Thank you for the advice.
I'm not used to RasPi So it's very helpful for me.
As you advised, I used the converter with TensorFlow==2.14.0.
But I got still the following error.
As the last line of the error saying, I failed to load the .keras model.
So I tried to generate the .keras model again according to the following notebook on the google colab.
https://github.com/sony/model_optimizat ... x500.ipynb
As default, TensorFlow version is 2.17.0 in my google colab.So I run following command and uninstall TensorFlow==2.17.0.Then I tried to run following command in the sample Notebook, I got an error saying there are dependency conflicts about [tf-keras], [tensorstore].
So I uninstall [tf-keras], [tensorstore].
And I reinstall [tf-keras==2.14.1], [tensrostore==0.1.45].
But still I got a conflict error about [orbax-checkpoint].
So I uninstalled it and reinstalled [orbax-checkpoint==0.3.5]
After that, I did't get an error.
Then I export the .keras file according to the sample Notebook and tried to use converter on command prompt.
But I still got an error saying failed to load the model.
Is it caused by Python dependency conflicts?
Or any other problems?
I'm not used to RasPi So it's very helpful for me.
As you advised, I used the converter with TensorFlow==2.14.0.
But I got still the following error.
- Image may be NSFW.
Clik here to view.
- Image may be NSFW.
Clik here to view.
As the last line of the error saying, I failed to load the .keras model.
So I tried to generate the .keras model again according to the following notebook on the google colab.
https://github.com/sony/model_optimizat ... x500.ipynb
As default, TensorFlow version is 2.17.0 in my google colab.
- Image may be NSFW.
Clik here to view.
Code:
!pip uninstall tensorflowCode:
TF_VER = '2.14'!pip install -q tensorflow[and-cuda]~={TF_VER}So I uninstall [tf-keras], [tensorstore].
And I reinstall [tf-keras==2.14.1], [tensrostore==0.1.45].
But still I got a conflict error about [orbax-checkpoint].
So I uninstalled it and reinstalled [orbax-checkpoint==0.3.5]
After that, I did't get an error.
Then I export the .keras file according to the sample Notebook and tried to use converter on command prompt.
But I still got an error saying failed to load the model.
Is it caused by Python dependency conflicts?
Or any other problems?
Statistics: Posted by k_sato — Tue Nov 12, 2024 7:09 am
