Training loop

Let's now write the training loop (this step is similar to what was done in previous labs).

Complete the code to implement the training loop:

# Hyperparameters

lr = 0.01

batch_size = 100

num_epoch = 500

hidden_size = 1

num_layers = 2

# Initialize the Vanilla RNN

# Your code here. Aim for 1 line

rnn =

# Initialize the Loss. Please, use torch.nn.BCEWithLogitsLoss

# Your code here. Aim for 1 line

loss =

# Initialize the Optimizer. Please, use torch.optim.Adam (with the lr specified above)

# Your code here. Aim for 1 line

optimizer =

# Training Loop

for epoch in range(num_epoch):

for i in range(0, N, batch_size):

# Read minibatches (for both X and y)

# Your code here. Aim for 2 lines

Xi =

yi =

# Run the model

# Your code here. Aim for 1 line

logits =

# Compute the loss (use l as variable for the loss)

# Your code here. Aim for 1 line

l =

# Update the parameters

# Your code here. Aim for 3 lines

# Print loss

if (epoch + 1) % 50 == 0:

print("Epoch %03d: Train_loss: %.4f " %(epoch+1, l.item()))