The WhisperPipeline.generate does not support tensor or numpy array

[codingl2k1]

I have to copy the data from the ndarray to Python and then convert the Python float list to a C++ float list. It would be beneficial if the API supported tensor or NumPy array input.

[as-suvorov]

Please provide a sample code

[codingl2k1]

Using ndarray directly performs better, while C++ accesses the buffer of numpy

import openvino_genai
import librosa

def read_wav(filepath):
    raw_speech, samplerate = librosa.load(filepath, sr=16000)
    return raw_speech  # Use the ndarray instead of Python float list

device = "CPU" # GPU can be used as well
pipe = openvino_genai.WhisperPipeline("whisper-base", device)
raw_speech = read_wav("sample.wav")
print(pipe.generate(raw_speech))