Closed jaketae closed 2 years ago
@fchollet for this issue, I think keras.application is expect to export the symbol in resnet_common.py which is aligned with tf.keras.application.
@qlzh727 Thanks for the input. Unfortunately, I’m not sure if I quite follow what you mean. Is resnet50.py still exported via TensorFlow? From what I see, TF nightly now has its own implementation of resnet independent of Keras.
Summary
I'm wondering if it is necessary to keep resnet50.py given that resnet_common.py already encapsulates all variants of ResNet, including ResNet50. They seem to be independent implementations of the same model? Please let me know if I'm missing something. Thanks in advance.
Environment
Leaving fields blank as the issue is irrelevant to build or runtime environment.
Logs or source codes for reproduction