use transform_for_execution to get callable for torch compile

[t-vi]

Fixes #1040

Note that back in the day, @IvanYashchuk added a comment about suspected bugs preventing this approach:

lightning-thunder/thunder/executors/torch_compile.py

Lines 69 to 85 in 7c1f94a

	# Here instead of using thunder.trace we could use torch_trace =
	# passes._transform_for_operator_executor_execution(region_trace, [torchex])
	# but then we would need to handle unpacking of the args explicitly For
	# example with:
	# try:
	# token = set_tracectx(region_trace)
	# col = CollectionProxy(region_trace.args, name="args")
	# _ = prims.unpack_sequence(col, len(region_trace.args))
	# finally:
	# reset_tracectx(token)
	# region_trace.bound_symbols.extend(bsyms)
	# But there are some issues with the
	# _transform_for_operator_executor_execution implementation that need to be
	# fixed first. One issue is that it doesn't maintain the ssa form of the
	# trace, which is needed for all the passes to work correctly.
	# TODO: issue "Try using _transform_for_operator_executor_execution for
	# torch.compile executor"

Now I seem to get good results with just calling transform_for_execution with only the torchex as int the executors_list. Maybe the bugs have been fixed since.

[lantiga]

Stamped!

[t-vi]

Seems I hit the exact problem that Ivan described.

[t-vi]

Turns out that here, it seems the problem was that the region_trace was not enough of a proper trace to work.
Note that the region trace still does not have unpack_trivial at the top.
Maybe we should either forego regions entirely in favour of traces or have a utility region->trace that does it.

[t-vi]

Merging, but don't hesitate to scream if you don't like it. 😉

[IvanYashchuk]

Maybe we should either forego regions entirely in favour of traces

Yes, that would be nice. Regions look like subclasses of TraceCtx with positional-only arguments and automatic identification of arguments and outputs.