[DSA][MLA] Tiny refactor on DeepSeek to make it reusable for different backends (vllm-project#26656)

Signed-off-by: MengqingCao <[email protected]>

Author: MengqingCao

vllm/attention/layer.py

@@ -587,6 +587,7 @@ def __init__(
         prefix: str = "",
         use_sparse: bool = False,
         indexer: object | None = None,
+        **extra_impl_args,
     ):
         super().__init__()
         self.num_heads = num_heads
@@ -639,6 +640,7 @@ def __init__(
             v_head_dim=self.v_head_dim,
             kv_b_proj=kv_b_proj,
             indexer=indexer,
+            **extra_impl_args,
         )
 
         self.use_direct_call = not current_platform.opaque_attention_op()

vllm/model_executor/models/deepseek_mtp.py

@@ -17,9 +17,13 @@
     VocabParallelEmbedding,
 )
 from vllm.model_executor.model_loader.weight_utils import default_weight_loader
+from vllm.platforms import current_platform
 from vllm.sequence import IntermediateTensors
 
-from .deepseek_v2 import DeepseekV2DecoderLayer, get_spec_layer_idx_from_weight_name
+from .deepseek_v2 import (
+    DeepseekV2DecoderLayer,
+    get_spec_layer_idx_from_weight_name,
+)
 from .interfaces import SupportsPP
 from .utils import maybe_prefix
 
@@ -56,14 +60,16 @@ def __init__(self, vllm_config: VllmConfig, prefix: str) -> None:
         self.hnorm = RMSNorm(config.hidden_size, eps=config.rms_norm_eps)
         self.eh_proj = nn.Linear(config.hidden_size * 2, config.hidden_size, bias=False)
 
+        self.device = current_platform.device_type
+
         self.is_v32 = hasattr(config, "index_topk")
         if self.is_v32:
             topk_tokens = config.index_topk
             topk_indices_buffer = torch.empty(
                 vllm_config.scheduler_config.max_num_batched_tokens,
                 topk_tokens,
                 dtype=torch.int32,
-                device="cuda",
+                device=self.device,
             )
         else:
             topk_indices_buffer = None

vllm/model_executor/models/deepseek_v2.py

@@ -1165,6 +1165,7 @@ def __init__(self, *, vllm_config: VllmConfig, prefix: str = ""):
         config = vllm_config.model_config.hf_config
         quant_config = vllm_config.quant_config
         self.config = config
+        self.device = current_platform.device_type
 
         self.vocab_size = config.vocab_size
         self.is_v32 = hasattr(config, "index_topk")
@@ -1174,7 +1175,7 @@ def __init__(self, *, vllm_config: VllmConfig, prefix: str = ""):
                 vllm_config.scheduler_config.max_num_batched_tokens,
                 topk_tokens,
                 dtype=torch.int32,
-                device="cuda",
+                device=self.device,
             )
         else:
             topk_indices_buffer = None