app.py

from fastapi import FastAPI, Request, Response

import gradio
from transformers import pipeline

from sklearn.metrics import f1_score, precision_score, recall_score
import prometheus_client as prom


app = FastAPI()

username = "yrajm1997"
repo_name = "finetuned-sentiment-model"
repo_path = username+ '/' + repo_name
sentiment_model = pipeline(model= repo_path)

import pandas as pd

test_data = pd.read_csv("test_reviews.csv")

f1_metric = prom.Gauge('sentiment_f1_score', 'F1 score for random 100 test samples')
precision_metric = prom.Gauge('sentiment_precision_score', 'Precision score for random 100 test samples')
recall_metric = prom.Gauge('sentiment_recall_score', 'Recall score for random 100 test samples')


# Function for response generation
def predict_sentiment(text):
    result = sentiment_model(text)
    if result[0]['label'].endswith('0'):
        return 'Negative'
    else:
        return 'Positive'

# Function for updating metrics
def update_metrics():
    global test_data
    test = test_data.sample(100)
    test_text = test['Text'].values
    test_pred = sentiment_model(list(test_text))
    pred_labels = [int(pred['label'].split("_")[1]) for pred in test_pred]
    f1 = f1_score(test['labels'], pred_labels).round(3)
    precision = precision_score(test['labels'], pred_labels).round(3)
    recall = recall_score(test['labels'], pred_labels).round(3)
    
    f1_metric.set(f1)
    precision_metric.set(precision)
    recall_metric.set(recall)


@app.get("/metrics")
async def get_metrics():
    update_metrics()
    return Response(media_type="text/plain", content= prom.generate_latest())


# Input from user
in_prompt = gradio.components.Textbox(lines=10, placeholder=None, label='Enter review text')

# Output response
out_response = gradio.components.Textbox(type="text", label='Sentiment')

# Gradio interface to generate UI link
title = "Sentiment Classification"
description = "Analyse sentiment of the given review"

iface = gradio.Interface(fn = predict_sentiment,
                         inputs = [in_prompt],
                         outputs = [out_response],
                         title = title,
                         description = description)

app = gradio.mount_gradio_app(app, iface, path="/")

#iface.launch(server_name = "0.0.0.0", server_port = 8001) # Ref. for parameters: https://www.gradio.app/docs/interface

if __name__ == "__main__":
    # Use this for debugging purposes only 
    import uvicorn
    uvicorn.run(app, host="0.0.0.0", port=8001)