- add float16 suppport
 - fix 64 bit input bug
 - fix int8 bug due to integer overflow
 - more checking on k
 - add license from pytorch