JSON serialization not available in pyannote.core 5.x

[hbredin]

https://github.com/huggingface/speechbox/blob/82a07082576ecbe9d3f2522e5909aede3b37abd4/src/speechbox/diarize.py#L90

This will break after pyannote.core 5.x branch is released. Source: https://github.com/pyannote/pyannote-core/commit/c89a9cfff103e3aa96c8518c4193a43e493c4790

[sanchit-gandhi]

Thanks for the heads-up @hbredin! 🙌 What's the recommended way to extract a Pythonic version from the diarization output (e.g. as a list or dict)?

[hbredin]

I'd use Annotation.itertracks.

[syffs]

This will break after pyannote.core 5.x branch is released.

@hbredin @sanchit-gandhi 7 months later, this is broken. Any insight to fix this?

[JBJoyce]

I've submitted PR addressing this issue

[2010b9]

Why did they remove the JSON serialization in pyannote.core 5.x?

[abhinavsingh]

Hi,

1) I am following a tutorial on huggingface and ran into this 🙆‍♂️🤦‍♂️ 2) Attempts to use older versions of dependencies simply doesn't work in 2024

Any workaround here?

Following step results into error due to missing for_json

pipeline(sample["audio"].copy())

Below is the exception stacktrace:

{
    "name": "AttributeError",
    "message": "'Annotation' object has no attribute 'for_json'",
    "stack": "---------------------------------------------------------------------------
AttributeError                            Traceback (most recent call last)
Cell In[51], line 1
----> 1 pipeline(sample[\"audio\"].copy())

File ~/.venv/lib/python3.10/site-packages/speechbox/diarize.py:90, in ASRDiarizationPipeline.__call__(self, inputs, group_by_speaker, **kwargs)
     83 inputs, diarizer_inputs = self.preprocess(inputs)
     85 diarization = self.diarization_pipeline(
     86     {\"waveform\": diarizer_inputs, \"sample_rate\": self.sampling_rate},
     87     **kwargs,
     88 )
---> 90 segments = diarization.for_json()[\"content\"]
     92 # diarizer output may contain consecutive segments from the same speaker (e.g. {(0 -> 1, speaker_1), (1 -> 1.5, speaker_1), ...})
     93 # we combine these segments to give overall timestamps for each speaker's turn (e.g. {(0 -> 1.5, speaker_1), ...})
     94 new_segments = []

AttributeError: 'Annotation' object has no attribute 'for_json'"
}

PS: Tutorial also uses for_json in previous steps. I was able to replace that using outputs.itertracks, but this last step usage is within speechbox library itself on huggingface and i don't see any newer release of speechbox in past 6-months.

[abhinavsingh]

Hi,

1. I am following a tutorial on huggingface and ran into this 🙆‍♂️🤦‍♂️
2. Attempts to use older versions of dependencies simply doesn't work in 2024

Any workaround here?

Following step results into error due to missing for_json

pipeline(sample["audio"].copy())

Below is the exception stacktrace:

{
  "name": "AttributeError",
  "message": "'Annotation' object has no attribute 'for_json'",
  "stack": "---------------------------------------------------------------------------
AttributeError                            Traceback (most recent call last)
Cell In[51], line 1
----> 1 pipeline(sample[\"audio\"].copy())

File ~/.venv/lib/python3.10/site-packages/speechbox/diarize.py:90, in ASRDiarizationPipeline.__call__(self, inputs, group_by_speaker, **kwargs)
     83 inputs, diarizer_inputs = self.preprocess(inputs)
     85 diarization = self.diarization_pipeline(
     86     {\"waveform\": diarizer_inputs, \"sample_rate\": self.sampling_rate},
     87     **kwargs,
     88 )
---> 90 segments = diarization.for_json()[\"content\"]
     92 # diarizer output may contain consecutive segments from the same speaker (e.g. {(0 -> 1, speaker_1), (1 -> 1.5, speaker_1), ...})
     93 # we combine these segments to give overall timestamps for each speaker's turn (e.g. {(0 -> 1.5, speaker_1), ...})
     94 new_segments = []

AttributeError: 'Annotation' object has no attribute 'for_json'"
}

PS: Tutorial also uses for_json in previous steps. I was able to replace that using outputs.itertracks, but this last step usage is within speechbox library itself on huggingface and i don't see any newer release of speechbox in past 6-months.

https://github.com/huggingface/speechbox/pull/26/files as per this, installing speechbox from git repo seems to do the job (also as recommended by tutorial, I ignored that and installed from pypi 🤦‍♂️)

Good to close.