Add float16 suport for scatter_max/scatter_min on gpu
This fix tries to address the issue raised in 24219 where
there were no float16 supports for scatter_max/scatter_min on gpu.
This fix adds the float16 support for scatter_max/scatter_min on gpu.
This fix fixes 24219.
Signed-off-by:
Yong Tang <yong.tang.github@outlook.com>
Loading
Please sign in to comment