The size of tensor a (41) must match the size of tensor b (13) at non-singleton dimension 3

[luoMonkeyKing]

Does whisper not support multi-threaded concurrent calls? The global variable I use when loading the model: whisper.load_model(name="large", download_root='/root').to(device), initialize the instance in the thread:
torch.hub.load(repo_or_dir='/root/snakers4_silero-vad_master',model='silero_vad',force_reload=False,trust_repo=True,source='local',onnx=False)
But when using multi-trip concurrent call whisper, the following error is reported:
RuntimeError: The size of tensor a (41) must match the size of tensor b (13) at non-singleton dimension 3
If whisper.load_model is loaded in a thread pool, concurrency can be supported, but this is too slow. The model needs to be loaded every time. Is there any better way to support multiple concurrency?
Flask + Python 3.11

[glangford]

The whisper API is not thread safe.

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