AttributeError: 'ConvolutionalBoxPredictor' object has no attribute 'inputs'

Prerequisites

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

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

1. The entire URL of the file you are using

colab notebook
python(3.7)
Tensorflow 2.4.0 (and 2.4.1)

2. Describe the bug

can't train model on tpu by using model_tpu_main.py file

error msg:

" File "/usr/local/lib/python3.7/dist-packages/object_detection/meta_architectures/faster_rcnn_meta_arch.py", line 2947, in updates
self._first_stage_box_predictor.inputs))
AttributeError: 'ConvolutionalBoxPredictor' object has no attribute 'inputs'"

3. Steps to reproduce

run the following line of code:
!python model_tpu_main.py --model_dir=model --pipeline_config_path=model/pipeline_tpu.config --job-dir=model

4. Expected behavior

running the training smoothly