Notes

I apologize for the confusion. If you're getting a `KeyError: 'input_question'`, it means that the dictionary returned by the `collate_fn` doesn't have the key `'input_question'`. This might be because you're trying to access `'input_question'` key within your training loop.

Since we're using a different key `'combined_sequence'` in the `collate_fn`, you need to access that key instead. Here's how you should modify your training loop:

```python
# Training loop
num_epochs = 10
for epoch in range(num_epochs):
model.train()
total_loss = 0.0

for batch in tqdm(train_loader, desc=f"Epoch {epoch + 1}"):
combined_sequence = batch['combined_sequence'].to(device) # Use 'combined_sequence' key

optimizer.zero_grad()
output = model(combined_sequence, combined_sequence) # Use the same sequence for both input and target
output_dim = output.shape[-1]
output = output.contiguous().view(-1, output_dim)
target_sequence = combined_sequence.view(-1) # Flatten the target

loss = criterion(output, target_sequence)
loss.backward()
optimizer.step()

total_loss += loss.item()

avg_loss = total_loss / len(train_loader)
print(f"Epoch {epoch + 1}, Loss: {avg_loss:.4f}")
```

Make sure to use the key `'combined_sequence'` in your training loop instead of `'input_question'`.