'Using S3 bucket for SavedModel with tensorflow/serving:2.7.0-gpu Docker image

I'm trying to upgrade the Docker base image that we build on top of, for our TF-serving server. Currently, we use tensorflow/serving:2.3.0-gpu as the base image. I'm now trying out tensorflow/serving:2.7.0-gpu.

When I run the server with the exact same command as before, I get the following error:

FileSystemStoragePathSource encountered a filesystem access error: Could not find base path s3://{s3_bucket}/{model_name}/ for servable {model_name} with error UNIMPLEMENTED: File system scheme 's3' not implemented (file: 's3://{s3_bucket}/{model_name}/')

I want to mention again that the only thing I have changed is the TF-serving base image, and the same thing works with the earlier base image.

My model config file looks as follows:

model_config_list: {
  config: {
     name: "{model_name}",
     base_path: "s3://{s3_bucket}/{model_name}/",
     model_platform: "tensorflow",
     model_version_policy {
        specific {
            versions: {version}
        }
     }
  }
}

I understand that this error is because the TF I/O capabilities have been abstracted out into tensorflow-io. What is the recommended resolution for this? Is S3 no longer supported with TF-serving?

Sources

This article follows the attribution requirements of Stack Overflow and is licensed under CC BY-SA 3.0.

Source: Stack Overflow

Solution	Source

'Using S3 bucket for SavedModel with tensorflow/serving:2.7.0-gpu Docker image

Sources

Related Questions