Modelpublic

Atreyee-Halder/mlops-imdb-sentiment

sourceHugging Facemitupdated 4mo agoView on Hugging Face

1likes9downloads

Model Card

MLOps IMDB Sentiment Analysis Model ---‐----------------------------------

Model Description

Fine-tuned distilbert-base-uncased for binary sentiment classification on IMDB movie reviews.

Training Details

—Base Model: distilbert-base-uncased
—Dataset: IMDB Movie Reviews (50,000 samples)
—Task: Binary Text Classification
—Platform: Kaggle GPU T4 x2

Performance (run-v2 - Best Model)

Metric	Score
Accuracy	91.70%
F1 Score	91.70%
Validation Loss	0.7424

Hyperparameters

Parameter	Value
Learning Rate	5e-5
Epochs	3
Batch Size	16
Max Length	256

Experiment Comparison

Run	Learning Rate	Accuracy	F1
run-v1	3e-5	91.54%	91.53%
run-v2	5e-5	91.70%	91.70%

Usage

from transformers import pipeline classifier = pipeline('text-classification', model='Atreyee-Halder/mlops-imdb-sentiment') result = classifier("This movie was absolutely amazing!") print(result)

Labels

—0 = negative
—1 = positive

Project Links

—GitHub: https://github.com/halderatreyee-hash/mlops-pipeline-a3
—W&B: https://wandb.ai/g25ait2023-iit-jodhpur/mlops-assignment3
—Kaggle Notebook V1 https://www.kaggle.com/code/ahalderg25ait2023/mlops-group-project
—Kaggle Notebook V2 https://www.kaggle.com/code/ahalderg25ait2023/mlops-group-project-v2