Skip to content

better error handling when window size is -1 and no max seq len is provided #4979

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

Closed
wants to merge 1 commit into from
+16 −4
Closed

better error handling when window size is -1 and no max seq len is provided #4979

Show file tree
Hide file tree
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
10 changes: 8 additions & 2 deletions ...m_gpu/experimental/gen_ai/src/attention/cuda/cutlass_blackwell_fmha/blackwell_fmha_bwd.cu
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -41,10 +41,16 @@ std::tuple<at::Tensor, at::Tensor, at::Tensor> dispatch_fmha_bwd(
causal = false; // Use local attention instead of causal
}
// Expand -1 window sizes to full sequence length if available
if (window_size_left < 0 && max_seq_len_k.has_value()) {
if (window_size_left < 0) {
TORCH_CHECK(
max_seq_len_k.has_value(),
"window_size_left is negative but max_seq_len_k is not provided");
window_size_left = max_seq_len_k.value();
}
if (window_size_right < 0 && max_seq_len_k.has_value()) {
if (window_size_right < 0) {
TORCH_CHECK(
max_seq_len_k.has_value(),
"window_size_right is negative but max_seq_len_k is not provided");
window_size_right = max_seq_len_k.value();
}
}
Expand Down
10 changes: 8 additions & 2 deletions ...m_gpu/experimental/gen_ai/src/attention/cuda/cutlass_blackwell_fmha/blackwell_fmha_fwd.cu
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -26,10 +26,16 @@ std::tuple<at::Tensor, at::Tensor> dispatch_fmha_fwd(
causal = false; // Use local attention instead of causal
}
// Expand -1 window sizes to full sequence length if available
if (window_size_left < 0 && max_seq_len_k.has_value()) {
if (window_size_left < 0) {
TORCH_CHECK(
max_seq_len_k.has_value(),
"window_size_left is negative but max_seq_len_k is not provided");
window_size_left = max_seq_len_k.value();
}
if (window_size_right < 0 && max_seq_len_k.has_value()) {
if (window_size_right < 0) {
TORCH_CHECK(
max_seq_len_k.has_value(),
"window_size_right is negative but max_seq_len_k is not provided");
window_size_right = max_seq_len_k.value();
}
}
Expand Down
Loading