RuntimeError: The size of tensor a (12) must match the size of tensor b (3) at non-singleton dimension 3

[Nitin-deloitte]

I am implementing multithreading on whisper Api. I getting this error RuntimeError: The size of tensor a (6) must match the size of tensor b (3) at non-singleton dimension 3

Can anyone help me please. I am new to this I searched but didn't get anything related to this error.

from concurrent.futures import ThreadPoolExecutor
import numpy as np
def transcribe_chunk(model, file):
    # print(file)
    result = model.transcribe(file)
    print(result)
    return result["text"]

def process_audio(filename):
    transcript = open("transcript.txt", "w+")
    myaudio = AudioSegment.from_file(filename)
    chunks_length_ms = 600000
    chunks = make_chunks(myaudio, chunks_length_ms)
    print(chunks)
    model = whisper.load_model("small")

    with ThreadPoolExecutor() as executor:
        futures = []
        for i, chunk in enumerate(chunks):
            chunk_array = np.array(chunk.get_array_of_samples())
            audio = chunk_array.astype(np.float32)/32768.0
            print("audio", audio)
            future = executor.submit(transcribe_chunk, model, audio)
            futures.append(future)

        for future in futures:
            result = future.result()
            transcript.write(result + ".\n")

process_audio("user_story.mp3")

Error:

RuntimeError                              Traceback (most recent call last)
[<ipython-input-34-a40170bc077f>](https://localhost:8080/#) in <cell line: 1>()
----> 1 process_audio_by_name("user_story.mp3")

17 frames
[/usr/local/lib/python3.10/dist-packages/whisper/model.py](https://localhost:8080/#) in qkv_attention(self, q, k, v, mask)
    102         qk = q @ k
    103         if mask is not None:
--> 104             qk = qk + mask[:n_ctx, :n_ctx]
    105         qk = qk.float()

[glangford]

• Is all API thread-safe? (e.g. create two threads to invoke model.transcribe for two files) #296

"The KV caching mechanism uses forward hooks which are installed in the module objects and will cause race issues when used by multiple threads. The --threads option provides a more low-level control on how the CPU operations are parallelized, but it's less relevant if you're using GPU.

Even if you disabled KV caching, multi-threaded usage will be generally inefficient because of the GIL. "