Using Julia with Coral google board

Hello

I am doing a research where I have to do a benchmark between a PC with GPU vs the Coral board’s TPU. But I have to use Julia.

The Coral board uses TensorFlow Lite.

Any guidance will be appreciated.

Tell google to publish documentation about the board so we can properly target it :wink:

1 Like

My plan is to use TensorFlow.jl https://www.youtube.com/watch?v=n2MwJ1guGVQ
In minute 4 with 13 seconds, the presenter says that it is possible to share code with people who are using python.

Once I have created the models I will convert them into TensorFlow Lite. Then compile and send that to the board. Like this:

What do you think?

I was able to install it, however, I couldn’t use it because Tensorflow.jl required conda to install some dependencies and conda didn’t have aarch64 support at the time. Now, I believe it has but haven’t tried again.