Has anybody tried to run whisper model in the API?

SambaNova Devs Discussion
1

I am getting a “model not found” error when using “whisper-large-v3”

1 Like
2

Welcome to the community, @janakagrawal1309. Glad to have you here!

Regarding your question:

The model "whisper-large-v3" is currently not available on the SambaNova Cloud platform. If you’re looking for speech-to-text functionality, you might consider using the Qwen2-Audio-7B-Instruct model, which is available on SambaNova Cloud and designed for audio processing tasks.

1 Like
3

@prajwal.balapure any idea when the whisper model will be available?

2 Likes
4

Hi @janakagrawal1309,

As of now, there’s no confirmed timeline for when the Whisper model will be available on the SambaNova Cloud platform.
Our team is always evaluating new models to add based on demand and compatibility.
We’ll be sure to share updates as soon as there’s any news on this

5

@janakagrawal1309

I cannot commit to a date but Whisper is on the near term roadmap .

-Coby

6

@janakagrawal1309 we have soft launched it via the API only. feel free to try and test it out: Audio - SambaNova Documentation

7

@vasanth.mohan still getting a Model not found error.

8

Hi @janakagrawal1309, can you please use Model Id as Whisper-Large-v3 and please try it again.
Thanks

9

Would you give us an example? I’d like to make sure this is a right way to use it. Thank you for the great tech!

headers = {
    "Authorization": f"Bearer {API_KEY}"
}

files = {
    'file': open('sample.m4a', 'rb')  # Change to your actual file path
}

data = {
    "model": "qwen2-audio-7b-instruct",
    "prompt": "Please transcribe carefully, including pauses and hesitations.",
    "response_format": "json",
    "temperature": 0,
    "language": "en"
}

response = requests.post(URL, headers=headers, data=data, files=files)

if response.ok:
    print("Transcription:")
    print(response.json())
else:
    print("Error:")
    print(response.status_code, response.text)

Error:
404 {"error":"Model not found"}
1 Like
10

Hi @yuikita21,
Can you please use "model": "Qwen2-Audio-7B-Instruct" and try it again.

here is the code which works for me:

import requests

def transcribe_audio(audio_file_path, api_key, language="french"):
  headers = {"Authorization": f"Bearer {api_key}"}

  files = {"file": open(audio_file_path, "rb")}

  data = {
      "model": "Qwen2-Audio-7B-Instruct",
      "language": language,
      "response_format": "json",
      "temperature": 0.01,
  }

  response = requests.post(
      "https://api.sambanova.ai/v1/audio/transcriptions",
      headers=headers,
      files=files,
      data=data,
  )

  return response.json()

result = transcribe_audio('/Users/omkarg/output.mp3','SN_api_key')