Can't deploy sagemaker object detection model on DeepLens

I trained a sagemaker model instead and get the following compile error "expecting bytes like object, str found" after compiling.

These are the steps I took:

1. Labeled images in ground truth myself.
2. Trained the model using sagemaker training job gui, with the following options:
   a) Vision Object Detection (Mxnet)
   b) resnet50
   c) image shape 300
3. Training job completed successfully.
4. Imported model from sagmaker in deeplens
     a) selected "Amazon Sagemaker Trained Model"
     b) Selected training job
     c) selected mxnet as framework
5. downloaded project with new model
    a) verified model downloaded and extracted in /opt/awscam/artifacts
    b) mo.optimize('model_algo_1', 300,300,'mx)
    c) compile error "expecting bytes like object, str found"

So here the debugging task is much simplified for your team: a) I have removed any external dependency using built in sagemaker vision model. b) NEO optimization is not part of the workflow - import model directly from training job in deeplens console. c) awscam updated - rebooted

I am completely operating in AWS managed services now with no external dependencies. All software versions are known to AWS and I am using AWS recommended workflow.

Please can someone suggest what is wrong.