feat: support flashinfer mla with prefix cache (#3643)

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

@@ -54,7 +54,9 @@ class DecodeMetadata:
 
 @dataclass
 class PrefillMetadata:
-    prefill_wrappers: List[BatchPrefillWithPagedKVCacheWrapper]
+    prefill_wrappers: List[
+        Union[BatchPrefillWithPagedKVCacheWrapper, BatchMLAPagedAttentionWrapper]
+    ]
     use_ragged: bool
     extend_no_prefix: bool
 
@@ -160,6 +162,24 @@ class FlashInferAttnBackend(AttentionBackend):
         self.decode_wrappers = []
         for _ in range(self.num_wrappers):
             if not skip_prefill:
+                if (
+                    self.enable_flashinfer_mla
+                    and not global_server_args_dict["disable_radix_cache"]
+                ):
+                    # use mla paged prefill
+                    self.prefill_wrappers_paged.append(
+                        BatchMLAPagedAttentionWrapper(
+                            self.workspace_buffer,
+                            backend="fa2",
+                        )
+                    )
+                    self.prefill_wrappers_verify.append(
+                        BatchMLAPagedAttentionWrapper(
+                            self.workspace_buffer,
+                            backend="fa2",
+                        )
+                    )
+                else:
                     self.prefill_wrappers_paged.append(
                         BatchPrefillWithPagedKVCacheWrapper(
                             self.workspace_buffer,
@@ -168,7 +188,9 @@ class FlashInferAttnBackend(AttentionBackend):
                         )
                     )
                     self.prefill_wrappers_verify.append(
-                    BatchPrefillWithPagedKVCacheWrapper(self.workspace_buffer, "NHD")
+                        BatchPrefillWithPagedKVCacheWrapper(
+                            self.workspace_buffer, "NHD"
+                        )
                     )
             if self.enable_flashinfer_mla:
                 self.decode_wrappers.append(
@@ -237,7 +259,10 @@ class FlashInferAttnBackend(AttentionBackend):
         else:
             prefix_lens = forward_batch.extend_prefix_lens
 
-            if self.is_multimodal:
+            if self.is_multimodal or (
+                self.enable_flashinfer_mla
+                and not global_server_args_dict["disable_radix_cache"]
+            ):
                 use_ragged = False
                 extend_no_prefix = False
             else:
@@ -419,7 +444,9 @@ class FlashInferAttnBackend(AttentionBackend):
 
             logits_soft_cap = layer.logit_cap
 
-            o1, _ = self.prefill_wrapper_ragged.forward_return_lse(
+            if global_server_args_dict["disable_radix_cache"]:
+                # use mla ragged prefill
+                o, _ = self.prefill_wrapper_ragged.forward_return_lse(
                     q.view(-1, layer.tp_q_head_num, layer.head_dim),
                     k.view(-1, layer.tp_k_head_num, layer.head_dim),
                     v.view(-1, layer.tp_v_head_num, layer.v_head_dim),
@@ -428,8 +455,6 @@ class FlashInferAttnBackend(AttentionBackend):
                     logits_soft_cap=logits_soft_cap,
                 )
 
-            o = o1
-
                 if save_kv_cache:
                     forward_batch.token_to_kv_pool.set_kv_buffer(
                         layer,
@@ -437,6 +462,26 @@ class FlashInferAttnBackend(AttentionBackend):
                         k,
                         v,
                     )
+            else:
+                # use mla paged prefill
+                prefill_wrapper_paged = self.forward_metadata.prefill_wrappers[
+                    self._get_wrapper_idx(layer)
+                ]
+                if k is not None:
+                    assert v is not None
+                    if save_kv_cache:
+                        forward_batch.token_to_kv_pool.set_kv_buffer(
+                            layer, cache_loc, k, v
+                        )
+                qall = q.view(-1, layer.tp_q_head_num, layer.head_dim)
+                k_buf = forward_batch.token_to_kv_pool.get_key_buffer(layer.layer_id)
+
+                o = prefill_wrapper_paged.run(
+                    qall[:, :, : layer.v_head_dim],
+                    qall[:, :, layer.v_head_dim :],
+                    k_buf[:, :, : layer.v_head_dim],
+                    k_buf[:, :, layer.v_head_dim :],
+                )
 
             return o.view(-1, layer.tp_q_head_num * layer.v_head_dim)
         else:
@@ -800,7 +845,9 @@ class FlashInferIndicesUpdaterPrefill:
         seq_lens: torch.Tensor,
         seq_lens_sum: int,
         prefix_lens: torch.Tensor,
-        prefill_wrappers: List[BatchPrefillWithPagedKVCacheWrapper],
+        prefill_wrappers: List[
+            Union[BatchPrefillWithPagedKVCacheWrapper, BatchMLAPagedAttentionWrapper]
+        ],
         use_ragged: bool,
         encoder_lens: Optional[torch.Tensor],
         spec_info: Optional[SpecInfo],
@@ -814,7 +861,9 @@ class FlashInferIndicesUpdaterPrefill:
         seq_lens: torch.Tensor,
         seq_lens_sum: int,
         prefix_lens: torch.Tensor,
-        prefill_wrappers: List[BatchPrefillWithPagedKVCacheWrapper],
+        prefill_wrappers: List[
+            Union[BatchPrefillWithPagedKVCacheWrapper, BatchMLAPagedAttentionWrapper]
+        ],
         use_ragged: bool,
         encoder_lens: Optional[torch.Tensor],
         spec_info: Optional[SpecInfo],
@@ -923,7 +972,9 @@ class FlashInferIndicesUpdaterPrefill:
     def call_begin_forward(
         self,
         wrapper_ragged: BatchPrefillWithRaggedKVCacheWrapper,
-        wrapper_paged: BatchPrefillWithPagedKVCacheWrapper,
+        wrapper_paged: Union[
+            BatchPrefillWithPagedKVCacheWrapper, BatchMLAPagedAttentionWrapper
+        ],
         req_pool_indices: torch.Tensor,
         paged_kernel_lens: torch.Tensor,
         paged_kernel_lens_sum: int,
@@ -1004,6 +1055,26 @@ class FlashInferIndicesUpdaterPrefill:
                 custom_mask=custom_mask,
                 non_blocking=True,
             )
+        elif (
+            global_config.enable_flashinfer_mla
+            and not global_server_args_dict["disable_radix_cache"]
+        ):
+            # mla paged prefill
+            kv_len_arr = kv_indptr[1:] - kv_indptr[:-1]
+            wrapper_paged.plan(
+                qo_indptr,
+                kv_indptr,
+                kv_indices,
+                kv_len_arr,
+                self.num_qo_heads,
+                512,
+                64,
+                1,
+                True,
+                1 / math.sqrt(192),
+                self.data_type,
+                self.data_type,
+            )
 
 
 class FlashInferMultiStepDraftBackend:

python/sglang/srt/managers/schedule_batch.py

@@ -66,6 +66,7 @@ global_server_args_dict = {
     "enable_ep_moe": ServerArgs.enable_ep_moe,
     "device": ServerArgs.device,
     "enable_flashinfer_mla": ServerArgs.enable_flashinfer_mla,
+    "disable_radix_cache": ServerArgs.disable_radix_cache,
 }
 
 logger = logging.getLogger(__name__)

python/sglang/srt/model_executor/model_runner.py

@@ -177,6 +177,7 @@ class ModelRunner:
                 "enable_ep_moe": server_args.enable_ep_moe,
                 "device": server_args.device,
                 "enable_flashinfer_mla": server_args.enable_flashinfer_mla,
+                "disable_radix_cache": server_args.disable_radix_cache,
             }
         )
 

python/sglang/srt/models/deepseek_v2.py

@@ -511,10 +511,13 @@ class DeepseekV2AttentionMLA(nn.Module):
         forward_batch: ForwardBatch,
     ) -> torch.Tensor:
         if global_server_args_dict["enable_flashinfer_mla"]:
+            if global_server_args_dict["disable_radix_cache"]:
                 if forward_batch.forward_mode.is_extend():
                     return self.forward_normal(positions, hidden_states, forward_batch)
                 else:
                     return self.forward_absorb(positions, hidden_states, forward_batch)
+            else:
+                return self.forward_absorb(positions, hidden_states, forward_batch)
         else:
             # Triton: Use normal computation for prefill and use weight absorption for extend/decode
             if (