rgov
rgov

Reputation: 4369

Building TensorFlow Lite with all TF ops?

How can I build the C++ TensorFlow Lite runtime library with all the select TF operators? I can't feed in a model at build time to selectively build a subset.

I'd like a cross-platform solution but for my immediate development needs, I am on macOS.

Upvotes: 0

Views: 897

Answers (1)

Karim Nosseir
Karim Nosseir

Reputation: 590

You need to build the TFLite runtime with the flex delegate as dependency, specifically this rule. This bazel macro builds a shared library for android and if no models specified will link all allowed Select TF ops, you can use it as an example.

Upvotes: 1

Related Questions