ium_444354/pytorch/pytorch.py

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

# In[18]:


import torch
import jovian
import torchvision
import matplotlib
import torch.nn as nn
import pandas as pd
import matplotlib.pyplot as plt
import seaborn as sns
import torch.nn.functional as F
from torchvision.datasets.utils import download_url
from torch.utils.data import DataLoader, TensorDataset, random_split
import random
import os
import sys
from sacred import Experiment
from sacred.observers import FileStorageObserver
from sacred.observers import MongoObserver
import mlflow
import mlflow.pytorch
from mlflow.models import infer_signature
from urllib.parse import urlparse

# In[2]:
ex = Experiment(save_git_info=False)

ex.observers.append(FileStorageObserver('s444354_runs'))

# ex.observers.append(MongoObserver(url='mongodb://mongo_user:mongo_password_IUM_2021@localhost:27017', db_name='sacred'))
ex.observers.append(MongoObserver(url='mongodb://admin:IUM_2021@172.17.0.1:27017', db_name='sacred'))

#mlflow.set_tracking_uri("http://172.17.0.1:5000")
mlflow.set_experiment('s444354')

try: 
    numberOfEpochParam = int(sys.argv[1])
except:
    numberOfEpochParam = 1500




dataframe_raw = pd.read_csv("winequality-red.csv")
dataframe_raw.head()


# In[3]:


input_cols=list(dataframe_raw.columns)[:-1]
output_cols = ['quality']
input_cols,output_cols


# In[4]:


def dataframe_to_arrays(dataframe):
    dataframe1 = dataframe_raw.copy(deep=True)
    inputs_array = dataframe1[input_cols].to_numpy()
    targets_array = dataframe1[output_cols].to_numpy()
    return inputs_array, targets_array

inputs_array, targets_array = dataframe_to_arrays(dataframe_raw)
inputs_array, targets_array


# In[5]:


inputs = torch.from_numpy(inputs_array).type(torch.float)
targets = torch.from_numpy(targets_array).type(torch.float)
inputs,targets


# In[6]:


dataset = TensorDataset(inputs, targets)
dataset


# In[7]:


train_ds, val_ds = random_split(dataset, [1300, 299])
batch_size=50
train_loader = DataLoader(train_ds, batch_size, shuffle=True)
val_loader = DataLoader(val_ds, batch_size)
lr = 1e-6

# In[8]:


class WineQuality(nn.Module):
    def __init__(self):
        super().__init__()
        self.linear = nn.Linear(input_size,output_size) 
        
    def forward(self, xb): 
        out = self.linear(xb)
        return out
    
    def training_step(self, batch):
        inputs, targets = batch 
        # Generate predictions
        out = self(inputs)          
        # Calcuate loss
        loss = F.l1_loss(out,targets) 
        return loss
    
    def validation_step(self, batch):
        inputs, targets = batch
        # Generate predictions
        out = self(inputs)
        # Calculate loss
        loss = F.l1_loss(out,targets)   
        return {'val_loss': loss.detach()}
        
    def validation_epoch_end(self, outputs):
        batch_losses = [x['val_loss'] for x in outputs]
        epoch_loss = torch.stack(batch_losses).mean() 
        return {'val_loss': epoch_loss.item()}
    
    def epoch_end(self, epoch, result, num_epochs):
        # Print result every 100th epoch
        if (epoch+1) % 100 == 0 or epoch == num_epochs-1:
            print("Epoch [{}], val_loss: {:.4f}".format(epoch+1, result['val_loss']))


# In[9]:


input_size = len(input_cols)
output_size = len(output_cols)


# In[10]:
@ex.config
def my_config():
    #epochs = numberOfEpochParam
    #epoki pobrane albo z CLI (try/catch), a jak nie przejdzie to ustawione w ex.config
    epochs = numberOfEpochParam
    lr=lr
    model=model 
    train_loader=train_loader
    val_loader=val_loader

model=WineQuality()


# In[11]:


def evaluate(model, val_loader):
    outputs = [model.validation_step(batch) for batch in val_loader]
    return model.validation_epoch_end(outputs)
    
@ex.capture
def fit(epochs, lr, model, train_loader, val_loader, _log, _run, opt_func=torch.optim.SGD):
    _log.info("log info test ")
    epochs=epochs
    history = []
    optimizer = opt_func(model.parameters(), lr)
    with mlflow.start_run() as run:
    
        for epoch in range(epochs):
            for batch in train_loader:
                loss = model.training_step(batch)
                loss.backward()
                optimizer.step()
                optimizer.zero_grad()
            result = evaluate(model, val_loader)
            model.epoch_end(epoch, result, epochs)
            history.append(result)
            
        mlflow.log_param('epochs', epochs)
        mlflow.log_param('lr', lr)
    
    torch.save(model, 'saved_model.pth')
    ex.add_artifact("saved_model.pth")
    
    _run.info["epochs"] = epochs
    
    #signature = mlflow.models.signature.infer_signature(inputs_array)
  
    #tracking_url_type_store = urlparse(mlflow.get_tracking_uri()).scheme
  
    input_example = [0.1, 0.1, 546.0, 555.223, 1., 1., 33.16376, 84.12426]
  
    #if tracking_url_type_store != "file":
        #mlflow.pytorch.log_model(model, "model", registered_model_name="s444354", signature=signature, input_example=input_example)
    #else:
        #mlflow.pytorch.log_model(model, "model", signature=signature, input_example=input_example)
        #mlflow.pytorch.save_model(model, "my_model", signature=signature, input_example=input_example)
    
    
  
    return history



def predict_single(input, target, model):
    inputs = input.unsqueeze(0)
    predictions = model(inputs)
    prediction = predictions[0].detach()

    return "Target: "+str(target)+"-----   Prediction: "+str(prediction)+"\n"


# In[32]:


#wylosuj 10 próbek predykcji
for i in random.sample(range(0, len(val_ds)), 10):
    input_, target = val_ds[i]
    print(predict_single(input_, target, model),end="")
    


# In[36]:


with open("result.txt", "w+") as file:
    for i in range(0, len(val_ds), 1):
        input_, target = val_ds[i]
        file.write(str(predict_single(input_, target, model)))


@ex.automain
def main():
    #my_config()
    #print("number of epochs is: ", epochs)
    history5 = fit()
    
   

#ex.run()