How to install deepspeech tflite on armv7 cortex A7

DeepSpeech
#1

Hi , i want to install deepspeech tflite on imx7 (using openembedded linux , yocto) , but i failed to add deepspeech layer , so now , i want to install it using docker , pip can’t find switable whl package , that 's why i should build it from source . Have you any idea, how i can install deepspeech on imx7? Thank you.

#2

If you don’t care to share more explicit things you performed, there’s no way we can help you.

We have python wheels published for linux/armv7 and we have documentation on how to rebuild … Please read it?

#3

Here, e.g., https://files.pythonhosted.org/packages/a8/01/693eb642003e1a15e41d1224334d44934325e20f3d01a9a7831cc416eb70/deepspeech-0.9.3-cp37-cp37m-linux_armv7l.whl

As documented, we only support those on Debian/Armbian/Raspbian and we only test on RPi3 cores. I’m unsure it can run directly.

#4

https://deepspeech.readthedocs.io/en/r0.9/BUILDING.html#cross-building

And you can rely on existing guides for Python bindings, just add TARGET=xxx.

#5

That being said, given the specs of i.MX7 Cortex-A7, I’m doubtful a full blown generic model will work, and it’s very likely you will have to train your own model with a lower n_hidden value to reduce the complexity of the network.