[MLA] Fuse cat and qaunt for fp8 kv-cache (#32950)

Signed-off-by: Lucas Wilkinson <lwilkins@redhat.com>

vllm/model_executor/layers/attention/mla_attention.py

@@ -202,13 +202,16 @@ from vllm._aiter_ops import rocm_aiter_ops
 from vllm.config import ModelConfig, VllmConfig, get_current_vllm_config
 from vllm.distributed.parallel_state import get_dcp_group, is_global_first_rank
 from vllm.logger import init_logger
+from vllm.model_executor.custom_op import CustomOp
 from vllm.model_executor.layers.batch_invariant import (
     vllm_is_batch_invariant,
 )
 from vllm.model_executor.layers.linear import (
     ColumnParallelLinear,
 )
+from vllm.model_executor.layers.quantization.input_quant_fp8 import QuantFP8
 from vllm.model_executor.layers.quantization.utils.quant_utils import (
+    GroupShape,
     get_and_maybe_dequant_weights,
 )
 from vllm.platforms import current_platform
@@ -287,6 +290,37 @@ def dynamic_per_batched_tensor_quant(
 
 logger = init_logger(__name__)
 
+
+@CustomOp.register("mla_decode_concat_quant_fp8")
+class _DecodeConcatQuantFP8(QuantFP8):
+    """
+    QuantFP8 variant that concatenates decode_ql_nope and decode_q_pe before
+    quantization. When disabled, forward_native is compiled via torch.compile,
+    fusing cat/reshape/quant/view together.
+    """
+
+    def _make_forward(quant_fn):  # noqa: N805
+        """Factory to create forward methods that concat before quantization."""
+
+        def forward(
+            self,
+            decode_ql_nope: torch.Tensor,
+            decode_q_pe: torch.Tensor,
+            scale: torch.Tensor,
+            scale_ub: torch.Tensor | None = None,
+        ) -> torch.Tensor:
+            decode_q0 = torch.cat((decode_ql_nope, decode_q_pe), dim=-1)
+            decode_q_flat = decode_q0.reshape(decode_q0.shape[0], -1)
+            decode_q, _ = quant_fn(self, decode_q_flat, scale, scale_ub)
+            return decode_q.view(decode_q0.shape)
+
+        return forward
+
+    forward_native = _make_forward(QuantFP8.forward_native)  # type: ignore[arg-type]
+    forward_cuda = _make_forward(QuantFP8.forward_cuda)  # type: ignore[arg-type]
+    forward_hip = _make_forward(QuantFP8.forward_hip)  # type: ignore[arg-type]
+
+
 CUDNN_WORKSPACE_SIZE = 12800
 
 
@@ -1398,6 +1432,11 @@ class MLACommonImpl(MLACommonBaseImpl[M], Generic[M]):
         self.cp_kv_cache_interleave_size: int = (
             get_current_vllm_config().parallel_config.cp_kv_cache_interleave_size
         )
+        self._decode_concat_quant_fp8_op = _DecodeConcatQuantFP8(
+            static=True,
+            group_shape=GroupShape.PER_TENSOR,
+            compile_native=True,
+        )
 
     def _flash_attn_varlen_diff_headdims(
         self, q, k, v, return_softmax_lse=False, softmax_scale=None, **kwargs
@@ -2048,29 +2087,11 @@ class MLACommonImpl(MLACommonBaseImpl[M], Generic[M]):
                 decode_ql_nope = decode_ql_nope.transpose(0, 1)
 
             if fp8_attention:
-                ql_nope_shape = decode_ql_nope.shape
-                q_pe_shape = decode_q_pe.shape
                 assert decode_ql_nope.shape[0] == decode_q_pe.shape[0]
                 assert decode_ql_nope.shape[1] == decode_q_pe.shape[1]
-                decode_q_shape = (
-                    ql_nope_shape[0],
-                    ql_nope_shape[1],
-                    ql_nope_shape[2] + q_pe_shape[2],
-                )
-                # Using empty and copy since torch.cat introduces significant overhead.
-                decode_q0 = torch.empty(
-                    decode_q_shape,
-                    device=decode_ql_nope.device,
-                    dtype=decode_ql_nope.dtype,
-                )
-                decode_q0[..., : ql_nope_shape[2]].copy_(decode_ql_nope)
-                decode_q0[..., ql_nope_shape[2] :].copy_(decode_q_pe)
-
-                decode_q, _ = ops.scaled_fp8_quant(
-                    decode_q0.view(decode_q_shape[0], -1),
-                    layer._q_scale,
+                decode_q = self._decode_concat_quant_fp8_op(
+                    decode_ql_nope, decode_q_pe, layer._q_scale
                 )
-                decode_q = decode_q.view(decode_q_shape)
             else:
                 decode_q = (decode_ql_nope, decode_q_pe)
             if self.dcp_world_size > 1: