Streamlit app failes to load weights

Deployment
, ,
1

Hello!

This is my first app in streamlit. I trained a model to segment images. I used keras 2 to save the weights of model. Then, I achieved deploy my model in streamlit in my local computer (without errors).

But, I have tried to deploy the same model as app in https://share.streamlit.io/, but, I have the next error:
“Keras 3 only supports V3 .keras and .weights.h5 files, or legacy V1/V2 .h5 files.”

My files are:
weights.data-00000-of-00001
weights.index
checkpoint

To load the weigths, I used:
import tensorflow as tf
unet.load_weights(‘weights’)

My requirement file is:
matplotlib
numpy
Pillow
scipy
opencv-python-headless
pytube
ipython
streamlit
tensorflow
keras

Please, do you have some idea how I can to load the weights of my model?

Thank you!

2

Hi @Artificial_Intellige

Have you try specifying the same version number for tensorflow that you used locally with that used on the cloud?

You can do:

pip list

to figure out which specific version that you’re using and specify that in the requirements.txt file.