Developed by: Yuvaraj K
Model type: DistilBERT for Sequence Classification
Language: English
Finetuned from: distilbert-base-cased
Training platform: Kaggle GPU
Experiment tracking: Weights & Biases

Model Description

This model is a fine-tuned DistilBERT transformer model for multi-class text classification on Goodreads book review text.

The objective of this project was to build a complete MLOps workflow including:

Kaggle GPU training
secure API token handling with Kaggle Secrets
experiment tracking using Weights & Biases
evaluation artifact logging
model versioning and deployment using Hugging Face Hub
reproducibility with GitHub

DistilBERT was selected because it is computationally efficient, lightweight, and well-suited for free GPU environments while maintaining competitive NLP performance.

Training Data

Dataset: Goodreads book review classification dataset

Classification task: Multi-class genre classification

Number of classes: 8

Example classes:

Fiction
Fantasy
Mystery
Romance
Thriller
History
Science Fiction
Biography

Training Configuration

Base model: distilbert-base-cased

Hyperparameters:

Epochs: 3
Training batch size: 8
Evaluation batch size: 16
Weight decay: 0.01
Warmup steps: 100
Evaluation strategy: epoch
Save strategy: epoch
Optimizer: AdamW

Training platform:

Kaggle Notebook
GPU enabled

Evaluation Results

Metric	Score
Accuracy	0.6031
F1 Score	0.6007
Eval Loss	2.2373

Usage

python

from transformers import AutoTokenizer, AutoModelForSequenceClassification
import torch

model_name = "YuvarajK-g25ait2054/distilbert-goodreads-genres-V2"

tokenizer = AutoTokenizer.from_pretrained(model_name)
model = AutoModelForSequenceClassification.from_pretrained(model_name)

text = "This mystery novel was suspenseful and brilliantly written."

inputs = tokenizer(
    text,
    return_tensors="pt",
    truncation=True,
    padding=True
)

outputs = model(**inputs)
prediction = torch.argmax(outputs.logits, dim=1)

print(prediction)

Intended Use

This model is intended for:

educational NLP experimentation
text classification demonstrations
academic MLOps assignments
transformer fine-tuning examples

Limitations

trained on assignment dataset only
performance may vary on unseen real-world text
not validated for production deployment
classification quality depends on dataset label quality

Resources

Weights & Biases Dashboard:
https://wandb.ai/g25ait2054-iit-jodhpur/mlops-assignment2

Hugging Face Model:
https://huggingface.co/YuvarajK-g25ait2054/distilbert-goodreads-genres-V2

Kaggle Notebook:
https://www.kaggle.com/code/kyuvarajg25ait2054/g25ait2054-yuvaraj

GitHub Repository:
https://github.com/YuvarajK-g25ait2054/Mlops_assignment2