-
Notifications
You must be signed in to change notification settings - Fork 428
Open
Description
Using the "eager" flag simultaneously with the "tensorflow_runtime_linking" flag results in the compilation of the entire tensorflow C api. This is not ideal.
I have compiled an optimized tensorflow for an aarch64 production environment, so runtime linking is what I'm looking for. My project uses the eager API (has worked fine so far). I'm trying to compile the project using target=aarch64_unknown_linux_gnu and features= "eager", "tensorflow_runtime_linking", "ndarray".
Metadata
Metadata
Assignees
Labels
No labels