【Huggingface Transformers】英語の質問応答の実装 〜学習から推論まで〜

from google.colab import drive 
drive.mount('/content/drive')
!mkdir -p '/content/drive/My Drive/huggingface_transformers_demo/'
%cd '/content/drive/My Drive/huggingface_transformers_demo/'

!git clone https://github.com/huggingface/transformers
%cd transformers

!pip install git+https://github.com/huggingface/transformers

!pip install -r ./examples/pytorch/question-answering/requirements.txt

# Huggingface Datasetsのインストール
!pip install datasets

from datasets import load_dataset
import pandas as pd

# データセット「squad」の中身を確認する
test = load_dataset("squad")
test

DatasetDict({
    train: Dataset({
        features: ['id', 'title', 'context', 'question', 'answers'],
        num_rows: 87599
    })
    validation: Dataset({
        features: ['id', 'title', 'context', 'question', 'answers'],
        num_rows: 10570
    })
})

# データセット「squad」の中身を確認する
dfs = {split: dset.to_pandas() for split, dset in test.flatten().items()}
qa_cols = ["title", "question", "answers.text", 
           "answers.answer_start", "context"]
sample_df = dfs["train"][qa_cols][:3]
sample_df

!python ./examples/pytorch/question-answering/run_qa.py \
  --model_name_or_path bert-base-uncased \
  --dataset_name squad \
  --do_train \
  --do_eval \
  --per_device_train_batch_size 12 \
  --learning_rate 3e-5 \
  --num_train_epochs 2 \
  --max_seq_length 384 \
  --doc_stride 128 \
  --output_dir /tmp/debug_squad/

***** eval metrics *****
  epoch            =     2.0
  eval_exact_match = 81.0501
  eval_f1          = 88.3606
  eval_samples     =   10784

from transformers import AutoTokenizer, AutoModelForQuestionAnswering
import torch

# トークナイザーとモデルをインスタンス化
model = AutoModelForQuestionAnswering.from_pretrained('/tmp/debug_squad/')  
tokenizer = AutoTokenizer.from_pretrained("bert-large-uncased-whole-word-masking-finetuned-squad") 

# 元の文章（入力）を定義
text = r"""
 Transformers (formerly known as pytorch-transformers and pytorch-pretrained-bert) provides general-purpose
architectures (BERT, GPT-2, RoBERTa, XLM, DistilBert, XLNet…) for Natural Language Understanding (NLU) and Natural
Language Generation (NLG) with over 32+ pretrained models in 100+ languages and deep interoperability between
TensorFlow 2.0 and PyTorch.
"""
# 質問を定義
questions = [
    "How many pretrained models are available in  Transformers?",
    "What does  Transformers provide?",
    " Transformers provides interoperability between which frameworks?",
]

for question in questions:
    # 質問を反復処理し、テキストと現在の質問からシーケンスを構築
    inputs = tokenizer(question, text, add_special_tokens=True, return_tensors="pt")
    input_ids = inputs["input_ids"].tolist()[0]
    # このシーケンスをモデルに渡して、開始位置と終了位置の両方について、シーケンストークン全体 (質問とテキスト) のスコアの範囲を出力
    outputs = model(**inputs)
    answer_start_scores = outputs.start_logits
    answer_end_scores = outputs.end_logits

    # スコアのargmaxで最も可能性の高い解答の先頭を取得する
    answer_start = torch.argmax(answer_start_scores)

    # スコアのargmaxで最も可能性の高い解答の末尾を取得する
    answer_end = torch.argmax(answer_end_scores) + 1
    
    # top_kメソッドを使用して上位 5 つのトークンを取得
    answer = tokenizer.convert_tokens_to_string(
        tokenizer.convert_ids_to_tokens(input_ids[answer_start:answer_end])
    )

    print(f"Question: {question}")
    print(f"Answer: {answer}")

Question: How many pretrained models are available in  Transformers?
Answer: 32 +
Question: What does  Transformers provide?
Answer: general - purpose architectures
Question:  Transformers provides interoperability between which frameworks?
Answer: tensorflow 2. 0 and pytorch