[Object Detection API] Object Detection API with Multi-Worker Mirrored Strategy training

System information

• What is the top-level directory of the model you are using: models/research/object-detection
• Have I written custom code (as opposed to using a stock example script provided in TensorFlow):
• OS Platform and Distribution (e.g., Linux Ubuntu 16.04): Ubuntu 16.04
• TensorFlow installed from (source or binary):
• TensorFlow version (use command below): 1.14
• Bazel version (if compiling from source):
• CUDA/cuDNN version: cuda 10.1 / cudnn 7.6.2
• GPU model and memory: RTX 2080ti - 11GB, P100 - 16GB
• Exact command to reproduce:

Describe the problem

Hello. I 'm looking to speed-up Object-Detection API by using 2(or more) individual multi-gpu machines. (machine 1 has 8*2080ti gpus, machine 2 has 2 * p100 gpus). I've tested 1-machine, multi-gpu training with both legacy/train.py and model_main.py. Now I have no idea of using both machines for training.

Q1) Does object-detection api support Multi-Worker Mirrored Strategy with Tensorflow 1.14?
(with model_main.py)
Q2) Should I upgrade tensorflow version up to 2.0 to use Multi-Worker Mirrored Strategy?

Q3) Or, should I use legacy/train.py for multi-machine training?
If so, are there any tutorial or idea for multi-machine training?

Thanks..

Source code / logs

Include any logs or source code that would be helpful to diagnose the problem. If including tracebacks, please include the full traceback. Large logs and files should be attached. Try to provide a reproducible test case that is the bare minimum necessary to generate the problem.