Wav2Vec2 Transcriber
speechline.transcribers.wav2vec2.Wav2Vec2Transcriber (AudioTranscriber)
Wav2Vec2-CTC model for speech recognition.
Parameters:
| Name | Type | Description | Default |
|---|---|---|---|
model_checkpoint |
str |
HuggingFace model hub checkpoint. |
required |
Source code in speechline/transcribers/wav2vec2.py
class Wav2Vec2Transcriber(AudioTranscriber):
"""
Wav2Vec2-CTC model for speech recognition.
Args:
model_checkpoint (str):
HuggingFace model hub checkpoint.
"""
def __init__(self, model_checkpoint: str) -> None:
super().__init__(model_checkpoint)
def predict(
self,
dataset: Dataset,
chunk_length_s: int = 30,
output_offsets: bool = False,
return_timestamps: str = "word",
keep_whitespace: bool = False,
) -> Union[List[str], List[List[Dict[str, Union[str, float]]]]]:
"""
Performs inference on `dataset`.
Args:
dataset (Dataset):
Dataset to be inferred.
chunk_length_s (int):
Audio chunk length during inference. Defaults to `30`.
output_offsets (bool, optional):
Whether to output timestamps. Defaults to `False`.
return_timestamps (str, optional):
Returned timestamp level. Defaults to `"word"`.
keep_whitespace (bool, optional):
Whether to presere whitespace predictions. Defaults to `False`.
Returns:
Union[List[str], List[List[Dict[str, Union[str, float]]]]]:
Defaults to list of transcriptions.
If `output_offsets` is `True`, return list of offsets.
### Example
```pycon title="example_transcriber_predict.py"
>>> from speechline.transcribers import Wav2Vec2Transcriber
>>> from datasets import Dataset, Audio
>>> transcriber = Wav2Vec2Transcriber("bookbot/wav2vec2-ljspeech-gruut")
>>> dataset = Dataset.from_dict({"audio": ["sample.wav"]}).cast_column(
... "audio", Audio(sampling_rate=transcriber.sr)
... )
>>> transcripts = transcriber.predict(dataset)
>>> transcripts
["ɪ t ɪ z n oʊ t ʌ p"]
>>> offsets = transcriber.predict(dataset, output_offsets=True)
>>> offsets
[
[
{"text": "ɪ", "start_time": 0.0, "end_time": 0.02},
{"text": "t", "start_time": 0.26, "end_time": 0.3},
{"text": "ɪ", "start_time": 0.34, "end_time": 0.36},
{"text": "z", "start_time": 0.42, "end_time": 0.44},
{"text": "n", "start_time": 0.5, "end_time": 0.54},
{"text": "oʊ", "start_time": 0.54, "end_time": 0.58},
{"text": "t", "start_time": 0.58, "end_time": 0.62},
{"text": "ʌ", "start_time": 0.76, "end_time": 0.78},
{"text": "p", "start_time": 0.92, "end_time": 0.94},
]
]
```
"""
return self.inference(
dataset,
chunk_length_s=chunk_length_s,
output_offsets=output_offsets,
offset_key="text",
return_timestamps=return_timestamps,
keep_whitespace=keep_whitespace,
)
predict(self, dataset, chunk_length_s=30, output_offsets=False, return_timestamps='word', keep_whitespace=False)
Performs inference on dataset.
Parameters:
| Name | Type | Description | Default |
|---|---|---|---|
dataset |
Dataset |
Dataset to be inferred. |
required |
chunk_length_s |
int |
Audio chunk length during inference. Defaults to |
30 |
output_offsets |
bool |
Whether to output timestamps. Defaults to |
False |
return_timestamps |
str |
Returned timestamp level. Defaults to |
'word' |
keep_whitespace |
bool |
Whether to presere whitespace predictions. Defaults to |
False |
Returns:
| Type | Description |
|---|---|
Union[List[str], List[List[Dict[str, Union[str, float]]]]] |
Defaults to list of transcriptions.
If |
Example
example_transcriber_predict.py
>>> from speechline.transcribers import Wav2Vec2Transcriber
>>> from datasets import Dataset, Audio
>>> transcriber = Wav2Vec2Transcriber("bookbot/wav2vec2-ljspeech-gruut")
>>> dataset = Dataset.from_dict({"audio": ["sample.wav"]}).cast_column(
... "audio", Audio(sampling_rate=transcriber.sr)
... )
>>> transcripts = transcriber.predict(dataset)
>>> transcripts
["ɪ t ɪ z n oʊ t ʌ p"]
>>> offsets = transcriber.predict(dataset, output_offsets=True)
>>> offsets
[
[
{"text": "ɪ", "start_time": 0.0, "end_time": 0.02},
{"text": "t", "start_time": 0.26, "end_time": 0.3},
{"text": "ɪ", "start_time": 0.34, "end_time": 0.36},
{"text": "z", "start_time": 0.42, "end_time": 0.44},
{"text": "n", "start_time": 0.5, "end_time": 0.54},
{"text": "oʊ", "start_time": 0.54, "end_time": 0.58},
{"text": "t", "start_time": 0.58, "end_time": 0.62},
{"text": "ʌ", "start_time": 0.76, "end_time": 0.78},
{"text": "p", "start_time": 0.92, "end_time": 0.94},
]
]
Source code in speechline/transcribers/wav2vec2.py
def predict(
self,
dataset: Dataset,
chunk_length_s: int = 30,
output_offsets: bool = False,
return_timestamps: str = "word",
keep_whitespace: bool = False,
) -> Union[List[str], List[List[Dict[str, Union[str, float]]]]]:
"""
Performs inference on `dataset`.
Args:
dataset (Dataset):
Dataset to be inferred.
chunk_length_s (int):
Audio chunk length during inference. Defaults to `30`.
output_offsets (bool, optional):
Whether to output timestamps. Defaults to `False`.
return_timestamps (str, optional):
Returned timestamp level. Defaults to `"word"`.
keep_whitespace (bool, optional):
Whether to presere whitespace predictions. Defaults to `False`.
Returns:
Union[List[str], List[List[Dict[str, Union[str, float]]]]]:
Defaults to list of transcriptions.
If `output_offsets` is `True`, return list of offsets.
### Example
```pycon title="example_transcriber_predict.py"
>>> from speechline.transcribers import Wav2Vec2Transcriber
>>> from datasets import Dataset, Audio
>>> transcriber = Wav2Vec2Transcriber("bookbot/wav2vec2-ljspeech-gruut")
>>> dataset = Dataset.from_dict({"audio": ["sample.wav"]}).cast_column(
... "audio", Audio(sampling_rate=transcriber.sr)
... )
>>> transcripts = transcriber.predict(dataset)
>>> transcripts
["ɪ t ɪ z n oʊ t ʌ p"]
>>> offsets = transcriber.predict(dataset, output_offsets=True)
>>> offsets
[
[
{"text": "ɪ", "start_time": 0.0, "end_time": 0.02},
{"text": "t", "start_time": 0.26, "end_time": 0.3},
{"text": "ɪ", "start_time": 0.34, "end_time": 0.36},
{"text": "z", "start_time": 0.42, "end_time": 0.44},
{"text": "n", "start_time": 0.5, "end_time": 0.54},
{"text": "oʊ", "start_time": 0.54, "end_time": 0.58},
{"text": "t", "start_time": 0.58, "end_time": 0.62},
{"text": "ʌ", "start_time": 0.76, "end_time": 0.78},
{"text": "p", "start_time": 0.92, "end_time": 0.94},
]
]
```
"""
return self.inference(
dataset,
chunk_length_s=chunk_length_s,
output_offsets=output_offsets,
offset_key="text",
return_timestamps=return_timestamps,
keep_whitespace=keep_whitespace,
)