[kv_offload+HMA][3/N]: Remove block_size from KVEvents (#36644)

Signed-off-by: Or Ozeri <oro@il.ibm.com>
This commit is contained in:
Or Ozeri
2026-04-15 11:53:19 +03:00
committed by GitHub
parent 431cea3eea
commit 235e1f930a
6 changed files with 10 additions and 38 deletions
@@ -424,7 +424,7 @@ class OffloadingConnectorScheduler:
parent_block_hash=None,
token_ids=[],
lora_id=None,
block_size=event.block_size,
block_size=0,
medium=event.medium,
lora_name=None,
)
-1
View File
@@ -79,7 +79,6 @@ class PrepareStoreOutput:
@dataclass
class OffloadingEvent:
keys: list[OffloadKey]
block_size: int
medium: str
# True if blocks are removed, False if stored
removed: bool
-4
View File
@@ -33,12 +33,10 @@ class CPUOffloadingManager(OffloadingManager):
def __init__(
self,
block_size: int,
num_blocks: int,
cache_policy: Literal["lru", "arc"] = "lru",
enable_events: bool = False,
):
self.block_size: int = block_size
self.medium: str = CPULoadStoreSpec.medium()
self._num_blocks: int = num_blocks
self._num_allocated_blocks: int = 0
@@ -145,7 +143,6 @@ class CPUOffloadingManager(OffloadingManager):
self.events.append(
OffloadingEvent(
keys=to_evict,
block_size=self.block_size,
medium=self.medium,
removed=True,
)
@@ -188,7 +185,6 @@ class CPUOffloadingManager(OffloadingManager):
self.events.append(
OffloadingEvent(
keys=stored_keys,
block_size=self.block_size,
medium=self.medium,
removed=False,
)
-5
View File
@@ -60,12 +60,7 @@ class CPUOffloadingSpec(OffloadingSpec):
kv_events_config is not None and kv_events_config.enable_kv_cache_events
)
assert len(self.gpu_block_size) == 1
gpu_block_size = self.gpu_block_size[0]
offloaded_block_size = gpu_block_size * self.block_size_factor
self._manager = CPUOffloadingManager(
block_size=offloaded_block_size,
num_blocks=self.num_blocks,
cache_policy=self.eviction_policy, # type: ignore[arg-type]
enable_events=enable_events,