Merge branch 'v0.15.1-fused_fill_rms_cat' into 'v0.15.1-dev'

feat:接入VLLM_USE_FUSED_FILL_RMS_CAT优化

See merge request dcutoolkit/deeplearing/vllm!512

vllm/model_executor/models/deepseek_mtp.py

@@ -41,7 +41,7 @@ from .interfaces import SupportsPP
 from vllm import _custom_ops as ops
 from vllm.model_executor.layers.quantization.blockwise_int8 import BlockInt8Config
 import vllm.envs as envs
-
+from vllm.utils import direct_register_custom_op
 logger = init_logger(__name__)
 
 
@@ -102,6 +102,24 @@ class DeepSeekMultiTokenPredictorLayer(nn.Module):
             config=self.config,
             topk_indices_buffer=topk_indices_buffer,
         )
+    def fuse_fill_rms_x2_concat(hidden_states_fuse: torch.Tensor, positions: torch.Tensor, inputs_embeds: torch.Tensor,
+                                    previous_hidden_states: torch.Tensor, weight_inputs_embeds: torch.Tensor, 
+                                    weight_previous_hidden_states: torch.Tensor, epsilon: float) -> None:
+        from lightop import fuse_fill_rms_x2_concat
+        fuse_fill_rms_x2_concat(hidden_states_fuse, positions, inputs_embeds, previous_hidden_states, weight_inputs_embeds, weight_previous_hidden_states, epsilon)
+
+    def fuse_fill_rms_x2_concat_fake(hidden_states_fuse: torch.Tensor, positions: torch.Tensor, inputs_embeds: torch.Tensor,
+                                    previous_hidden_states: torch.Tensor, weight_inputs_embeds: torch.Tensor, 
+                                    weight_previous_hidden_states: torch.Tensor, epsilon: float) -> None:
+        pass
+
+    direct_register_custom_op(
+        op_name="fuse_fill_rms_x2_concat",
+        op_func=fuse_fill_rms_x2_concat,
+        mutates_args=["hidden_states_fuse", "inputs_embeds"], 
+        fake_impl=fuse_fill_rms_x2_concat_fake,
+    )
+
 
     def forward(
         self,