Flex Ops about tensorflowlite-flexdelegate HOT 4 CLOSED

Thank you for the suggestion. My working policy is clear. I will generate a Flex-enabled installer the day after the official release is announced. For example, the following repository is the repository where the pip package with FlexDelegate enabled is committed. I plan to build it the day after the v2.5.0 branch is created and moved to release. It is released every three months, so I expect v2.5.0 to be released soon.
https://github.com/PINTO0309/TensorflowLite-bin

from tensorflowlite-flexdelegate.

Thats great as been trying to work out the release schedule as the FlexDelegate I think will be a big one for choice of models in 2.5 as really interested in what the perf gains will be for TFL with FlexOps.
I have been aware of your great TFL Pi community wheels for a while but the natural streaming models often used such as GRU, CRNN or SVDF until FlexOps has excluded my interest.
MFCC is also in the Core now but the CRNN is a true hybrid of CNN then GRU and wondered if not CRNN then maybe you are also curious to add more audio based models to your model zoo?

I will be eagerly checking for releases many thanks.

from tensorflowlite-flexdelegate.

Hey pinto, would you update this repository since flexdelegate in tensorflow 2.5 has much improved

from tensorflowlite-flexdelegate.

@etyhh
https://github.com/PINTO0309/TensorflowLite-bin

from tensorflowlite-flexdelegate.