From 4dd49b06f81af2238ac5a86cfb0b7220083eb125 Mon Sep 17 00:00:00 2001 From: Greg Pereira Date: Sun, 5 Apr 2026 12:11:58 -0700 Subject: [PATCH] [Bug] Fix Import paths for `encoder_cudagraph` modules (#38997) Signed-off-by: greg pereira Signed-off-by: Robert Shaw <114415538+robertgshaw2-redhat@users.noreply.github.com> Co-authored-by: Robert Shaw <114415538+robertgshaw2-redhat@users.noreply.github.com> --- tests/v1/cudagraph/test_encoder_cudagraph.py | 8 ++++---- vllm/model_executor/models/interfaces.py | 2 +- vllm/model_executor/models/qwen3_vl.py | 6 +++--- vllm/v1/worker/encoder_cudagraph.py | 2 +- vllm/v1/worker/gpu_model_runner.py | 4 ++-- 5 files changed, 11 insertions(+), 11 deletions(-) diff --git a/tests/v1/cudagraph/test_encoder_cudagraph.py b/tests/v1/cudagraph/test_encoder_cudagraph.py index 322fcb3caa..543dfc8bb3 100644 --- a/tests/v1/cudagraph/test_encoder_cudagraph.py +++ b/tests/v1/cudagraph/test_encoder_cudagraph.py @@ -14,17 +14,17 @@ from typing import Any import pytest import torch -from vllm.v1.worker.gpu.mm.encoder_cudagraph import ( + +from vllm.platforms import current_platform +from vllm.v1.worker.encoder_cudagraph import ( EncoderCudaGraphManager, ) -from vllm.v1.worker.gpu.mm.encoder_cudagraph_defs import ( +from vllm.v1.worker.encoder_cudagraph_defs import ( EncoderCudaGraphCaptureInputs, EncoderCudaGraphConfig, EncoderCudaGraphReplayBuffers, ) -from vllm.platforms import current_platform - # --------------------------------------------------------------------------- # Helpers # --------------------------------------------------------------------------- diff --git a/vllm/model_executor/models/interfaces.py b/vllm/model_executor/models/interfaces.py index df7170aab3..1f1d57493c 100644 --- a/vllm/model_executor/models/interfaces.py +++ b/vllm/model_executor/models/interfaces.py @@ -46,7 +46,7 @@ if TYPE_CHECKING: from vllm.multimodal.inputs import MultiModalFeatureSpec from vllm.multimodal.registry import _ProcessorFactories from vllm.sequence import IntermediateTensors - from vllm.v1.worker.gpu.mm.encoder_cudagraph_defs import ( + from vllm.v1.worker.encoder_cudagraph_defs import ( EncoderCudaGraphCaptureInputs, EncoderCudaGraphConfig, EncoderCudaGraphReplayBuffers, diff --git a/vllm/model_executor/models/qwen3_vl.py b/vllm/model_executor/models/qwen3_vl.py index cb48ceb0c7..1aa5dec539 100644 --- a/vllm/model_executor/models/qwen3_vl.py +++ b/vllm/model_executor/models/qwen3_vl.py @@ -1733,7 +1733,7 @@ class Qwen3VLForConditionalGeneration( # -- SupportsEncoderCudaGraph protocol methods -- def get_encoder_cudagraph_config(self): - from vllm.v1.worker.gpu.mm.encoder_cudagraph_defs import ( + from vllm.v1.worker.encoder_cudagraph_defs import ( EncoderCudaGraphConfig, ) @@ -1818,7 +1818,7 @@ class Qwen3VLForConditionalGeneration( device: torch.device, dtype: torch.dtype, ): - from vllm.v1.worker.gpu.mm.encoder_cudagraph_defs import ( + from vllm.v1.worker.encoder_cudagraph_defs import ( EncoderCudaGraphCaptureInputs, ) @@ -1872,7 +1872,7 @@ class Qwen3VLForConditionalGeneration( mm_kwargs: dict[str, Any], max_batch_size: int, ): - from vllm.v1.worker.gpu.mm.encoder_cudagraph_defs import ( + from vllm.v1.worker.encoder_cudagraph_defs import ( EncoderCudaGraphReplayBuffers, ) diff --git a/vllm/v1/worker/encoder_cudagraph.py b/vllm/v1/worker/encoder_cudagraph.py index b2930a2347..0fabbc77c0 100644 --- a/vllm/v1/worker/encoder_cudagraph.py +++ b/vllm/v1/worker/encoder_cudagraph.py @@ -16,7 +16,7 @@ from vllm.distributed import ( from vllm.logger import init_logger from vllm.model_executor.models.interfaces import SupportsEncoderCudaGraph from vllm.model_executor.models.vision import get_load_balance_assignment -from vllm.v1.worker.gpu.mm.encoder_cudagraph_defs import ( +from vllm.v1.worker.encoder_cudagraph_defs import ( EncoderCudaGraphConfig, ) diff --git a/vllm/v1/worker/gpu_model_runner.py b/vllm/v1/worker/gpu_model_runner.py index 7a21117fb6..8dfa65da13 100644 --- a/vllm/v1/worker/gpu_model_runner.py +++ b/vllm/v1/worker/gpu_model_runner.py @@ -211,7 +211,7 @@ 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__) @@ -5988,7 +5988,7 @@ class GPUModelRunner( SupportsEncoderCudaGraph, supports_encoder_cudagraph, ) - from vllm.v1.worker.gpu.mm.encoder_cudagraph import ( + from vllm.v1.worker.encoder_cudagraph import ( EncoderCudaGraphManager, )