temp.py

import torch
import torch.nn as nn
import numpy as np
import time
import math
from matplotlib import pyplot
import pandas as pd
from sklearn.preprocessing import MinMaxScaler

torch.manual_seed(0)
np.random.seed(0)


input_window = 100 
output_window = 1 
block_len = input_window + output_window 
batch_size = 10
train_size = 0.8
device = torch.device("cuda" if torch.cuda.is_available() else "cpu")

class PositionalEncoding(nn.Module):

    def __init__(self, d_model, max_len=5000):
        super(PositionalEncoding, self).__init__()       
        pe = torch.zeros(max_len, d_model)
        position = torch.arange(0, max_len, dtype=torch.float).unsqueeze(1)
        div_term = 1 / (10000 ** ((2 * np.arange(d_model)) / d_model))
        pe[:, 0::2] = torch.sin(position * div_term[0::2])
        pe[:, 1::2] = torch.cos(position * div_term[1::2])

        pe = pe.unsqueeze(0).transpose(0, 1) 
        self.register_buffer('pe', pe)

    def forward(self, x):
        return x + self.pe[:x.size(0), :].repeat(1,x.shape[1],1)
          

class JV2former(nn.Module):
    def __init__(self,feature_size=250,num_layers=1,dropout=0.1):
        super(JV2former, self).__init__()
        self.model_type = 'Transformer'
        self.input_embedding  = nn.Linear(1,feature_size)
        self.src_mask = None

        self.pos_encoder = PositionalEncoding(feature_size)
        self.encoder_layer = nn.TransformerEncoderLayer(d_model=feature_size, nhead=10, dropout=dropout)
        self.transformer_encoder = nn.TransformerEncoder(self.encoder_layer, num_layers=num_layers)
        self.decoder = nn.Linear(feature_size,1)
        self.init_weights()

    def init_weights(self):
        initrange = 0.1    
        self.decoder.bias.data.zero_()
        self.decoder.weight.data.uniform_(-initrange, initrange)

    def forward(self,src):
        if self.src_mask is None or self.src_mask.size(0) != len(src):
            device = src.device
            mask = self._generate_square_subsequent_mask(len(src)).to(device)
            self.src_mask = mask

        src = self.input_embedding(src) 
        src = self.pos_encoder(src)
        output = self.transformer_encoder(src,self.src_mask)
        output = self.decoder(output)
        return output

    def _generate_square_subsequent_mask(self, sz):
        mask = (torch.triu(torch.ones(sz, sz)) == 1).transpose(0, 1)
        mask = mask.float().masked_fill(mask == 0, float('-inf')).masked_fill(mask == 1, float(0.0))
        return mask

def create_inout_sequences(input_data, input_window ,output_window):
    inout_seq = []
    L = len(input_data)
    block_num =  L - block_len + 1

    for i in range( block_num ):
        train_seq = input_data[i : i + input_window]
        train_label = input_data[i + output_window : i + input_window + output_window]
        inout_seq.append((train_seq ,train_label))

    return torch.FloatTensor(np.array(inout_seq))

def get_data(csv_file, train_size, input_window, output_window, device):
    data = pd.read_csv(csv_file)

    data['DateTime'] = pd.to_datetime(data['DateTime'])

    scaler = MinMaxScaler(feature_range=(-1, 1)) 
    data['Vehicles'] = scaler.fit_transform(data['Vehicles'].values.reshape(-1, 1)).reshape(-1)

    sampels = int(len(data) * train_size)
    train_data = data['Vehicles'][:sampels]
    test_data = data['Vehicles'][sampels:]

    train_sequence = create_inout_sequences(train_data, input_window, output_window)
    test_sequence = create_inout_sequences(test_data, input_window, output_window)

    train_sequence = torch.FloatTensor(train_sequence).to(device)
    test_sequence = torch.FloatTensor(test_sequence).to(device)

    return train_sequence, test_sequence


def get_batch(input_data, i , batch_size):

    batch_len = min(batch_size, len(input_data) - i)
    data = input_data[ i:i + batch_len ]
    input = torch.stack([item[0] for item in data]).view((input_window,batch_len,1))
    target = torch.stack([item[1] for item in data]).view((input_window,batch_len,1))
    return input, target

def train(train_data):
    model.train() 
    total_loss = 0.
    start_time = time.time()

    for batch, i in enumerate(range(0, len(train_data), batch_size)):
        data, targets = get_batch(train_data, i , batch_size)
        optimizer.zero_grad()
        output = model(data)
        loss = criterion(output, targets)
        loss.backward()
        torch.nn.utils.clip_grad_norm_(model.parameters(), 0.7)
        optimizer.step()

        total_loss += loss.item()
        log_interval = int(len(train_data) / batch_size / 5)
        if batch % log_interval == 0 and batch > 0:
            cur_loss = total_loss / log_interval
            elapsed = time.time() - start_time
            print('| epoch {:3d} | {:5d}/{:5d} batches | '
                  'lr {:02.6f} | {:5.2f} ms | '
                  'loss {:5.5f} | ppl {:8.2f}'.format(
                    epoch, batch, len(train_data) // batch_size, scheduler.get_lr()[0],
                    elapsed * 1000 / log_interval,
                    cur_loss, math.exp(cur_loss)))
            total_loss = 0
            start_time = time.time()

def plot_and_loss(eval_model, data_source,epoch):
    eval_model.eval() 
    total_loss = 0.
    test_result = torch.Tensor(0)    
    truth = torch.Tensor(0)
    with torch.no_grad():
        for i in range(len(data_source)):  
            data, target = get_batch(data_source, i , 1) 
            output = eval_model(data)            
            total_loss += criterion(output, target).item()
            test_result = torch.cat((test_result, output[-1].view(-1).cpu()), 0)
            truth = torch.cat((truth, target[-1].view(-1).cpu()), 0)
            
    len(test_result)

    pyplot.plot(test_result,color="red")
    pyplot.plot(truth[:500],color="blue")
    pyplot.plot(test_result-truth,color="green")
    pyplot.grid(True, which='both')
    pyplot.axhline(y=0, color='k')
    pyplot.close()
    return total_loss / i

 
def predict_future(eval_model, data_source,steps):
    eval_model.eval() 
    total_loss = 0.
    test_result = torch.Tensor(0)    
    truth = torch.Tensor(0)
    data, _ = get_batch(data_source , 0 , 1)
    with torch.no_grad():
        for i in range(0, steps):            
            output = eval_model(data[-input_window:])

            data = torch.cat((data, output[-1:])) 

    data = data.cpu().view(-1)

    pyplot.plot(data,color="red")       
    pyplot.plot(data[:input_window],color="blue")    
    pyplot.grid(True, which='both')
    pyplot.axhline(y=0, color='k')

    pyplot.show()
    pyplot.close()
        

def evaluate(eval_model, data_source):
    eval_model.eval() 
    total_loss = 0.
    eval_batch_size = 1000
    with torch.no_grad():
        for i in range(0, len(data_source), eval_batch_size):
            data, targets = get_batch(data_source, i,eval_batch_size)
            output = eval_model(data)            
            total_loss += len(data[0]) * criterion(output, targets).cpu().item()
    return total_loss / len(data_source)

train_data, val_data = get_data("/kaggle/input/vehicle-count/traffic.csv", train_size, input_window, output_window, device)
model = JV2former().to(device)

criterion = nn.L1Loss()
lr = 0.005 
optimizer = torch.optim.AdamW(model.parameters(), lr=lr)
scheduler = torch.optim.lr_scheduler.StepLR(optimizer, 1, gamma=0.95)

best_val_loss = float("inf")
epochs = 10 
best_model = None

for epoch in range(1, epochs + 1):
    epoch_start_time = time.time()
    train(train_data)
    if ( epoch % 5 == 0 ):
        val_loss = plot_and_loss(model, val_data,epoch)

    else:
        val_loss = evaluate(model, val_data)
   
    print('-' * 89)
    print('| end of epoch {:3d} | time: {:5.2f}s | valid loss {:5.5f} | valid ppl {:8.2f}'.format(epoch, (time.time() - epoch_start_time),
                                     val_loss, math.exp(val_loss)))
    print('-' * 89)


    scheduler.step() 

def print_model_summary(model):
  """Prints a summary of the JV2former model architecture.

  Args:
    model: The JV2former model to summarize.
  """
  print("JV2former Model Summary:")
  print("Input Features:", model.input_embedding.in_features)  # Input feature dimension
  print("Positional Encoding:", model.pos_encoder.__class__.__name__)  # Positional encoding layer type
  print("Transformer Encoder:")
  print("  - Layers:", model.transformer_encoder.num_layers)  # Number of encoder layers
  print("  - Heads:", model.transformer_encoder.layer[0].nhead)  # Number of heads in each encoder layer
  print("  - d_model:", model.transformer_encoder.layer[0].d_model)  # Dimension of the model (d_model)
  print("Decoder:", model.decoder.in_features, "->", model.decoder.out_features)  # Decoder input and output dimensions
  print("Total Trainable Parameters:", sum(p.numel() for p in model.parameters() if p.requires_grad))

# Example usage
model = JV2former().to(device)
print_model_summary(model)