Fairseq MMS HuggingFace model deployment

Amazon SageMaker
1

I have been trying to deploy the Fairseq MMS model onto AWS Sagemaker and I keep running into error code: 500 when trying to predict the asr for an audio file.

The detailed error:

An error occurred (ModelError) when calling the InvokeEndpoint operation: Received server error (500) from primary with message "{
  "code": 500,
  "type": "InternalServerException",
  "message": "Worker died."
}

My driver code:

# initialize model
huggingface_model = HuggingFaceModel(
    transformers_version='4.26.0',
    pytorch_version='1.13.1',
    py_version='py39',
    env=hub,
    role=role, 
)

# deploy model
predictor = huggingface_model.deploy(
    initial_instance_count=1, # number of instances
    instance_type='ml.m5.xlarge' # ec2 instance type
)
 
# add serializer for audio files   
predictor.serializer = DataSerializer(content_type='audio/x-audio')

audio_path="test.flac"

# read audio file
with open("test.flac", "rb") as f:
    data = f.read()

#transcribe
predictor.predict(data)

The error occurs at predictor.predict(data)

2

Did you resolve this? I have a very similar issue.