Skip to content

Fix bug in dglu xfilter math #568

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
Anerudhan merged 1 commit into from
Aug 20, 2026
Merged

Fix bug in dglu xfilter math #568

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
15 changes: 6 additions & 9 deletions python/cudnn/gemm/cutedsl/grouped/dglu/moe_blockscaled_grouped_gemm_dglu_rubin.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -2004,17 +2004,15 @@ def dgeglu(
if cutlass.const_expr(self.has_prob):
dy1_0, dy1_1 = fmul2((dy1_0, dy1_1), mprob2)

x1_filter_0 = y1_0 if x1_0 <= geglu_max_value else cutlass.Float32(0.0)
x1_filter_1 = y1_1 if x1_1 <= geglu_max_value else cutlass.Float32(0.0)
x1_filter_0 = cutlass.Float32(1.0) if x1_0 <= geglu_max_value else cutlass.Float32(0.0)
x1_filter_1 = cutlass.Float32(1.0) if x1_1 <= geglu_max_value else cutlass.Float32(0.0)

dx1_vec[i], dx1_vec[i + 1] = fmul2((dy1_0, dy1_1), (cutlass.Float32(x1_filter_0), cutlass.Float32(x1_filter_1)))

# dy2 = g * y1 * sigmoid_out * mProb
dy2_0, dy2_1 = fmul2(y1, acc_mul_sigmoid_prob)
x2_filter_0 = x2_0 if x2_0 <= geglu_max_value else x_dtype(0.0)
x2_filter_1 = x2_1 if x2_1 <= geglu_max_value else x_dtype(0.0)
x2_filter_0 = y2_0 if x2_filter_0 >= geglu_min_value else cutlass.Float32(0.0)
x2_filter_1 = y2_1 if x2_filter_1 >= geglu_min_value else cutlass.Float32(0.0)
x2_filter_0 = cutlass.Float32(1.0) if (x2_0 >= geglu_min_value and x2_0 <= geglu_max_value) else cutlass.Float32(0.0)
x2_filter_1 = cutlass.Float32(1.0) if (x2_1 >= geglu_min_value and x2_1 <= geglu_max_value) else cutlass.Float32(0.0)
dx2_vec[i], dx2_vec[i + 1] = fmul2((dy2_0, dy2_1), (cutlass.Float32(x2_filter_0), cutlass.Float32(x2_filter_1)))

if cutlass.const_expr(self.generate_dprob):
Expand Down Expand Up @@ -2052,9 +2050,8 @@ def dgeglu(
dy1 = g * sigmoid_out * (1 + 1.702 * y1 * (1 - sigmoid_out)) * (y2 + linear_offset)
dy2 = g * y1 * sigmoid_out

x1_filter = x1_vec_load[i] if x1_vec_load[i] <= 7.0 else 0.0
x2_filter = x2_vec_load[i] if x2_vec_load[i] <= 7.0 else 0.0
x2_filter = x2_filter if x2_filter >= -7.0 else 0.0
x1_filter = 1.0 if x1_vec_load[i] <= 7.0 else 0.0
x2_filter = 1.0 if (x2_vec_load[i] >= -7.0 and x2_vec_load[i] <= 7.0) else 0.0

dx1_vec[i] = x1_filter * dy1
dx2_vec[i] = x2_filter * dy2
Expand Down