Hey all does anyone have an example with | OpenAI | Page 1

#

I have a relevant gist, but it's basically:

#


from transformers import WhisperProcessor, WhisperForConditionalGeneration
from datasets import Audio, load_dataset
import torch

processor = WhisperProcessor.from_pretrained("openai/whisper-large-v2")
model = WhisperForConditionalGeneration.from_pretrained("openai/whisper-large-v2").to("cuda")
forced_decoder_ids = processor.get_decoder_prompt_ids(language="russian", task="translate")


ds = load_dataset("mozilla-foundation/common_voice_11_0", "ru", split="test", streaming=True)
ds = ds.cast_column("audio", Audio(sampling_rate=16_000))
input_speech = next(iter(ds))["audio"]
input_array = torch.from_numpy(input_speech["array"]).cuda()
input_features = processor(input_array, sampling_rate=input_speech["sampling_rate"], return_tensors="pt").input_features

#

CPU works fine, I'm just trying to get it over to GPU before I move over to pyydub etc

#Hey all does anyone have an example with