[Refactor][kv_offload] KV Offloading maintainability improvements (#40538)

Signed-off-by: Martin Hickey <martin.hickey@ie.ibm.com>
Co-authored-by: Or Ozeri <oro@il.ibm.com>
This commit is contained in:
Martin Hickey
2026-04-30 05:55:31 +03:00
committed by GitHub
co-authored by Or Ozeri
parent 0ff1bf9bb1
commit c42981d034
24 changed files with 446 additions and 440 deletions
@@ -15,7 +15,7 @@ from vllm.distributed.kv_transfer.kv_connector.v1.offloading.scheduler import (
OffloadingConnectorScheduler,
)
from vllm.v1.core.kv_cache_utils import BlockHash
from vllm.v1.kv_offload.abstract import (
from vllm.v1.kv_offload.base import (
OffloadingEvent,
OffloadingManager,
ReqContext,
@@ -20,7 +20,7 @@ from vllm.v1.kv_cache_interface import (
MLAAttentionSpec,
UniformTypeKVCacheSpecs,
)
from vllm.v1.kv_offload.spec import (
from vllm.v1.kv_offload.base import (
CanonicalKVCacheRef,
CanonicalKVCaches,
OffloadingSpec,
@@ -37,15 +37,15 @@ from vllm.v1.kv_cache_interface import (
KVCacheConfig,
KVCacheGroupSpec,
)
from vllm.v1.kv_offload.abstract import (
from vllm.v1.kv_offload.base import (
GPULoadStoreSpec,
LoadStoreSpec,
OffloadingManager,
OffloadingSpec,
OffloadKey,
PrepareStoreOutput,
make_offload_key,
)
from vllm.v1.kv_offload.mediums import GPULoadStoreSpec
from vllm.v1.kv_offload.spec import OffloadingSpec
from vllm.v1.kv_offload.worker.worker import (
OffloadingHandler,
TransferResult,
@@ -9,14 +9,15 @@ import torch
from vllm.platforms import current_platform
from vllm.utils.torch_utils import set_random_seed
from vllm.v1.kv_offload.cpu.shared_offload_region import SharedOffloadRegion
from vllm.v1.kv_offload.mediums import CPULoadStoreSpec, GPULoadStoreSpec
from vllm.v1.kv_offload.spec import (
from vllm.v1.kv_offload.base import (
CanonicalKVCacheRef,
CanonicalKVCaches,
CanonicalKVCacheTensor,
GPULoadStoreSpec,
)
from vllm.v1.kv_offload.worker.cpu_gpu import CpuGpuOffloadingHandlers
from vllm.v1.kv_offload.cpu.common import CPULoadStoreSpec
from vllm.v1.kv_offload.cpu.gpu_worker import CpuGpuOffloadingHandlers
from vllm.v1.kv_offload.cpu.shared_offload_region import SharedOffloadRegion
NUM_GPU_BLOCKS = [64]
NUM_CPU_BLOCKS = [256]
@@ -6,7 +6,7 @@ from dataclasses import dataclass
import numpy as np
import pytest
from vllm.v1.kv_offload.abstract import (
from vllm.v1.kv_offload.base import (
LoadStoreSpec,
OffloadingEvent,
OffloadKey,
@@ -14,9 +14,9 @@ from vllm.v1.kv_offload.abstract import (
ReqContext,
make_offload_key,
)
from vllm.v1.kv_offload.cpu.common import CPULoadStoreSpec
from vllm.v1.kv_offload.cpu.manager import CPUOffloadingManager
from vllm.v1.kv_offload.cpu.policies.arc import ARCCachePolicy
from vllm.v1.kv_offload.mediums import CPULoadStoreSpec
from vllm.v1.kv_offload.reuse_manager import FilterReusedOffloadingManager
+1 -1
View File
@@ -1,6 +1,6 @@
# SPDX-License-Identifier: Apache-2.0
# SPDX-FileCopyrightText: Copyright contributors to the vLLM project
from vllm.v1.kv_offload.abstract import LoadStoreSpec
from vllm.v1.kv_offload.base import LoadStoreSpec
from vllm.v1.kv_offload.worker.worker import (
OffloadingHandler,
OffloadingWorker,