Skip to content

Fix missing stream in layer_norm and sdpa fallbacks #4663

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
zcbenz merged 2 commits into from
Oct 10, 2026
Merged

Fix missing stream in layer_norm and sdpa fallbacks #4663

Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
64a6f77
Pass stream to all ops in layer_norm fallback and SDPA bool mask conv…
AkshayShah03 Oct 8, 2026
5ce4ec9
nit
zcbenz Oct 10, 2026
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: 4 additions & 3 deletions mlx/fast.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -352,7 +352,7 @@ array layer_norm(
auto xc = subtract(x, mu, s);
auto v = mean(square(xc, s), /* axis= */ -1, /* keepdims= */ true, s);

x = multiply(xc, rsqrt(add(v, array(eps, float32), s), s));
x = multiply(xc, rsqrt(add(v, array(eps, float32), s), s), s);
x = astype(x, out_type, s);

// If the LN is affine then transform x according to the weight and bias
Expand Down Expand Up @@ -945,8 +945,9 @@ array scaled_dot_product_attention(
array& mask = inputs[3];
mask = where(
mask,
full_like(mask, 0, final_type, s),
full_like(mask, -inf, final_type, s));
full_like(mask, 0, final_type, stream),
full_like(mask, -inf, final_type, stream),
stream);
}
Shape out_shape{q.shape(0), q.shape(1), q.shape(2), v.shape(-1)};
auto primitive = std::make_shared<ScaledDotProductAttention>(
Expand Down
Loading