update deepspeed to run with the most recent triton 2.1.0 (#4278)
* fix codes to work with triton 2.1
tl.libdevice and triton.testing.allclose are gone with triton2.1
* formatting
* formatting
---------
Co-authored-by: Stephen Youn <styoun@microsoft.com>
Co-authored-by: Lev Kurilenko <113481193+lekurile@users.noreply.github.com>