loss mask question

```python
sample.tokens = prompt_tokens_ids + response_token_ids
sample.response_length = len(response_token_ids)
sample.response = response
sample.loss_mask = loss_masks
```

这里loss_masks只需要计算在(system+user_query)之后的response就可以了吗？不需要包括的[0]*len(prompt_tokens_ids)吗？