import tensorflow as tf
from sklearn.model_selection import train_test_split
from sklearn.preprocessing import MinMaxScaler, OneHotEncoder
from keras.models import Sequential
from keras.layers import Dense, Dropout
from keras import regularizers
import numpy as np
import pandas as pd
from datetime import datetime
from importlib.metadata import version

from sacred import Experiment
from sacred.observers import FileStorageObserver

ex = Experiment("464903", interactive=True)
ex.observers.append(FileStorageObserver('my_runs'))

@ex.config
def my_config():
    num_epochs = 200
    dropout_layer_value = 0.3  

@ex.capture
def prepare_info(num_epochs, dropout_layer_value, _run):
    _run.info["num_epochs"] = num_epochs
    _run.info["dropout_layer_value"] = dropout_layer_value
    _run.info["training_ts"] = datetime.now()

@ex.main
def train_and_evaluate(num_epochs, dropout_layer_value, _run):
    prepare_info()

    ex.open_resource('./lettuce_dataset_updated.csv', "r")

    dataset = pd.read_csv('./lettuce_dataset_updated.csv', encoding="ISO-8859-1")

    print(version('tensorflow')) 
    print(version('scikit-learn'))
    print(version('keras'))
    print(version('pandas'))

    ph_level = dataset['pH Level'].values.tolist()
    temp_F = dataset['Temperature (F)'].values.tolist()
    humid = dataset['Humidity'].values.tolist()
    days = dataset['Growth Days'].values.tolist()
    plant_id = dataset['Plant_ID'].values.tolist()

    X = []
    Y = []

    id = plant_id[0]
    temp_sum = 0
    humid_sum = 0
    ph_level_sum = 0
    day = 1

    for i in range(0, len(plant_id)):
        if plant_id[i] == id:
            temp_sum += temp_F[i]
            humid_sum += humid[i]
            ph_level_sum += ph_level[i]
            day = days[i]
        else:
            temp = []
            temp.append(temp_sum/day)
            temp.append(humid_sum/day)
            temp.append(ph_level_sum/day)
            X.append(temp)
            Y.append(day)
            temp_sum = 0
            humid_sum = 0
            ph_level_sum = 0
            day = 1
            id = plant_id[i]

    scaler = MinMaxScaler()
    X = scaler.fit_transform(X)

    X = np.array(X)
    Y = np.array(Y)

    encoder = OneHotEncoder(sparse=False)
    y_onehot = encoder.fit_transform(Y.reshape(-1,1))

    X_train, X_test, y_train, y_test = train_test_split(X, y_onehot, test_size=0.4, random_state=42)

    model = Sequential([
        Dense(8, activation='relu', input_dim=3, kernel_regularizer=regularizers.l2(0.04)),
        Dropout(dropout_layer_value),
        Dense(8, activation='relu', kernel_regularizer=regularizers.l2(0.04)),
        Dropout(dropout_layer_value),
        Dense(4, activation='softmax', kernel_regularizer=regularizers.l2(0.04)),
    ])

    model.compile(optimizer='sgd',
                loss='categorical_crossentropy',
                metrics=['accuracy'])

    history = model.fit(X_train, y_train, epochs=num_epochs, validation_data=(X_test, y_test), verbose=2)

    test_loss, test_accuracy = model.evaluate(X_test, y_test, verbose=2)

    print(history.history['val_accuracy'])

    _run.log_scalar("train loss", history.history['loss'])
    _run.log_scalar("train accuracy", history.history['accuracy'])
    _run.log_scalar("test loss", test_loss)
    _run.log_scalar("test accuracy", test_accuracy)

    print(f"Dokładność testowa: {test_accuracy:.2%}")

    model.evaluate(X_test, y_test)[1]

    model.save('./model.keras') 
    ex.add_artifact("./model.keras")

ex.run()
print(my_config())