import sys
import pandas as pd
import numpy as np
from tensorflow import keras
from sklearn.metrics import accuracy_score
from tensorflow.keras import layers
import mlflow

def train(epochs, batch_size):
    model_name = "model"
    train = pd.read_csv('train.csv', header=None, skiprows=1)
    indexNames = train[train[1] == 2].index
    train.drop(indexNames, inplace=True)
    cols = [0, 2, 3]
    X = train[cols].to_numpy()
    y = train[1].to_numpy()
    X = np.asarray(X).astype('float32')

    model = keras.Sequential(name="winner")
    model.add(keras.Input(shape=(3), name="game_info"))
    model.add(layers.Dense(4, activation="relu", name="layer1"))
    model.add(layers.Dense(8, activation="relu", name="layer2"))
    model.add(layers.Dense(8, activation="relu", name="layer3"))
    model.add(layers.Dense(5, activation="relu", name="layer4"))
    model.add(layers.Dense(1, activation="relu", name="output"))

    model.compile(
        optimizer=keras.optimizers.Adam(),
        loss=keras.losses.MeanSquaredError(),
    )

    history = model.fit(
        X,
        y,
        batch_size=batch_size,
        epochs=epochs, )

    model.save(model_name)

    test = pd.read_csv('test.csv', header=None, skiprows=1)
    cols = [0, 2, 3]
    indexNames = test[test[1] == 2].index
    test.drop(indexNames, inplace=True)
    X_test = test[cols].to_numpy()
    y_test = test[1].to_numpy()
    X_test = np.asarray(X_test).astype('float32')

    predictions = model.predict(X_test)
    pd.DataFrame(predictions).to_csv('results.csv', sep='\t', index=False, header=False)

    acc = accuracy_score(y_test, predictions)
    print('Accuracy: ', acc)
    return acc, model


epochs = int(sys.argv[1]) if len(sys.argv) > 1 else 15
batch_size = int(sys.argv[2]) if len(sys.argv) > 2 else 16

with mlflow.start_run():
    acc, model = train(epochs, batch_size)
    mlflow.log_param("epochs", epochs)
    mlflow.log_param("batch_size", batch_size)
    mlflow.log_metric("accuracy", acc)
    mlflow.keras.log_model(model, 'model')