Inference with Tensorflow-gpu 1.15.4

[Equationliu]

I got into trouble when test ResNet/inference-example.py and EfficientNet/eval_ckpt_main.py.

It seems that there are missing args to define a ResNetModel (like group, activation_name etc), but after I added them, it still raise:

ValueError: The custom_gradient decorator currently supports keywords arguments only when eager execution is enabled.

As for EfficientNet/eval_ckpt_main.py, it raise:

ModuleNotFoundError: No module named 'condconv'

and there is a similiar issue https://github.com/qubvel/efficientnet/issues/79