🧠 Q&AMODEL-SQUAD
A roberta-base-squad2 extractive Question Answering model fine-tuned on the SQuAD v2.0 dataset to predict precise answers from context passages, including handling unanswerable questions.
✨ Model Highlights
- 📌 Based on roberta-base-squad2
- 🔍 Fine-tuned on SQuAD v2.0 (or your custom QA dataset)
- ⚡ Supports extractive question answering finds precise answers from context passages
- 💾 Suitable for real-time inference with minimal latency on both CPU and GPU
- 🛠️ Easily integrable into web apps, enterprise tools, and virtual assistants
- 🔒 Handles unanswerable questions gracefully with no-answer detection (if trained on SQuAD v2)
🧠 Intended Uses
- ✅Customer support bots that extract answers from product manuals or FAQs
- ✅ Educational tools that answer student queries based on textbooks or syllabus
- ✅ Legal, financial, or technical document analysis
- ✅ Search engines with context-aware question answering
- ✅ Chatbots that require contextual comprehension for precise responses
🚫 Limitations
❌Trained primarily on formal text performance may degrade on informal or slang-heavy input
❌Does not support multi-hop questions requiring reasoning across multiple paragraphs
❌ May struggle with ambiguous questions or context with multiple possible answers
❌ Not designed for very long documents (performance may drop for inputs >512 tokens)
🏋️♂️ Training Details
| Field | Value |
|---|---|
| Base Model | roberta-base-squad2 |
| Dataset | SQuAD v2.0 |
| Framework | PyTorch with Transformers |
| Epochs | 3 |
| Batch Size | 16 |
| Optimizer | AdamW |
| Loss | CrossEntropyLoss (token-level) |
| Device | Trained on CUDA-enabled GPU |
📊 Evaluation Metrics
| Metric | Score |
|---|---|
| Accuracy | 0.80 |
| F1-Score | 0.78 |
| Precision | 0.79 |
| Recall | 0.78 |
🚀 Usage
from transformers import BertTokenizerFast, BertForTokenClassification
from transformers import pipeline
import torch
model_name = "AventIQ-AI/QA-Squad-Model"
tokenizer = AutoTokenizer.from_pretrained(model_checkpoint)
model = AutoModelForQuestionAnswering.from_pretrained(model_checkpoint)
model.eval()
#Inference
qa_pipeline = pipeline("question-answering", model="./qa_model", tokenizer="./qa_model")
# Provide a context and a question
context = """
The Amazon rainforest, also known as Amazonia, is a moist broadleaf tropical rainforest in the Amazon biome
that covers most of the Amazon basin of South America. This region includes territory belonging to nine nations.
"""
question = "What is the Amazon rainforest also known as?"
# Run inference
result = qa_pipeline(question=question, context=context)
# Print the result
print(f"Question: {question}")
print(f"Answer: {result['answer']}")
print(f"Score: {result['score']:.4f}")
- 🧩 Quantization
- Post-training static quantization applied using PyTorch to reduce model size and accelerate inference on edge devices.
🗂 Repository Structure
.
├── model/ # Quantized model files
├── tokenizer_config/ # Tokenizer and vocab files
├── model.safensors/ # Fine-tuned model in safetensors format
├── README.md # Model card
🤝 Contributing
Open to improvements and feedback! Feel free to submit a pull request or open an issue if you find any bugs or want to enhance the model.
- Downloads last month
- 1