Relocate Encoder CUDA graph manager (#38116)

Signed-off-by: Woosuk Kwon <woosuk@inferact.ai>
Signed-off-by: Nick Hill <nickhill123@gmail.com>
Co-authored-by: Nick Hill <nickhill123@gmail.com>

tests/v1/cudagraph/test_encoder_cudagraph.py

@@ -14,8 +14,6 @@ from typing import Any
 
 import pytest
 import torch
-
-from vllm.platforms import current_platform
 from vllm.v1.worker.gpu.mm.encoder_cudagraph import (
     EncoderCudaGraphManager,
 )
@@ -25,6 +23,8 @@ from vllm.v1.worker.gpu.mm.encoder_cudagraph_defs import (
     EncoderCudaGraphReplayBuffers,
 )
 
+from vllm.platforms import current_platform
+
 # ---------------------------------------------------------------------------
 # Helpers
 # ---------------------------------------------------------------------------

vllm/v1/worker/gpu_model_runner.py

@@ -208,7 +208,7 @@ from .utils import (
 if TYPE_CHECKING:
     from vllm.v1.core.sched.output import GrammarOutput, SchedulerOutput
     from vllm.v1.spec_decode.ngram_proposer import NgramProposer
-    from vllm.v1.worker.gpu.mm.encoder_cudagraph import EncoderCudaGraphManager
+    from vllm.v1.worker.encoder_cudagraph import EncoderCudaGraphManager
 
 logger = init_logger(__name__)
 
@@ -5972,9 +5972,7 @@ class GPUModelRunner(
                 SupportsEncoderCudaGraph,
                 supports_encoder_cudagraph,
             )
-            from vllm.v1.worker.gpu.mm.encoder_cudagraph import (
-                EncoderCudaGraphManager,
-            )
+            from vllm.v1.worker.encoder_cudagraph import EncoderCudaGraphManager
 
             raw_model = self.get_model()
             if supports_encoder_cudagraph(raw_model):