Remove num_frames in ASR pipeline

[jiqing-feng]

Script:

import torch
from transformers import pipeline
from datasets import load_dataset

device = "cuda:0" if torch.cuda.is_available() else "cpu"
pipe = pipeline(
  "automatic-speech-recognition",
  model="openai/whisper-small",
  device=device,
)
ds = load_dataset("hf-internal-testing/librispeech_asr_dummy", "clean", split="validation")
sample = ds[0]["audio"]
prediction = pipe(sample, batch_size=8)["text"]

error:

Traceback (most recent call last):
  File "/home/jiqing/HuggingFace/tests/workloads/test_whisper.py", line 16, in <module>
    prediction = pipe(sample, batch_size=8)["text"]
                 ^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/home/jiqing/transformers/src/transformers/pipelines/automatic_speech_recognition.py", line 266, in __call__
    return super().__call__(inputs, **kwargs)
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/home/jiqing/transformers/src/transformers/pipelines/base.py", line 1266, in __call__
    return next(
           ^^^^^
  File "/home/jiqing/transformers/src/transformers/pipelines/pt_utils.py", line 126, in __next__
    item = next(self.iterator)
           ^^^^^^^^^^^^^^^^^^^
  File "/home/jiqing/transformers/src/transformers/pipelines/pt_utils.py", line 271, in __next__
    processed = self.infer(next(self.iterator), **self.params)
                           ^^^^^^^^^^^^^^^^^^^
  File "/opt/venv/lib/python3.12/site-packages/torch/utils/data/dataloader.py", line 741, in __next__
    data = self._next_data()
           ^^^^^^^^^^^^^^^^^                                                                                                           File "/opt/venv/lib/python3.12/site-packages/torch/utils/data/dataloader.py", line 801, in _next_data
    data = self._dataset_fetcher.fetch(index)  # may raise StopIteration                                                                        ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/opt/venv/lib/python3.12/site-packages/torch/utils/data/_utils/fetch.py", line 35, in fetch
    data.append(next(self.dataset_iter))
                ^^^^^^^^^^^^^^^^^^^^^^^
  File "/home/jiqing/transformers/src/transformers/pipelines/pt_utils.py", line 188, in __next__                                         processed = next(self.subiterator)
                ^^^^^^^^^^^^^^^^^^^^^^                                                                                                 File "/home/jiqing/transformers/src/transformers/pipelines/automatic_speech_recognition.py", line 486, in preprocess                   extra["num_frames"] = processed.pop("num_frames")
                          ^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "<frozen _collections_abc>", line 945, in pop
  File "/home/jiqing/transformers/src/transformers/feature_extraction_utils.py", line 90, in __getitem__
    return self.data[item]
           ~~~~~~~~~^^^^^^
KeyError: 'num_frames'

This change can fix this error by returning None if there is no num_frames in processors.

[jiqing-feng]

Hi @zucchini-nlp . Would you please review this PR? Thanks!

[zucchini-nlp]

Thanks! The CI is a bit flaky, i'll retrigger and merge

[HuggingFaceDocBuilderDev]

The docs for this PR live here. All of your documentation changes will be reflected on that endpoint. The docs are available until 30 days after the last update.

[jiqing-feng]

Hi @zucchini-nlp . The CI seems ok now.