diff options
author | Dwight Lidman <dwight.lidman@arm.com> | 2021-03-18 14:51:42 +0100 |
---|---|---|
committer | patrik.gustavsson <patrik.gustavsson@arm.com> | 2021-03-26 10:07:24 +0000 |
commit | 9bb1e2ed361286769c362f002910d1dbd1736b05 (patch) | |
tree | cf16840abfe8c5494678d5ff824d37eb7417c8e9 /ethosu/vela/scaling.py | |
parent | c629129f79666eeea1d86a779c8cb245e052672f (diff) | |
download | ethos-u-vela-9bb1e2ed361286769c362f002910d1dbd1736b05.tar.gz |
MLBEDSW-4163: OFM zero point outside valid range
This commit fixes a bug where the OFM zero point
would assume values outside of [0, 255] due to it's
usage as a stand-in for a bias when emulating the
TensorFlow Lite implementation of MEAN.
The solution is to adjust for the bias using an
ADD operator with the bias value as an int16 const
tensor. The 16-bit integer is needed as the bias
is 32 bits in the original implementation but can
effectively assume values in the range [-255, 255].
Signed-off-by: Dwight Lidman <dwight.lidman@arm.com>
Change-Id: I84df48ea89bb559954f1b2c289b65e08a6418274
Diffstat (limited to 'ethosu/vela/scaling.py')
0 files changed, 0 insertions, 0 deletions