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

[functorch] fix AOTAutograd tutorial #87415

Closed
wants to merge 2 commits into from
Closed

[functorch] fix AOTAutograd tutorial #87415

Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
d8346d7
[functorch] fix AOTAutograd tutorial
zou3519 Oct 20, 2022
d90c285
Update on "[functorch] fix AOTAutograd tutorial"
zou3519 Oct 20, 2022
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
8 changes: 7 additions & 1 deletion functorch/notebooks/aot_autograd_optimizations.ipynb
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -126,7 +126,10 @@
"aot_print_fn = aot_function(fn, fw_compiler=compiler_fn, bw_compiler=compiler_fn)\n",
"\n",
"# Run the aot_print_fn once to trigger the compilation and print the graphs\n",
"res = aot_print_fn(a, b, c, d).sum().backward()\n",
"cloned_inputs = [x.clone().detach().requires_grad_(True) for x in (a, b, c, d)]\n",
"cloned_a, cloned_b, cloned_c, cloned_d = cloned_inputs\n",
"res = aot_print_fn(cloned_a, cloned_b, cloned_c, cloned_d)\n",
"res.sum().backward()\n",
"assert torch.allclose(ref, res)"
]
},
Expand Down Expand Up @@ -300,6 +303,9 @@
"source": [
"from functorch.compile import min_cut_rematerialization_partition\n",
"\n",
"# Zero out the gradients so we can do a comparison later\n",
"a.grad, b.grad, c.grad, d.grad = (None,) * 4\n",
"\n",
"# Lets set up the partitioner. Also set the fwd and bwd compilers to the printer function that we used earlier.\n",
"# This will show us how the recomputation has modified the graph.\n",
"aot_fn = aot_function(fn, fw_compiler=compiler_fn, bw_compiler=compiler_fn, partition_fn=min_cut_rematerialization_partition)\n",
Expand Down