"unable to evaluate payload provided" invoking SageMaker Factorization Machines endpoint using AWS CLI

[fnavarro94]

Hi,

I have trained and deployed an inference model using Amazon SageMaker's Factorization Machines algorithm. However, when I try to call the inference model using the AWS CLI command below, I encounter an error message. I have reviewed the documentation, and my input JSON format seems correct. I would appreciate help figuring out what is going wrong.

I trained the model using the MovieLens 1 million ratings dataset. The training data is a sparse matrix with 9992 features. The first 6040 columns correspond to users, and the next 3952 columns correspond to movies. Each row has only two non-zero entries, one corresponding to the user and one to the movie.

Here's the AWS CLI command I used to invoke the endpoint:

aws sagemaker-runtime invoke-endpoint ^
  --endpoint-name factorization-machines-2023-04-03-17-20-00-090 ^
  --body "{\"instances\": [{\"data\": {\"features\": {\"keys\": [1, 8000], \"shape\": [2], \"values\": [1, 1]}}}]}" ^
  --content-type application/json ^
  --accept application/json ^
  results

Error message:

An error occurred (ModelError) when calling the InvokeEndpoint operation: Received client error (400) from primary with message "unable to evaluate payload provided". See https://us-east-1.console.aws.amazon.com/cloudwatch/home?region=us-east-1#logEventViewer:group=/aws/sagemaker/Endpoints/*********** in account ******************* for more information.