[1/n] Enable DCA CUDA graph capture (#9537)

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

@@ -1537,7 +1537,7 @@ class DualChunkFlashAttentionBackend(AttentionBackend):
                     query_inter,
                     key_cache,
                     value_cache,
-                    block_table[:, : decode_meta.max_seq_len_inter],
+                    block_table,
                     decode_meta.seq_lens_inter,
                     softmax_scale,
                     causal=False,

python/sglang/srt/server_args.py

@@ -878,10 +878,9 @@ class ServerArgs:
 
         if self.attention_backend == "dual_chunk_flash_attn":
             logger.warning(
-                "Mixed chunk, radix cache, and cuda graphs are disabled because of using dual chunk flash attention backend"
+                "Mixed chunk and radix cache are disabled when using dual-chunk flash attention backend"
             )
             self.enable_mixed_chunk = False
-            self.disable_cuda_graph = True
             self.disable_radix_cache = True
 
     def _handle_page_size(self):