Thanks for the information.
Are there any pre-trained checkpoints available for v0.6?
I’m trying to see if it is at all possible to get speech inference running on raspberry pis with an edgetpu accelerator to compare the speed against a desktop CPU.