Skip to content

optimize pad operations in fa3 to accelarate 100+us #6077

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 4 commits into from
May 8, 2025
Merged

optimize pad operations in fa3 to accelarate 100+us #6077

Changes from all commits
Commits
Show all changes
4 commits
Select commit Hold shift + click to select a range
9e6ce0a
optimize pad operations in fa3 to accelarate 100+us
May 7, 2025
0eb1816
fix formatting
zminglei May 7, 2025
c9f97ae
Merge branch 'main' into zminglei/remove-pad
hebiao064 May 7, 2025
880b35f
Merge branch 'main' into zminglei/remove-pad
May 8, 2025
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
56 changes: 17 additions & 39 deletions python/sglang/srt/layers/attention/flashattention_backend.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -1525,12 +1525,9 @@ def init_forward_metadata_replay_cuda_graph(
metadata.max_seq_len_k = seq_lens_cpu.max().item() + (
self.speculative_step_id + 1
)
metadata.cu_seqlens_k.copy_(
torch.nn.functional.pad(
torch.cumsum(
metadata.cache_seqlens_int32, dim=0, dtype=torch.int32
),
(1, 0),
metadata.cu_seqlens_k[1:].copy_(
torch.cumsum(
metadata.cache_seqlens_int32, dim=0, dtype=torch.int32
)
)

Expand All @@ -1554,12 +1551,9 @@ def init_forward_metadata_replay_cuda_graph(
# metadata.max_seq_len_q = self.topk, already set in capture
metadata.max_seq_len_k = seq_lens_cpu.max().item()
# metadata.cu_seqlens_q already set in capture
metadata.cu_seqlens_k.copy_(
torch.nn.functional.pad(
torch.cumsum(
metadata.cache_seqlens_int32, dim=0, dtype=torch.int32
),
(1, 0),
metadata.cu_seqlens_k[1:].copy_(
torch.cumsum(
metadata.cache_seqlens_int32, dim=0, dtype=torch.int32
)
)

Expand Down Expand Up @@ -1616,13 +1610,8 @@ def init_forward_metadata_replay_cuda_graph(
metadata.max_seq_len_k = (
seq_lens_cpu.max().item() + self.speculative_num_draft_tokens
)
metadata.cu_seqlens_k.copy_(
torch.nn.functional.pad(
torch.cumsum(
metadata.cache_seqlens_int32, dim=0, dtype=torch.int32
),
(1, 0),
)
metadata.cu_seqlens_k[1:].copy_(
torch.cumsum(metadata.cache_seqlens_int32, dim=0, dtype=torch.int32)
)
max_seq_pages = (
metadata.max_seq_len_k + self.page_size - 1
Expand All @@ -1641,13 +1630,8 @@ def init_forward_metadata_replay_cuda_graph(
# metadata.max_seq_len_q = self.speculative_num_draft_tokens, already set in capture
metadata.max_seq_len_k = seq_lens_cpu.max().item()
# metadata.cu_seqlens_q already set in capture
metadata.cu_seqlens_k.copy_(
torch.nn.functional.pad(
torch.cumsum(
metadata.cache_seqlens_int32, dim=0, dtype=torch.int32
),
(1, 0),
)
metadata.cu_seqlens_k[1:].copy_(
torch.cumsum(metadata.cache_seqlens_int32, dim=0, dtype=torch.int32)
)
page_table = self.req_to_token[
req_pool_indices, : metadata.max_seq_len_k
Expand Down Expand Up @@ -1705,14 +1689,11 @@ def init_forward_metadata_replay_cuda_graph(
metadata_expand.cache_seqlens_int32.copy_(
mask.sum(dim=1).to(torch.int32)
)
metadata_expand.cu_seqlens_k.copy_(
torch.nn.functional.pad(
torch.cumsum(
metadata_expand.cache_seqlens_int32,
dim=0,
dtype=torch.int32,
),
(1, 0),
metadata_expand.cu_seqlens_k[1:].copy_(
torch.cumsum(
metadata_expand.cache_seqlens_int32,
dim=0,
dtype=torch.int32,
)
)
metadata_expand.max_seq_len_k = (
Expand All @@ -1723,11 +1704,8 @@ def init_forward_metadata_replay_cuda_graph(
# Only support encoder size 1 for now
metadata.encoder_max_seq_len_k = encoder_lens[0]
metadata.encoder_lens_int32.copy_(encoder_lens[:1])
metadata.encoder_cu_seqlens_k.copy_(
torch.nn.functional.pad(
torch.cumsum(metadata.encoder_lens_int32, dim=0, dtype=torch.int32),
(1, 0),
)
metadata.encoder_cu_seqlens_k[1:].copy_(
torch.cumsum(metadata.encoder_lens_int32, dim=0, dtype=torch.int32)
)

metadata.encoder_page_table[:, : metadata.encoder_max_seq_len_k].copy_(
Expand Down
Loading