# A simple Linear Regression example for Celsius to Fahrenheit conversion with TensorFlow

import tensorflow as tf
import numpy as np
import streamlit as st
import matplotlib.pyplot as plt

# Streamlit UI
st.title('Celsius to Fahrenheit Conversion with TensorFlow')

# Define the model
model = tf.keras.Sequential([
    tf.keras.layers.Dense(units=1, input_shape=[1])
])

# Compile the model with an optimizer and loss function
model.compile(optimizer='sgd', loss='mse')

# Training data (Celsius to Fahrenheit)
celsius = np.array([-40, -10, 0, 8, 15, 22, 38], dtype=float)  # Celsius
fahrenheit = np.array([-40, 14, 32, 46.4, 59, 71.6, 100.4], dtype=float)  # Corresponding Fahrenheit

# Display example input and output
st.write("Example Celsius values (input):", celsius)
st.write("Corresponding Fahrenheit values (output):", fahrenheit)

# User input for the Celsius value to predict Fahrenheit
input_celsius = st.number_input('Enter Celsius value:', value=0.0, format="%.1f")

# User input for epochs
epochs = st.sidebar.slider("Number of epochs", 10, 500, 10)

# Button to train the model and make prediction
if st.button('Train Model and Predict Fahrenheit'):
    with st.spinner('Training...'):
        model.fit(celsius, fahrenheit, epochs=epochs)
    st.success('Training completed!')

    # Make prediction
    predicted_fahrenheit = model.predict([input_celsius])
    st.write(f'For input of {input_celsius}°C, the predicted Fahrenheit value is {predicted_fahrenheit[0][0]:.1f}°F')

    # Predictions for visualization
    predictions = model.predict(celsius)

    # Plotting
    plt.scatter(celsius, fahrenheit, label='Actual Conversion')
    plt.plot(celsius, predictions, color='red', label='Predicted Conversion')
    plt.xlabel('Celsius')
    plt.ylabel('Fahrenheit')
    plt.legend()
    st.pyplot(plt)