Question Decomposer T5

## 🚀 質問分解器（T5とSeq2seqベース）

*このプロジェクトはT5とSeq2seqをベースにした質問分解器で、複合的な質問を複数のサブ質問に分解します。*

### データセットとモデル情報
| プロパティ | 詳細 |
|----------|---------|
| データセット | microsoft/ms_marco |
| 言語 | en |
| ベースモデル | google-t5/t5-base |
| パイプラインタグ | text2text-generation |

### 例
例: What is the capital of France and when it entablish ?
- What is the capital of France ?
- When was the capital of France entablish ?

👉🏻 [デモ](https://huggingface.co/spaces/thenHung/Demo-question-decomposer) をこちらで確認できます。

## 🚀 クイックスタート

### 💻 使用例

#### 基本的な使用法
```python
from transformers import T5Tokenizer, T5ForConditionalGeneration
import torch

# Set device
device = "mps" if torch.backends.mps.is_available() else "cuda" if torch.cuda.is_available() else "cpu"

# Load model and tokenizer
model_path = "thenHung/question_decomposer_t5"
tokenizer = T5Tokenizer.from_pretrained(model_path)
model = T5ForConditionalGeneration.from_pretrained(model_path)
model.to(device)
model.eval()

# Decompose question
question = "Who is taller between John and Mary?"
input_text = f"decompose question: {question}"
input_ids = tokenizer(
    input_text,
    max_length=128,
    padding="max_length",
    truncation=True,
    return_tensors="pt"
).input_ids.to(device)

with torch.no_grad():
    outputs = model.generate(
        input_ids,
        max_length=128,
        num_beams=4,
        early_stopping=True
    )

# Decode output
decoded_output = tokenizer.decode(outputs[0], skip_special_tokens=True)
sub_questions = decoded_output.split(" [SEP] ")

# Print sub-questions
print(sub_questions)
# ['What is the height of John?', 'What is the height of Mary?']