Fall back to eager mode when viewing with differing bitwidths (#120998) (#121786)
The inductor lowering code for viewing a tensor as a type with a different bitwidth currently doesn't generate valid triton code. This change looks for a source and destination dtype and, if different sizes, falls back to the eager mode aten implementation. Prior to this change, this condition would throw an exception.
Fixes #120998.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/121786
Approved by: https://github.com/peterbell10, https://github.com/bertmaher