[Core][Refactor] Move KV cache planning logic to kv_cache_planning.py (#61183)
Signed-off-by: Dakai An <dakaian108@gmail.com> Signed-off-by: Yifan Qiao <yifanqiao@inferact.ai> Co-authored-by: Yifan Qiao <yifanqiao@inferact.ai> Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
This commit is contained in:
co-authored by
Yifan Qiao
Claude Opus 5.5
parent
c077dc663d
commit
df6c7ad031
@@ -394,7 +394,7 @@ def test_k3_dspark_mla_kv_cache_spec_groups_with_target_mla():
|
||||
"""The draft's MLA layers must share a KV cache group with the target's."""
|
||||
from vllm.model_executor.layers.attention.mla_attention import MLAAttention
|
||||
from vllm.models.kimi_k3.nvidia.mla import MultiHeadLatentAttention
|
||||
from vllm.v1.core.kv_cache_utils import _get_kv_cache_groups_uniform_page_size
|
||||
from vllm.v1.core.kv_cache_planning import _get_kv_cache_groups_uniform_page_size
|
||||
|
||||
vllm_config = SimpleNamespace(
|
||||
model_config=None, cache_config=SimpleNamespace(block_size=64)
|
||||
|
||||
@@ -14,10 +14,8 @@ from vllm.logger import init_logger
|
||||
from vllm.platforms import current_platform
|
||||
from vllm.utils.mem_constants import GiB_bytes
|
||||
from vllm.v1.attention.backends.utils import resolve_kv_cache_layout
|
||||
from vllm.v1.core.kv_cache_utils import (
|
||||
generate_scheduler_kv_cache_config,
|
||||
get_kv_cache_configs,
|
||||
)
|
||||
from vllm.v1.core.kv_cache_planning import get_kv_cache_configs
|
||||
from vllm.v1.core.kv_cache_utils import generate_scheduler_kv_cache_config
|
||||
from vllm.v1.engine.core import EngineCore as V1EngineCore
|
||||
|
||||
from ..utils import (
|
||||
|
||||
Reference in New Issue
Block a user