I was reading through the notebook, and I saw that the `extract_pipe_output` function is filtering out elements without a "POSITIVE" label. There's a snippet li

I was reading through the notebook, and I saw that the `extract_pipe_output` function is filtering out elements without a "POSITIVE" label. There's a snippet like this: logits = extract_pipe_output(sentiment_pipe(texts, **sentiment_pipe_kwargs)) rewards = pos_logit_to_reward(logits, task_list) `pos_logit_to_reward` expects `logits` and `task_list` to be the same length, but `logits` is the result of some filtering. I must be missing something about the code and libraries (I'm a PyTorch newb), because I'd otherwise expect this to be a bug.