Clone a loaded object detection model in memory fails

Prerequisites

Please answer the following questions for yourself before submitting an issue.

• [Done] I am using the latest TensorFlow Model Garden release and TensorFlow 2.
• [Done] I am reporting the issue to the correct repository. (Model Garden official or research directory)
• [Done] I checked to make sure that this issue has not been filed already.

1. The entire URL of the file you are using

https://github.com/tensorflow/models.git

2. Describe the bug

Trained my model from ssd_mobilenet_v2_320x320_coco17_tpu-8 as base, saved it to disk with exporter_main_v2.py. After loading the model with detect_fn = tf.saved_model.load(model_dir) I want to clone the python object with tf.keras.models.clone_model(detect_fn). This fails with Expected model argument to be a Model instance, got ', <tensorflow.python.saved_model.load.Loader._recreate_base_user_object.._UserObject object at 0x7f7195138e80>

3. Steps to reproduce

See 2.

4. Expected behavior

Get a in-memory clone of the model

5. Additional context

Expected model argument to be a Model instance, got ', <tensorflow.python.saved_model.load.Loader._recreate_base_user_object.._UserObject object at 0x7f7195138e80>

6. System information

• Ubuntu 18.04
• TensorFlow installed from binary
• TensorFlow version: v2.4.0-rc4-71-g582c8d236cb 2.4.0
• Python version: 3.6.9
• Bazel version (if compiling from source): na
• GCC/Compiler version (if compiling from source): na
• CUDA/cuDNN version: na
• GPU model and memory: AMD