Optimize a pad operation to accelerate 25us (sgl-project#5945)

Author: hebiao064

python/sglang/srt/layers/attention/flashattention_backend.py

@@ -1617,8 +1617,9 @@ def init_forward_metadata_replay_cuda_graph(
                 metadata.max_seq_len_k = max_len
 
                 metadata.cache_seqlens_int32 = seq_lens.to(torch.int32)
-                metadata.cu_seqlens_k = torch.nn.functional.pad(
-                    torch.cumsum(seq_lens, dim=0, dtype=torch.int32), (1, 0)
+                # Optimize cumulative sequence length calculation
+                metadata.cu_seqlens_k[1:].copy_(
+                    torch.cumsum(seq_lens, dim=0, dtype=torch.int32)
                 )
 
                 max_seq_pages = (