1

I want to use pretrained SqueezeNet model implemented in Tensorflow in another Python package. The problem is that the current version of Tenorflow does not support Python 3.7 and also including whole Tensorflow package as a requirement is not a good idea since it is very space consuming and can be an additional source of issues in the package. The best option would be to avoid Tensorflow as a requirement.

Is there any option to convert Tensorflow model to operations in Numpy or any other representation which is lighter. I do not need the support for training the model since it is already pretrained. I will use the model just for prediction purposes. The solution must have an option to be packed in a Python package (installable through pip).

Any other suggestions welcome.

Improve this question

1 Answer 1

Reset to default
2

You might want to try exporting your model to tflite and see if you can afford to depend on that. There is no currently supported way to automatically convert any tensorflow model to numpy but you might be able to do so manually by inspecting your model.

Improve this answer
Sign up to request clarification or add additional context in comments.

2 Comments

Thank you for the suggestion. I will try with to export the model to the tflite.
I do not find the solution to run tflite from Python without having Tensorflow installed (and then using its lite interpreter). I think I will need to find another solution.

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Start asking to get answers

Find the answer to your question by asking.

Ask question

Explore related questions

See similar questions with these tags.