ium_444421/training_mlflow.py

#!/usr/bin/env python
# coding: utf-8

# In[ ]:


import numpy as np
import pandas as pd
from sklearn.metrics import accuracy_score
import torch
from torch import nn, optim
import torch.nn.functional as F
import sys
import mlflow
from urllib.parse import urlparse


# In[ ]:

mlflow.set_tracking_uri("http://172.17.0.1:5000")
mlflow.set_experiment("s444421")


# In[ ]:


epochs = int(sys.argv[1])


# In[ ]:


def prepare_data():
    X_train = pd.read_csv('X_train.csv')
    y_train = pd.read_csv('y_train.csv')
    X_train = torch.from_numpy(np.array(X_train)).float()
    y_train = torch.squeeze(torch.from_numpy(y_train.values).float())
    return X_train, y_train


# In[ ]:


class Net(nn.Module):
    def __init__(self, n_features):
        super(Net, self).__init__()
        self.fc1 = nn.Linear(n_features, 5)
        self.fc2 = nn.Linear(5, 3)
        self.fc3 = nn.Linear(3, 1)
    def forward(self, x):
        x = F.relu(self.fc1(x))
        x = F.relu(self.fc2(x))
        return torch.sigmoid(self.fc3(x))


# In[ ]:


def calculate_accuracy(y_true, y_pred):
    predicted = y_pred.ge(.5).view(-1)
    return (y_true == predicted).sum().float() / len(y_true)


# In[ ]:


def round_tensor(t, decimal_places=3):
    return round(t.item(), decimal_places)

# In[ ]:


def train_model(X_train, y_train, device, epochs):
    net = Net(X_train.shape[1])
    criterion = nn.BCELoss()
    optimizer = optim.Adam(net.parameters(), lr=0.001)
    
    X_train = X_train.to(device)
    y_train = y_train.to(device)
    
    net = net.to(device)
    criterion = criterion.to(device)

    for epoch in range(epochs):
        y_pred = net(X_train)
        y_pred = torch.squeeze(y_pred)
        train_loss = criterion(y_pred, y_train)
        if epoch % 100 == 0:
            train_acc = calculate_accuracy(y_train, y_pred)
            print(
                f'''epoch {epoch}
                Train set - loss: {round_tensor(train_loss)}, accuracy: {round_tensor(train_acc)}
                ''')
    optimizer.zero_grad()
    train_loss.backward()
    optimizer.step()
    return net, round_tensor(train_loss)


# In[ ]:

def my_main(epochs):
    X_train, y_train = prepare_data()
    device = torch.device("cuda:0" if torch.cuda.is_available() else "cpu")
    
    model, loss = train_model(X_train, y_train, device, epochs)
    torch.save(model, 'model.pth')
    mlflow.log_param("epochs", epochs)
    mlflow.log_metric("loss", loss)
    
    X_test = pd.read_csv('X_test.csv')
    X_test = torch.from_numpy(np.array(X_test)).float()
    X_test = X_test.to(device)
    y_pred = model(X_test)
    y_pred = y_pred.ge(.5).view(-1).cpu()
    
    signature = mlflow.models.signature.infer_signature(X_train.numpy(), np.array(y_pred))
    tracking_url_type_store = urlparse(mlflow.get_tracking_uri()).scheme
    
    if tracking_url_type_store != "file":
        mlflow.sklearn.log_model(model, "s444421", registered_model_name="s444421", signature=signature, input_example=X_test.numpy()[:5])
    else:
        mlflow.sklearn.log_model(model, "s444421", signature=signature, input_example=X_test.numpy()[:5])


# In[ ]:

with mlflow.start_run() as run:
    print("MLflow run experiment_id: {0}".format(run.info.experiment_id))
    print("MLflow run artifact_uri: {0}".format(run.info.artifact_uri))
    my_main(epochs)