[NVIDIA] Add Cutlass MLA backend (#17625)

csrc/attention/mla/cutlass_mla_kernels.cu

@@ -119,7 +119,7 @@ typename T::Fmha::Arguments args_from_options(
       {static_cast<ElementOut*>(out.data_ptr()), stride_O,
        static_cast<ElementAcc*>(nullptr), stride_LSE},
       hw_info,
-      -1,       // split_kv
+      1,        // split_kv
       nullptr,  // is_var_split_kv
   };
   // TODO(kaixih@nvidia): When split_kv=-1 and is_var_split_kv=false, we compute

tests/kernels/test_cutlass_mla_decode.py

@@ -76,7 +76,9 @@ def test_cutlass_mla_decode(dtype: torch.dtype, mean_seq_len: int, bs: int,
     pack_factor = 128 // block_size
     block_num = ((block_num + pack_factor - 1) // pack_factor) * pack_factor
 
-    q = torch.randn(bs, h_q, d)
+    # Amplify input values to ensure test coverage of edge cases where CUTLASS
+    # kernel errors occur with split_k settings.
+    q = torch.randn(bs, h_q, d) * 100
     block_table = torch.randint(0,
                                 bs * block_num, (bs, block_num),
                                 dtype=torch.int32)

vllm/engine/arg_utils.py

@@ -1395,6 +1395,7 @@ class EngineArgs:
             "PALLAS_VLLM_V1",
             "TRITON_ATTN_VLLM_V1",
             "TRITON_MLA",
+            "CUTLASS_MLA_VLLM_V1",
             "FLASHMLA",
             "FLASHINFER",
             "FLASHINFER_VLLM_V1",

vllm/platforms/cuda.py

@@ -183,6 +183,14 @@ class CudaPlatformBase(Platform):
         if use_mla:
             # TODO(lucas): refactor to  be more concise
             #  we should probably consider factoring out V1 here
+            if selected_backend == _Backend.CUTLASS_MLA_VLLM_V1:
+                if use_v1:
+                    logger.info_once("Using Cutlass MLA backend on V1 engine.")
+                    return ("vllm.v1.attention.backends.mla."
+                            "cutlass_mla.CutlassMLABackend")
+                else:
+                    logger.warning(
+                        "Cutlass MLA backend is only supported on V1 engine")
             if selected_backend == _Backend.TRITON_MLA or block_size != 64:
                 if use_v1:
                     logger.info_once("Using Triton MLA backend on V1 engine.")

vllm/platforms/interface.py

@@ -51,6 +51,7 @@ class _Backend(enum.Enum):
     TRITON_MLA_VLLM_V1 = enum.auto()
     FLASHMLA_VLLM_V1 = enum.auto()
     FLASHMLA = enum.auto()  # Supported by V1
+    CUTLASS_MLA_VLLM_V1 = enum.auto()
     HPU_ATTN = enum.auto()
     PALLAS = enum.auto()
     PALLAS_VLLM_V1 = enum.auto()

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

@@ -350,7 +350,7 @@ class MLACommonMetadataBuilder(Generic[M]):
         self.num_heads = model_config.get_num_attention_heads(
             runner.parallel_config)
         self.mla_dims = get_mla_dims(model_config)
-        self.aot_schedule = is_vllm_fa and (get_flash_attn_version() == 3)
+        self.aot_schedule = current_platform.is_cuda()
         self.kv_cache_spec = kv_cache_spec
 
         # Dont try to access the runner on AMD

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

@@ -0,0 +1,96 @@
+# SPDX-License-Identifier: Apache-2.0
+
+from typing import Any, Optional
+
+import torch
+
+import vllm._custom_ops as ops
+from vllm.attention.backends.abstract import (AttentionType,
+                                              is_quantized_kv_cache)
+from vllm.logger import init_logger
+from vllm.v1.attention.backends.mla.common import (MLACommonBackend,
+                                                   MLACommonImpl,
+                                                   MLACommonMetadata)
+
+logger = init_logger(__name__)
+
+
+class CutlassMLABackend(MLACommonBackend):
+
+    @staticmethod
+    def get_name() -> str:
+        return "CUTLASS_MLA_VLLM_V1"
+
+    @staticmethod
+    def get_impl_cls() -> type["CutlassMLAImpl"]:
+        return CutlassMLAImpl
+
+
+class CutlassMLAImpl(MLACommonImpl[MLACommonMetadata]):
+
+    def __init__(
+            self,
+            num_heads: int,
+            head_size: int,
+            scale: float,
+            num_kv_heads: int,
+            alibi_slopes: Optional[list[float]],
+            sliding_window: Optional[int],
+            kv_cache_dtype: str,
+            blocksparse_params: Optional[dict[str, Any]],
+            logits_soft_cap: Optional[float],
+            attn_type: str,
+            # MLA Specific Arguments
+            **mla_args) -> None:
+        super().__init__(num_heads, head_size, scale, num_kv_heads,
+                         alibi_slopes, sliding_window, kv_cache_dtype,
+                         blocksparse_params, logits_soft_cap, attn_type,
+                         **mla_args)
+
+        unsupported_features = [
+            alibi_slopes, sliding_window, blocksparse_params, logits_soft_cap
+        ]
+        if any(unsupported_features):
+            raise NotImplementedError(
+                "CutlassMLAImpl does not support one of the following: "
+                "alibi_slopes, sliding_window, blocksparse_params, "
+                "logits_soft_cap")
+
+        if attn_type != AttentionType.DECODER:
+            raise NotImplementedError("Encoder self-attention and "
+                                      "encoder/decoder cross-attention "
+                                      "are not implemented for "
+                                      "CutlassMLAImpl")
+
+        if is_quantized_kv_cache(self.kv_cache_dtype):
+            raise NotImplementedError(
+                "CutlassMLA V1 with FP8 KV cache not yet supported")
+
+    def _forward_decode(
+        self,
+        q_nope: torch.Tensor,
+        q_pe: torch.Tensor,
+        kv_c_and_k_pe_cache: torch.Tensor,
+        attn_metadata: MLACommonMetadata,
+    ) -> torch.Tensor:
+        assert kv_c_and_k_pe_cache.numel() > 0
+        assert attn_metadata.decode is not None
+
+        if self.kv_cache_dtype.startswith("fp8"):
+            raise NotImplementedError("FP8 Cutlass MLA not yet supported")
+
+        B = q_nope.shape[0]
+
+        o = torch.empty((B, self.num_heads, self.kv_lora_rank),
+                        dtype=q_nope.dtype,
+                        device=q_nope.device)
+
+        # Run MLA
+        # Clone q_nope and q_pe to make sure strides computation is correct.
+        q_nope = q_nope.clone()
+        q_pe = q_pe.clone()
+        ops.cutlass_mla_decode(o, q_nope, q_pe, kv_c_and_k_pe_cache,
+                               attn_metadata.decode.seq_lens,
+                               attn_metadata.decode.block_table, self.scale)
+
+        return self._v_up_proj(o)