You can deploy this model and start earning money today!

Sadat07/
distilbert-base-uncased-SQuAD

Name: distilbert-base-uncased-SQuAD
Availability: InStock

safetensors

distilbert

Views

No views yet

Here's the revised documentation for your fine-tuned DistilBERT model:

`distilbert/distilbert-base-uncased` Fine-Tuned on SQuAD

distilbert_squad

Pretrained model on context-based Question Answering using the SQuAD dataset. This model is fine-tuned from the DistilBERT architecture for extracting answers from passages.

Model Description

distilbert_squad is a lightweight transformer-based model fine-tuned for context-based question-answering tasks. It adapts the pretrained DistilBERT architecture to extract precise answers from passages. This model was trained and fine-tuned on the Stanford Question Answering Dataset (SQuAD), leveraging its efficiency for resource-constrained environments.

Fine-tuned by: SADAT PARVEJ
Shared by: SADAT PARVEJ

Language(s) (NLP): ENGLISH
Finetuned from model: distilbert/distilbert-base-uncased

Training Objective

The model predicts the most relevant span of text in a given passage that answers a specific question. It fine-tunes DistilBERT's ability to analyze context using supervised data from SQuAD.

Performance Benchmarks

Training Loss: 0.464500
Validation Loss: 0.504376
Exact Match (EM): 85.59%
F1 Score: 85.59%

Intended Uses & Limitations

This model is designed for tasks such as:

Extractive Question Answering
Reading comprehension applications

Known Limitations:

As DistilBERT is distilled from BERT, its smaller architecture may limit its performance compared to the original BERT model on certain tasks.
The model's predictions may be biased or overly reliant on the training dataset, as SQuAD comprises structured and fact-based question-answer pairs.

How to Get Started with the Model

Use the code below to get started with the model:

python

import torch
from transformers import AutoTokenizer, AutoModelForQuestionAnswering

# Load the model and tokenizer
model_name = "YourHuggingFaceModelPath/distilbert_squad"
tokenizer = AutoTokenizer.from_pretrained(model_name)
model = AutoModelForQuestionAnswering.from_pretrained(model_name)

device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
model.to(device)

context = """
Thomas Edison is credited with inventing the light bulb. He was born in 1847 and was a prolific inventor.
"""
question = "Who invented the light bulb?"

inputs = tokenizer(question, context, return_tensors="pt", truncation=True, max_length=512)
input_ids = inputs["input_ids"].to(device)
attention_mask = inputs["attention_mask"].to(device)

# Perform inference
with torch.no_grad():
    outputs = model(input_ids=input_ids, attention_mask=attention_mask)
    start_scores = outputs.start_logits
    end_scores = outputs.end_logits

# Get start and end indices
start_idx = torch.argmax(start_scores)
end_idx = torch.argmax(end_scores) + 1

# Decode the answer
if start_idx >= end_idx:
    print("Model did not predict a valid answer. Please check context and question.")
else:
    answer = tokenizer.convert_tokens_to_string(
        tokenizer.convert_ids_to_tokens(input_ids[0][start_idx:end_idx])
    )
    print(f"Question: {question}")
    print(f"Answer: {answer}")

Training Details

Training Data

The model was fine-tuned using the SQuAD dataset, a benchmark for context-based question-answering tasks. It contains Wikipedia passages and human-annotated questions with answers.

Training Procedure

Training Objective
The model was trained to extract answers to specific questions based on a provided passage, leveraging DistilBERT's reduced architecture for faster inference and training.

Optimization

Optimizer: AdamW
Learning Rate Scheduler: Linear scheduler with warm-up steps
Steps: 2000 training steps with early stopping

Hardware and Resources

GPU: Tesla P100 or NVIDIA T4
Framework: Hugging Face Transformers

Training Metrics

The model achieved the following metrics during training and evaluation:

Step	Training Loss	Validation Loss	Exact Match	Squad F1	Start Accuracy	End Accuracy
100	0.719900	0.941330	84.66%	84.66%	84.66%	89.92%
500	0.640500	0.555793	84.87%	84.87%	84.87%	89.92%
1000	0.413100	0.551416	84.93%	84.93%	84.93%	89.92%
1500	0.522600	0.518057	85.17%	85.17%	85.17%	89.92%
2000	0.464500	0.504376	85.59%	85.59%	85.59%	89.92%

Results Summary

The fine-tuned DistilBERT model achieved:

Exact Match (EM): 85.59%
F1 Score: 85.59%
Validation Loss: 0.504376

This highlights DistilBERT’s efficiency and accuracy for context-based question-answering tasks when fine-tuned on SQuAD.

BibTeX

bibtex

@misc{distilbert_squad_finetune,
  title = {DistilBERT Fine-tuned for SQuAD},
  author = {Sadat Parvej},
  year = {2024},
  url = {https://huggingface.co/your-model-repository}
}

Let me know if you need further adjustments or additions!

Sadat07/distilbert-base-uncased-SQuAD

distilbert/distilbert-base-uncased Fine-Tuned on SQuAD