mirror of
https://github.com/m-bain/whisperX.git
synced 2025-07-01 18:17:27 -04:00
@ -639,9 +639,10 @@ def cli():
|
|||||||
if vad_filter:
|
if vad_filter:
|
||||||
if hf_token is None:
|
if hf_token is None:
|
||||||
print("Warning, no huggingface token used, needs to be saved in environment variable, otherwise will throw error loading VAD model...")
|
print("Warning, no huggingface token used, needs to be saved in environment variable, otherwise will throw error loading VAD model...")
|
||||||
from pyannote.audio import Inference
|
from pyannote.audio import Inference, Model
|
||||||
vad_pipeline = Inference(
|
vad_pipeline = Inference(
|
||||||
"pyannote/segmentation",
|
Model.from_pretrained("pyannote/segmentation",
|
||||||
|
use_auth_token=hf_token),
|
||||||
pre_aggregation_hook=lambda segmentation: segmentation,
|
pre_aggregation_hook=lambda segmentation: segmentation,
|
||||||
use_auth_token=hf_token,
|
use_auth_token=hf_token,
|
||||||
device=torch.device(device),
|
device=torch.device(device),
|
||||||
|
Reference in New Issue
Block a user