Added Q&A language pairs dataset and training

[FiveTechSoft]

No description provided.

[Copilot]

Pull Request Overview

This PR adds support for training on Q&A language pairs by introducing a new dataset builder and training script. The implementation enables the TinyRecursiveModels framework to work with natural language question-answer pairs.

Key changes:

• New Q&A dataset builder with templates for factual, mathematical, color, animal, time, and weather questions
• Training script configured for Q&A pairs with appropriate hyperparameters
• Documentation updates showing how to prepare data, train, and evaluate Q&A models

Reviewed Changes

Copilot reviewed 3 out of 3 changed files in this pull request and generated 10 comments.

File	Description
train_qa_pairs.sh	Shell script with hyperparameters for training the model on Q&A dataset
dataset/build_qa_dataset.py	Python script that generates Q&A pairs from templates and converts them to the expected dataset format
README.md	Documentation updates adding Q&A pairs example and correcting evaluation command

Comments suppressed due to low confidence (1)

dataset/build_qa_dataset.py:184

• Variable train_data is not used.

    train_data = []

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[Copilot]

The city name "Brasília" contains a special character (í) that may cause encoding issues depending on the tokenization approach used. Since the create_qa_sequence() function uses a simple regex-based word tokenizer that only matches \b\w+\b, the accented character might not be handled correctly. Consider using "Brasilia" without the accent or ensuring proper Unicode handling in the tokenizer.

Suggested change

	{"country": "Brazil", "capital": "Brasília"},
	{"country": "Brazil", "capital": "Brasilia"},