[feat]优化mtp/eagle的计算逻辑，减少第1层并行解码的计算重复（num_accepted_tokens_tensor修改暂未合入）

vllm/v1/attention/backends/mla/common.py

@@ -754,12 +754,8 @@ class MLACommonMetadataBuilder(AttentionMetadataBuilder[M]):
 
         decode_metadata = None
         if num_decodes > 0:
-            if self.use_spec_decode:
+            if self.use_spec_decode and not common_attn_metadata.spec_layer_decoding:
                 query_lens = self.num_scheduled_tokens_np[:num_decodes]
-                if common_attn_metadata.num_rejected_tokens is not None:
-                    num_rejected_tokens = common_attn_metadata.num_rejected_tokens[:num_decodes]
-                    query_lens = query_lens - np.array(num_rejected_tokens, dtype=np.int32)
-                    self._num_decode_tokens -= sum(num_rejected_tokens)
                 cu_num_blocks = np.cumsum(query_lens)
                 virtual_batches = cu_num_blocks[-1]
                 block_offsets = np.repeat(cu_num_blocks - query_lens, query_lens)
@@ -790,10 +786,20 @@ class MLACommonMetadataBuilder(AttentionMetadataBuilder[M]):
                         seq_lens=decode_seq_lens,
                     )
             else:
-                decode_metadata = self._build_decode(
-                    block_table_tensor=block_table_tensor[:num_decodes, ...],
-                    seq_lens=seq_lens[:num_decodes],
-                )
+                self._num_decode_tokens = num_decodes
+                if self.use_spec_decode and self.spec_decode_block_table_tensor is not None:
+                    self.spec_decode_block_table_tensor[:self._num_decode_tokens].copy_(block_table_tensor[:self._num_decode_tokens, ...])
+                    self.spec_decode_seq_lens[:self._num_decode_tokens].copy_(seq_lens[:self._num_decode_tokens])
+
+                    decode_metadata = self._build_decode(
+                        block_table_tensor=self.spec_decode_block_table_tensor[:self._num_decode_tokens, ...],
+                        seq_lens=self.spec_decode_seq_lens[:self._num_decode_tokens],
+                    )
+                else:
+                    decode_metadata = self._build_decode(
+                        block_table_tensor=block_table_tensor[:self._num_decode_tokens, ...],
+                        seq_lens=seq_lens[:self._num_decode_tokens],
+                    )
 
         attn_metadata = self.metadata_cls(
             num_reqs=common_attn_metadata.num_reqs,

vllm/v1/attention/backends/utils.py

@@ -58,12 +58,11 @@ class CommonAttentionMetadata:
     
     block_table_tensor: torch.Tensor
     
-    num_rejected_tokens: list[int] = None
-    """(batch_size,), record the rejected tokens number in cpu and gpu"""
     num_speculative_tokens: int = 0
     """Number of speculative tokens"""
     slot_mapping: torch.Tensor = None
     """(batch_size, seq_len), slot mapping"""
+    spec_layer_decoding: bool = False
 
 
 M = TypeVar("M")

vllm/v1/spec_decode/eagle.py

@@ -98,6 +98,7 @@ class EagleProposer:
         next_token_ids: torch.Tensor,
         common_attn_metadata: CommonAttentionMetadata,
         sampling_metadata: SamplingMetadata,
+        decoding: bool = False,
     ) -> torch.Tensor:
         num_tokens = target_token_ids.shape[0]
         batch_size = next_token_ids.shape[0]
@@ -141,7 +142,7 @@ class EagleProposer:
         self.positions[:num_tokens] = target_positions
         self.hidden_states[:num_tokens] = target_hidden_states
         
-        if (self.use_full_cuda_graph
+        if (decoding and self.use_full_cuda_graph
                 and num_tokens <= self.cudagraph_batch_sizes[-1]):
             assert self.attn_metadata_cudagraph
             if self.method == "deepseek_mtp":
@@ -166,7 +167,8 @@ class EagleProposer:
 
         with set_forward_context(per_layer_attn_metadata,
                                  self.vllm_config,
-                                 num_tokens=num_input_tokens):
+                                 num_tokens=num_input_tokens,
+                                 skip_cuda_graphs=not decoding):
             ret_hidden_states = self.model(
                 self.input_ids[:num_input_tokens],
                 self.positions[:num_input_tokens],
@@ -329,9 +331,11 @@ class EagleProposer:
 
     def prepare_inputs(
         self,
+        # cu_target_query_lens: torch.Tensor,
         common_attn_metadata: CommonAttentionMetadata,
         # [batch_size]
-        num_rejected_tokens: torch.Tensor
+        num_rejected_tokens: torch.Tensor,
+        # num_accepted_tokens_tensor: torch.Tensor,
     ) -> tuple[CommonAttentionMetadata, torch.Tensor]:
         """
         This function is used to prepare the inputs for the spec decode.
@@ -403,6 +407,7 @@ class EagleProposer:
         token_indices_np = token_offests + old_query_start_locs_expanded
         token_indices = torch.from_numpy(token_indices_np).to(
             device, non_blocking=True)
+        # token_indices = num_accepted_tokens_tensor + cu_target_query_lens[:-1]
 
         spec_common_attn_metadata = CommonAttentionMetadata(
             query_start_loc=new_query_start_loc_cpu.to(device,

vllm/v1/worker/gpu_model_runner.py

@@ -1732,7 +1732,6 @@ class GPUModelRunner(LoRAModelRunnerMixin):
                                           dtype=torch.int32,
                                           device=self.device)
 
-            num_rejected_tokens = None
             if spec_decode_metadata is None:
                 # input_ids can be None for multimodal models.
                 target_token_ids = self.input_ids[:num_scheduled_tokens]
@@ -1757,6 +1756,19 @@ class GPUModelRunner(LoRAModelRunnerMixin):
                     self.drafter.prepare_inputs(
                     common_attn_metadata, num_rejected_tokens_cpu)
 
+                # num_accepted_tokens = [len(s) - 1 for s in sampled_token_ids]
+                # num_accepted_tokens_tensor = async_tensor_h2d(
+                #     num_accepted_tokens,
+                #     dtype=torch.int32,
+                #     target_device=self.device,
+                #     pin_memory=True)
+
+                # num_accepted_tokens_cpu = torch.tensor(num_accepted_tokens,
+                #                                        dtype=torch.int32)
+                # common_attn_metadata, token_indices =\
+                #     self.drafter.prepare_inputs(
+                #     common_attn_metadata, num_accepted_tokens_cpu)
+
                 target_token_ids = self.input_ids[token_indices]
                 # TODO(woosuk): Support M-RoPE.
                 target_positions = self.positions[token_indices]
@@ -1772,7 +1784,7 @@ class GPUModelRunner(LoRAModelRunnerMixin):
                 next_token_ids=next_token_ids,
                 sampling_metadata=sampling_metadata,
                 common_attn_metadata=common_attn_metadata,
-                num_rejected_tokens=num_rejected_tokens
+                decoding=spec_decode_metadata is not None
             )
             spec_token_ids = draft_token_ids.tolist()