Skip to content

[release/2.8] layernorm tests: Tweak test thresholds for comparing tensors #2583

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

Merged
merged 1 commit into from
Aug 27, 2025
Merged

[release/2.8] layernorm tests: Tweak test thresholds for comparing tensors #2583

Changes from all commits
Commits
Show all changes
1 commit
Select commit
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
9 changes: 7 additions & 2 deletions test/test_nn.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -7437,10 +7437,15 @@ def test_layer_norm_backwards_eps(self):
ln_out_cuda = ln_cuda(x_cuda)
ln_out.backward(grad_output)
ln_out_cuda.backward(grad_output_cuda)
atol = 1e-4
rtol = 1e-5
if m > 64 * 1024:
atol = 1e-3
rtol = 1e-3
if elementwise_affine:
self.assertEqual(ln.weight.grad, ln_cuda.weight.grad, f"weight grad failed: {m=} {n=}", rtol=1e-4, atol=1e-4)
self.assertEqual(ln.weight.grad, ln_cuda.weight.grad, f"weight grad failed: {m=} {n=}", rtol=rtol, atol=atol)
if bias and elementwise_affine:
self.assertEqual(ln.bias.grad, ln_cuda.bias.grad, f"bias grad failed: {m=} {n=}", rtol=1e-5, atol=1e-4)
self.assertEqual(ln.bias.grad, ln_cuda.bias.grad, f"bias grad failed: {m=} {n=}", rtol=rtol, atol=atol)

@largeTensorTest("40GB", device="cuda")
def test_layer_norm_large_tensor(self):
Expand Down