Prerequisites

• Git LFS

We use it to manage large files. Anyone cloning this repo must install Git LFS first, then clone the repo.

Sentiment Analysis on Movie Reviews

This project is about sentiment analysis on movie reviews. We fine-tune a pre-trained model on the IMDB dataset for sentiment analysis. We provide a Gradio UI for sentiment analysis.

Setup

Clone the repo

git clone git@github.com:Hivekind/fine-tune-model.git

Install dependencies

# create python virtual env
python -m venv env

# activate the env
source env/bin/activate

# install dependencies
pip3 install -r requirements.txt

Run the model via Docker

You can run the model via Docker. The model API server is running on http://localhost:5000.

docker compose up

The API endpoint is /sentiment and it accepts a POST request with a JSON payload:

{
  "review": "This is a great movie!"
}

You can test the model API server using curl:

  curl -X POST http://localhost:5000/sentiment \
       -H "Content-Type: application/json" \
       -d '{"review": "This movie was absolutely fantastic, I loved every minute of it!"}'

Run the UI via Gradio

You can run the Gradio UI for sentiment analysis. The UI is running on http://localhost:7860.

python sentiment_analysis.py

It takes a text input and returns the sentiment of the text, by sending a POST request to the model API server. The UI looks like this:

Project Structure

Pre-trained Model

We use the distilbert-base-uncased model from the Hugging Face Transformers library.

Dataset

The dataset is the IMDB dataset from the Hugging Face Datasets library. The dataset contains 50,000 movie reviews, with 25,000 reviews for training and 25,000 reviews for validation. The dataset is split into two classes: positive and negative reviews.

Fine-tuning the model

We fine-tune the model on the IMDB dataset for sentiment analysis. The training was done on Google Colab. The fine-tuned model is saved in the movie_sentiment_model directory. The model is saved in the model.safetensors format.

You can find the fine-tuning code here.

Running the model for sentiment analysis on Colab

After fine-tuning the model, we can run the model on Google Colab for sentiment analysis, using the Gradio UI. You can find the code here.

Docker image for fine-tuned model

We package the fine-tuned model into a Docker image, and provide an API for sentiment analysis.

docker build -t sentiment-analysis-model .

Diagram

You can find the project structure in the diagram directory.

• fine-tuning
• dockerized-model

Managing Large Files with Git LFS

We use Git LFS to manage large files in this repository.

Initial Setup

Install Git LFS:

# for macOS
brew install git-lfs

# for Ubuntu
sudo apt install git-lfs

Set up Git LFS to track large file:

# Initialize Git LFS
git lfs install

# Track large file
git lfs track "movie_sentiment_model/model.safetensors"

# Add and commit
git add .gitattributes
git add .
git commit -m "Track large file with Git LFS"
git push origin main