Optimize official ssd_mobilenet model but get nothing valuable

I configured the tensorflow/models/research/object_detection project and installed tf-nightly and TF-TRT as directed by GitHub - tensorflow/tensorrt: TensorFlow/TensorRT integration.

However, when trying to perform the sample object detection project, the program is always in the execution state at the Optimize step. After waiting for a long time, there is no more response(except tf 2.0 function warning?). What is going on?

frozen_graph = optimize_model(
    config_path='models/ssd_mobilenet_v1_coco_2018_01_28/pipeline.config', 
    checkpoint_path='models/ssd_mobilenet_v1_coco_2018_01_28/model.ckpt',
    use_trt=True,
    precision_mode='FP16'
)

By the way the version of TensorRT is TensorRT5.1 and the GPU is GTX 1060