ium_478815/train-mlflow.py

import torch
import torch.nn as nn
import torch.nn.functional as F
import pandas as pd
import numpy as np
import sys
import os
from sklearn.model_selection import train_test_split
from sklearn import preprocessing
from fastai.basics import *
from sklearn.preprocessing import StandardScaler
scaler = StandardScaler()
from sacred import Experiment
from sacred.observers import FileStorageObserver
from sacred.observers import MongoObserver
from sklearn.metrics import accuracy_score
from sacred.observers.file_storage import file_storage_option
from sacred.observers.mongo import mongo_db_option
from sklearn.metrics import mean_squared_error
from sklearn.metrics import mean_absolute_error
import warnings
from sklearn.metrics import r2_score
from sklearn.model_selection import train_test_split
from sklearn.linear_model import ElasticNet
from urllib.parse import urlparse
import mlflow
import mlflow.sklearn
from urllib.parse import urlparse
from mlflow.models import infer_signature

mlflow.set_experiment("478815")
mlflow.set_tracking_uri("http://172.17.0.1:5000")

# Model
class Model(nn.Module):
    def __init__(self):
        super().__init__()
        self.linear = nn.Linear(1,1) 

    def forward(self, x):
        y_predicted = torch.sigmoid(self.linear(x))
        return y_predicted

data = pd.read_csv('data.csv')
data.dropna()

training_data = data.sample(frac=0.9, random_state=25)
testing_data = data.drop(training_data.index)

print(f"No. of training examples: {training_data.shape[0]}")
print(f"No. of testing examples: {testing_data.shape[0]}")

training_data = training_data[['sqft_living', 'price']]

testing_data = testing_data[['sqft_living', 'price']]

training_data[['price']] = training_data[['price']] / 10000000
training_data[['sqft_living']] = training_data[['sqft_living']] / 10000

testing_data[['price']] = testing_data[['price']] / 10000000
testing_data[['sqft_living']] = testing_data[['sqft_living']] / 10000

# Tensory
X_training = training_data[['sqft_living']].to_numpy()
X_testing = testing_data[['sqft_living']].to_numpy()
y_training = training_data[['price']].to_numpy()
y_testing = testing_data[['price']].to_numpy()

torch.from_file
X_training = torch.from_numpy(X_training.astype(np.float32))
X_testing = torch.from_numpy(X_testing.astype(np.float32))
y_training = torch.from_numpy(y_training.astype(np.float32))
y_testing = torch.from_numpy(y_testing.astype(np.float32))

model = Model()
criterion = nn.BCELoss()
optimizer = torch.optim.SGD(model.parameters(), lr=0.01)

epochs=1000

def my_main(epochs):
# Trening
   
    mlflow.log_param("epochs", epochs)

    for epochs in range(epochs):
        y_predicted_train = model(X_training)
        loss = criterion(y_predicted_train,y_training)
        loss.backward()
        optimizer.step()
        optimizer.zero_grad()
        with open ("output.txt",'a+') as f:
            if (epochs%100==0):
                f.write(f'epoch:{epochs+1},loss = {loss.item():.4f}')
        x = X_training.detach().numpy()
        y = y_predicted_train.detach().numpy()

        with torch.no_grad():
            y_predicted = model(X_testing)
            y_predicted_cls = y_predicted.round()
            acc = y_predicted_cls.eq(y_testing).sum()/float(y_testing.shape[0])
            print(f'{acc:.4f}')
            #result = open("output",'w+')
            #result.write(f'{y_predicted}')

            print(np.array(X_testing[0]))

            input_example = np.array(X_testing[0])
            siganture = mlflow.models.signature.infer_signature(np.array(X_training), np.array(y_predicted_train))
            tracking_url_type_store = urlparse(mlflow.get_tracking_uri()).scheme

            if tracking_url_type_store != "file":
                mlflow.pytorch.log_model(model, "model_new", registered_model_name="478815", signature=siganture,
                                    input_example=input_example)
            else:
                mlflow.pytorch.log_model(model, "model_new", signature=siganture, input_example=input_example)
                mlflow.pytorch.save_model(model, "model_new", signature=siganture, input_example=input_example)

        rmse = mean_squared_error(y_testing, y_predicted)
        #print(rmse)

        mae = mean_absolute_error(y_testing, y_predicted)
        #print(mae)

        mlflow.log_metric("rmse", rmse)
        mlflow.log_metric("mae", mae)

        #with open('metrics.txt', 'a+') as f:
        #f.write('Root mean squared error:' + str(rmse) + '\n')  
        #f.write('Mean absolute error:' + str(mae) + '\n')  
        #count = [float(line) for line in f if line]
        #builds = list(range(1, len(count)))
            
        #with open('metric.txt', 'a+') as f:
        #    f.write(str(rmse) + '\n')

        #with open('metric.txt') as file:
         #   y_rmse = [float(line) for line in file if line]
          #  x_builds = list(range(1, len(y_rmse) + 1))
        

with mlflow.start_run() as run:
    my_main(epochs)