How to perform backpropagation through time?

I am using an LSTM for a sequential classification task. I am attempting to backpropagate through time (non-truncated) and update the parameters for each file in my training set. However, I am a bit confused on how to implement BPTT.

Here is a snippet of my code.

    for file in file_list:
        # Clear existing gradients

        # Load features and labels for current file
        x_batch, y_batch = read_feat_file(file, conf_dict)

        # Normalize features
        x_batch = scaler.transform(x_batch)

        # Encode labels as integers
        y_batch = le.transform(y_batch).astype('long')

        # Move to GPU
        x_batch = (torch.from_numpy(x_batch)).to(device)
        y_batch = (torch.from_numpy(y_batch)).to(device)

        # Get outputs
        train_outputs = model(x_batch)

        # Calculate loss
        loss = F.nll_loss(train_outputs, y_batch, reduction='sum')

        # Backpropagate and update weights

However, based on others posts such as this one, it looks like the correct way to perform non-truncated BPTT is to pass the output at the t-1 time step as the input to the model at time t and then call .backward() at the end of the sequence.

My question: As written, is my code actually performing backpropagation through time?