What are differences between Espressif's TensorFlow Lite Micro and original?
Posted: Sun Nov 01, 2020 5:04 pm
TensorFlow Lite for Microcontrollers is part of the normal native TensorFlow repository. But Espressif has its own fork as well.
I usually assume an ESP-specific version of a component has simplification or improvements for esp32, as compared to the creator's generic repo version.
I am trying to determine what are the main differences between repos: github.com/espressif/tensorflow and github.com/tensorflow/tensorflow
When I run 'make' to create project templates, I get varying numbers of files for components/tfmicro/ depending on whose repo and which example project I specify.
E.g.
Any guidance as to what I am getting if I choose the Espressif repo would be helpful
And why is tfmicro/ dependent on which example project is chosen?
Thanks!
I usually assume an ESP-specific version of a component has simplification or improvements for esp32, as compared to the creator's generic repo version.
I am trying to determine what are the main differences between repos: github.com/espressif/tensorflow and github.com/tensorflow/tensorflow
When I run 'make' to create project templates, I get varying numbers of files for components/tfmicro/ depending on whose repo and which example project I specify.
E.g.
Code: Select all
make -f tensorflow/lite/micro/tools/make/Makefile TARGET=esp generate_hello_world_esp_project
And why is tfmicro/ dependent on which example project is chosen?
Thanks!