Closed miracseref closed 2 years ago
It's trained with TF2, but uses the TF1 operations. So it definitely runs with TF2 (see requirements.txt, there I used tensorflow==2.4.0).
Hello,
Since it uses TF1 operations I'm not able to save the model as a .h5
or .pkl
file by doing model.save()
.
And I don't know how to load TF1 models, so I wonder if you could help to load this model in my Flask web app?
Thanks
You can directly load the checkpoint file (as its done here), is there a reason why you want to export to h5?
I see, thank you for taking the time. I appreciate it.
The only reason I want to export as an .h5
is that it's easy to manipulate.
Hello Mr. Harald,
First of all, I want to thank you for sharing your work. I have trouble with the saved model. Since it was trained with TensorFlow 1, the format is slightly different from TensorFlow 2 models. I'm building a web app with Flask for deploying this model but I couldn't figure out how to implement the model. Can you help me? Thanks