Andy Kassen
Results
3
issues of
Andy Kassen
Follow-up to #3205 adding attention scale info to the SPDA verbose line.
component:common
Partially addresses [MFDNN-13444](https://jira.devtools.intel.com/browse/MFDNN-13444).
platform:gpu-intel
Testing with [1], shows that f32->f8 down-convert with stochastic rounding always rounds down, so in practice, it *almost* always rounds down. This is because there are 4-5 bits between the...
platform:gpu-intel
component:tests
component:common