Loading the model

[billyAlexio]

Hi, may I ask how can I load your pretrained .model file into Keras properly? I use tensorflow 1.13.1, Keras 2.2.4, and it seems there is a problem loading the model.

I would be very grateful if you can take a look of my problem or tell me your solution!

[tom-doerr]

Could it be that you need to select the correct network architecture for the checkpoints in the model_config?

[billyAlexio]

Thanks for your reply! I don't think network architecture could be the problem, as I simply import keras and use keras.model.load_model() function, there appears the same error.

[tom-doerr]

Could you just remove the decode part? Don't think you need the decode when it's a str.

[billyAlexio]

Yes, and I find a solution by reinstalling another version of h5py package, but when I run play.py another problem happened, as shown in the image. I have looked up for this issue, most answers revolves around the older version of keras should use lr rather than learning_rate as the parameter, however, I have checked the original code, sendex did pass lr as the parameter. And when I run train.py, there is no error, but the trained seems to stuck at waiting state. I am a little confused.

[tom-doerr]

Maybe ChatGPT could help

[billyAlexio]

Yeah maybe, thanks for the tip!