Sagemaker Neo compilation failed for TF Lite object detection model

aws / amazon-sagemaker-examples

Example 📓 Jupyter notebooks that demonstrate how to build, train, and deploy machine learning models using 🧠 Amazon SageMaker.

https://sagemaker-examples.readthedocs.io

Apache License 2.0

9.99k stars 6.74k forks source link

Sagemaker Neo compilation failed for TF Lite object detection model #1464

Open tmnasvo opened 4 years ago

tmnasvo commented 4 years ago

I am trying to compile a TFLite object detection model for imx8qm using Sagemaker Neo. I supplied a .tflite file and specified the input config as {"input":[1, 300, 300, 3]} But the compilation failed with the following error: ClientError: InputConfiguration: Framework cannot load TFLite model. Unable to infer tensor data type for all inputs/outputs: Please specify all input layers in data_shape.

How to fix this?

Alvark95 commented 4 years ago

I'm having the same issue. Could you solve it?

minlu1021 commented 4 years ago

Hi @tmnasvo,

Since we may need your model file to further troubleshoot the issue. Could you open a ticket through AWS support if possible? Thank you!

sbhavani commented 3 years ago

Same issue. Tested with https://tfhub.dev/google/lite-model/aiy/vision/classifier/plants_V1/3