Getting ImportError: cannot import name 'FP16_Optimizer' error

I am trying to use apex code with this repo - https://github.com/LuoweiZhou/VLP. I have installed the latest version of apex as per the guidelines here. When I try running the repo, I get below error:

Traceback (most recent call last):
  File "vlp/decode_img2txt.py", line 19, in <module>
    from pytorch_pretrained_bert.tokenization import BertTokenizer, WhitespaceTokenizer
  File "/home/default/ephemeral_drive/work/image_captioning/VLP/pytorch_pretrained_bert/__init__.py", line 6, in <module>
    from .optimization_fp16 import FP16_Optimizer_State
  File "/home/default/ephemeral_drive/work/image_captioning/VLP/pytorch_pretrained_bert/optimization_fp16.py", line 4, in <module>
    from apex.optimizers import FP16_Optimizer
ImportError: cannot import name 'FP16_Optimizer'

The recommended commit of apex to use with repo is this - https://github.com/NVIDIA/apex/tree/1603407bf49c7fc3da74fceb6a6c7b47fece2ef8. The possible cause of the error could be that the optimizer that was used with the repo have been replaced with different ones in the latest version of apex. How can I can replace FP16_Optimizer with a suitable optimizer in the latest version of apex without having major changes in the way model is trained?

Start with vlp/decode_img2txt.py and pytorch_pretrained_bert/optimization_fp16.py, then compare the repository's pinned Apex commit with the installed Apex version. Reproduce the import failure using the reported traceback and determine whether a supported replacement preserves the existing training behavior. Done means the repository imports and runs without this FP16_Optimizer error.