlee/flan-t5-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: 82.203
            name: Exact Match
          - type: f1
            value: 85.283
            name: F1
      - task:
          type: question-answering
          name: Question Answering
        dataset:
          name: squad
          type: squad
          config: plain_text
          split: validation
        metrics:
          - type: exact_match
            value: 86.32
            name: Exact Match
          - type: f1
            value: 92.947
            name: F1
      - task:
          type: question-answering
          name: Question Answering
        dataset:
          name: adversarial_qa
          type: adversarial_qa
          config: adversarialQA
          split: validation
        metrics:
          - type: exact_match
            value: 23.133
            name: Exact Match
          - type: f1
            value: 31.386
            name: F1
      - task:
          type: question-answering
          name: Question Answering
        dataset:
          name: squad_adversarial
          type: squad_adversarial
          config: AddOneSent
          split: validation
        metrics:
          - type: exact_match
            value: 68.159
            name: Exact Match
          - type: f1
            value: 71.876
            name: F1
      - task:
          type: question-answering
          name: Question Answering
        dataset:
          name: squadshifts
          type: squadshifts
          config: amazon
          split: test
        metrics:
          - type: exact_match
            value: 67.587
            name: Exact Match
          - type: f1
            value: 80.085
            name: F1
      - task:
          type: question-answering
          name: Question Answering
        dataset:
          name: squadshifts
          type: squadshifts
          config: new_wiki
          split: test
        metrics:
          - type: exact_match
            value: 77.261
            name: Exact Match
          - type: f1
            value: 85.068
            name: F1
      - task:
          type: question-answering
          name: Question Answering
        dataset:
          name: squadshifts
          type: squadshifts
          config: nyt
          split: test
        metrics:
          - type: exact_match
            value: 79.066
            name: Exact Match
          - type: f1
            value: 86.178
            name: F1

flan-t5-base for Extractive QA

This is the flan-t5-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.

NOTE: The <cls> token must be manually added to the beginning of the question for this model to work properly. It uses the <cls> token to be able to make "no answer" predictions. The t5 tokenizer does not automatically add this special token which is why it is added manually.

Overview

Language model: flan-t5-base
Language: English
Downstream-task: Extractive QA
Training data: SQuAD 2.0
Eval data: SQuAD 2.0
Infrastructure: 1x NVIDIA 3070

Model Usage

import torch
from transformers import(
  AutoModelForQuestionAnswering,
  AutoTokenizer,
  pipeline
)
model_name = "sjrhuschlee/flan-t5-base-squad2"

# a) Using pipelines
nlp = pipeline(
  'question-answering',
  model=model_name,
  tokenizer=model_name,
  trust_remote_code=True,
)
qa_input = {
'question': f'{nlp.tokenizer.cls_token}Where do I live?',  # '<cls>Where do I live?'
'context': 'My name is Sarah and I live in London'
}
res = nlp(qa_input)
# {'score': 0.980, 'start': 30, 'end': 37, 'answer': ' London'}

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

question = f'{tokenizer.cls_token}Where do I live?'  # '<cls>Where do I live?'
context = 'My name is Sarah and I live in London'
encoding = tokenizer(question, context, return_tensors="pt")
output = model(
  encoding["input_ids"],
  attention_mask=encoding["attention_mask"]
)

all_tokens = tokenizer.convert_ids_to_tokens(encoding["input_ids"][0].tolist())
answer_tokens = all_tokens[torch.argmax(output["start_logits"]):torch.argmax(output["end_logits"]) + 1]
answer = tokenizer.decode(tokenizer.convert_tokens_to_ids(answer_tokens))
# 'London'

Metrics

# Squad v2
{
    "eval_HasAns_exact": 79.97638326585695,
    "eval_HasAns_f1": 86.1444296592862,
    "eval_HasAns_total": 5928,
    "eval_NoAns_exact": 84.42388561816652,
    "eval_NoAns_f1": 84.42388561816652,
    "eval_NoAns_total": 5945,
    "eval_best_exact": 82.2033184536343,
    "eval_best_exact_thresh": 0.0,
    "eval_best_f1": 85.28292588395921,
    "eval_best_f1_thresh": 0.0,
    "eval_exact": 82.2033184536343,
    "eval_f1": 85.28292588395928,
    "eval_runtime": 522.0299,
    "eval_samples": 12001,
    "eval_samples_per_second": 22.989,
    "eval_steps_per_second": 0.96,
    "eval_total": 11873
}

# Squad
{
    "eval_exact_match": 86.3197729422895,
    "eval_f1": 92.94686836210295,
    "eval_runtime": 442.1088,
    "eval_samples": 10657,
    "eval_samples_per_second": 24.105,
    "eval_steps_per_second": 1.007
}

Training procedure

Training hyperparameters

The following hyperparameters were used during training:

learning_rate: 2e-05
train_batch_size: 16
eval_batch_size: 8
seed: 42
gradient_accumulation_steps: 6
total_train_batch_size: 96
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

Training results

Framework versions

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