Does Pipeline.from_pretrained downloads the model each time or uses a cache?

[lsshawn]

When we run Pipeline.from_pretrained using hugging face, does it download the model every time? Or will it read from my cache?

I'm trying to make this code run faster:

pipeline = Pipeline.from_pretrained(
        "pyannote/segmentation",
        use_auth_token="TOKEN",
    )

for wav in wavFiles:
    pipeline(wav)

[hbredin]

Please read the FAQ explaining how to use models and pipelines in an offline manner.