Bug - Issue loading `openai-whisper-large-v3-turbo-cuda-gpu`

[jtracey93]

Downloaded model fine and then when trying to load getting the below error:

foundry model load openai-whisper-large-v3-turbo-cuda-gpu --verbose
🕛 Loading model... [11:31:36 ERR] Failed loading model:openai-whisper-large-v3-turbo-cuda-gpu
Exception: Failed: Loading model openai-whisper-large-v3-turbo-cuda-gpu from http://127.0.0.1:50168/openai/load/openai-whisper-large-v3-turbo-cuda-gpu?ttl=600&ep=cuda
Internal Server Error
Failed loading model openai-whisper-large-v3-turbo-cuda-gpu
Error encountered while parsing 'C:\Users\<redacted>\.foundry\cache\models\Microsoft\openai-whisper-large-v3-turbo-cuda-gpu\v2\genai_config.json' JSON Error: model:decoder:outputs: Unknown value "output_cross_qk_names" at line 28 index 62

[bmehta001]

Hi @jtracey93, The Foundry Local CLI does not currently support running voice models. However, you can run voice models with the v2 SDKs.

AudioTranscriptionExample, a sample using whisper-tiny can be found under the samples/cs folder. You can see the code here and follow the compilation instructions for your preferred platform here.