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

aot_autograd: factor out runtime epilogue from aot_dispatch_base #100586

Closed
wants to merge 3 commits into from
Closed

aot_autograd: factor out runtime epilogue from aot_dispatch_base #100586

Changes from all commits
Commits
Show all changes
3 commits
Select commit Hold shift + click to select a range
54cec01
aot_autograd: factor out runtime epilogue from aot_dispatch_base
bdhirsh May 3, 2023
b8c53ee
Update on "aot_autograd: factor out runtime epilogue from aot_dispatc…
bdhirsh May 11, 2023
6a7a33f
Update on "aot_autograd: factor out runtime epilogue from aot_dispatc…
bdhirsh May 11, 2023
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
7 changes: 6 additions & 1 deletion torch/_functorch/aot_autograd.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -1306,7 +1306,7 @@ def call_func_with_args(f, args, steal_args=False, disable_amp=False):
del guard
return out

def aot_dispatch_base(flat_fn, flat_args: List[Tensor], aot_config: AOTConfig, *, fw_metadata: ViewAndMutationMeta):
def aot_dispatch_base_graph(flat_fn, flat_args: List[Tensor], aot_config: AOTConfig, *, fw_metadata: ViewAndMutationMeta):
# aot_dispatch_base requires functionalization, but doesn't need to handle as many cases as the autograd case.
# The cases that aot_dispatch_base doesn't need to handle include:
# - outputs that are aliases of graph intermediates
Expand Down Expand Up @@ -1342,6 +1342,11 @@ def aot_dispatch_base(flat_fn, flat_args: List[Tensor], aot_config: AOTConfig, *
if aot_config.enable_log:
aot_graphs_log.info("%s", lazy_format_graph_code("Forward graph", fw_module, aot_config.aot_id))

return fw_module

def aot_dispatch_base(flat_fn, flat_args: List[Tensor], aot_config: AOTConfig, *, fw_metadata: ViewAndMutationMeta):
fw_module = aot_dispatch_base_graph(flat_fn, flat_args, aot_config, fw_metadata=fw_metadata)

disable_amp = torch._C._is_any_autocast_enabled()
context = disable_autocast_manager if disable_amp else nullcontext

Expand Down