You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
asym8uxsym8s_asym8u will be similar( in terms of implementations) to asym8sxsym8s_asym8s kernels we have
but it would be use quint8 instead of qint8 as input.
The text was updated successfully, but these errors were encountered:
@han-qiu,
We do have uint8 support for couple of operators as TFLM team moved from uint8 to int8 support, our latest kernels supports int8 format to sync with TFLM(TensorFlow Lite Micro) development.
Currently we do not have any plan to support uint8 variant kernels.
@cad-audio
Hi,Thanks for getting back!
would it be ok that I implement it myself and create a pull request?
I think it's probaly not very hard to do a asym8uxsym8s_asym8u based on the sym8sxsym8s_sym8s implementation.
asym8uxsym8s_asym8u will be similar( in terms of implementations) to asym8sxsym8s_asym8s kernels we have
but it would be use quint8 instead of qint8 as input.
The text was updated successfully, but these errors were encountered: