I’m having trouble on saving/loading Tensorflow models.
The same code works on Google colab but not for submission environment.
I used tf.saved_model.save(model, {path}) and models.load_model({path}). It should create a folder named {path} containing the model files but the following error is resulted instead.
Logs
OSError: SavedModel file does not exist at: trained_model/{saved_model.pbtxt|saved_model.pb}
So I went into your model and I can’t see your saved object uploaded. I can see that you are using zip submissions so you should have been able to upload everything.
Can you confirm that you are actually uploading a saved model file somewhere inside your zip folder?
Also could you downgrade to tensorflow=2.2 for now?
@alfarzan Thanks for the reminder! After I reuploaded the saved model file and downgraded tensorflow, it now gives a different error message:
This function does not handle the case of the path where all inputs are not already EagerTensors.
Again, I was able to load the model just fine on Google Colab. I have to download the model and submit using zip as the python notebook failed to copy the model when running %aicrowd_submit.