Spaces:
Sleeping
Sleeping
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-tiny.en", | |
chunk_length_s=30, | |
device=device, | |
) | |
ds = load_dataset("hf-internal-testing/librispeech_asr_dummy", "clean", split="validation") | |
sample = ds[0]["audio"] | |
prediction = pipe(sample.copy(), batch_size=8)["text"] | |
# we can also return timestamps for the predictions | |
prediction = pipe(sample.copy(), batch_size=8, return_timestamps=True)["chunks"] | |