We read every piece of feedback, and take your input very seriously.
To see all available qualifiers, see our documentation.
1 parent adbfa12 commit 9453b7aCopy full SHA for 9453b7a
1 file changed
transformer_engine/plugin/core/backends/flagos/impl/softmax.py
@@ -60,4 +60,4 @@ def scaled_masked_softmax_backward_fl(
60
# grad_softmax = softmax_output_f32 * (output_grad_f32 - sum_term)
61
grad_softmax = flag_gems.mul(softmax_output_f32, flag_gems.sub(output_grad_f32, sum_term))
62
grad_scaled = flag_gems.mul(grad_softmax, scale_factor)
63
- return flag_gems.to_copy(grad_scaled, dtype=orig_dtype)
+ return flag_gems.to_copy(grad_scaled, dtype=orig_dtype)
0 commit comments