Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom

Print AOT Autograd graph name when accuracy failed #99366

Closed
wants to merge 3 commits into from
Closed

Print AOT Autograd graph name when accuracy failed #99366

Changes from 1 commit
Commits
Show all changes
3 commits
Select commit Hold shift + click to select a range
9f4f615
Print AOT Autograd graph name when accuracy failed
ezyang Apr 17, 2023
7044fbd
Update on "Print AOT Autograd graph name when accuracy failed"
ezyang Apr 18, 2023
712e8b0
Update on "Print AOT Autograd graph name when accuracy failed"
ezyang Apr 20, 2023
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Jump to
Jump to file
Failed to load files.
Diff view
Diff view
6 changes: 5 additions & 1 deletion torch/_dynamo/debug_utils.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -540,6 +540,10 @@ def debug_wrapper(gm, example_inputs, **kwargs):

compiler_fn = functools.partial(unconfigured_compiler_fn, **kwargs)

from torch._functorch.aot_autograd import get_aot_graph_name
graph_name = get_aot_graph_name()

# TODO: Why do we have to save the orig_graph?
orig_graph = copy.deepcopy(gm.graph)
assert config.repro_after in ("dynamo", "aot", None)
inner_compiled_fn = None
Expand Down Expand Up @@ -578,7 +582,7 @@ def deferred_for_real_inputs(real_inputs):
if inner_compiled_fn is None:
inner_compiled_fn = compiler_fn(gm, example_inputs)
if backend_aot_accuracy_fails(gm, real_inputs, compiler_fn):
log.warning("Accuracy failed for the AOT Autograd graph")
log.warning("Accuracy failed for the AOT Autograd graph %s", graph_name)
dump_compiler_graph_state(
fx.GraphModule(gm, orig_graph),
copy_tensor_attrs,
Expand Down