Kapilydv6/my-qa-model

My First Question Answering Model ❓

A fine-tuned DistilBERT model for extractive question answering. Given a context paragraph and a question, it extracts the answer directly from the text.

How to Use

from transformers import pipeline

qa = pipeline("question-answering", model="Kapilydv6/my-qa-model")

result = qa(
    question="Who created Python?",
    context="Python is a programming language created by Guido van Rossum."
)
print(result["answer"])  # "Guido van Rossum"

Training Details

Parameter	Value
Base model	distilbert-base-uncased
Dataset	SQuAD v1.1 (3000 samples)
Epochs	3
Learning rate	3e-5
Max sequence len	384
Framework	PyTorch + Transformers

What I Learned

This is my second Hugging Face model! Key concepts:

• Extractive Q&A: The model doesn't generate text — it finds the answer span within the given context
• Token-level prediction: Unlike classification (one label per text), Q&A predicts start and end token positions
• Stride: Long documents are split into overlapping chunks so no answer is missed

Limitations

Trained on a small subset of SQuAD for learning purposes. For production, train on the full dataset (~87k examples).