Skip to content

[FIX_FOR_VLLM_LATEST] adapt to upstream OffloadingSpec kv_cache_config refactor (GAUDISW-247307) #1159

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Closed
tzielinski-habana wants to merge 2 commits into from
Closed

[FIX_FOR_VLLM_LATEST] adapt to upstream OffloadingSpec kv_cache_config refactor (GAUDISW-247307) #1159

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
6d11095
fix: [vllm-hourly] adapt to upstream OffloadingSpec kv_cache_config r…
tzielinski-habana Mar 13, 2026
df890a8
Remove linter cosmetic changes, keep only functional fixes
tzielinski-habana Mar 13, 2026
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
20 changes: 18 additions & 2 deletions tests/unit_tests/kv_offload/test_offloading_connector.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -27,7 +27,7 @@
init_none_hash,
)
from vllm.v1.core.sched.scheduler import Scheduler
from vllm.v1.kv_cache_interface import KVCacheConfig
from vllm.v1.kv_cache_interface import FullAttentionSpec, KVCacheConfig, KVCacheGroupSpec
from vllm.v1.kv_offload.abstract import (
LoadStoreSpec,
OffloadingEvent,
Expand Down Expand Up @@ -160,8 +160,24 @@ def __init__(self, offloaded_block_size: int, gpu_block_size: int, num_gpu_block
},
)

kv_cache_config = KVCacheConfig(
num_blocks=num_gpu_blocks,
kv_cache_tensors=[],
kv_cache_groups=[
KVCacheGroupSpec(
["layer"],
FullAttentionSpec(
block_size=gpu_block_size,
num_kv_heads=1,
head_size=1,
dtype=torch.float32,
),
)
],
)

self.scheduler: Scheduler = create_scheduler(vllm_config, num_blocks=num_gpu_blocks)
self.worker_connector = OffloadingConnector(vllm_config, KVConnectorRole.WORKER)
self.worker_connector = OffloadingConnector(vllm_config, KVConnectorRole.WORKER, kv_cache_config)

# register worker kv_caches to enable OffloadingWorker creations
self.worker_connector.register_cross_layers_kv_cache(
Expand Down
6 changes: 4 additions & 2 deletions vllm_gaudi/v1/kv_offload/worker/cpu_hpu.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -360,10 +360,12 @@ def get_handlers(
attn_backends: dict[str, type[AttentionBackend]],
) -> Iterator[tuple[type[LoadStoreSpec], type[LoadStoreSpec], OffloadingHandler]]:
if not self._handlers:
assert len(self.gpu_block_size) == 1
gpu_block_size = self.gpu_block_size[0]
self._handlers = CpuGpuOffloadingHandlers(
attn_backends=attn_backends,
gpu_block_size=self.gpu_block_size,
cpu_block_size=self.offloaded_block_size,
gpu_block_size=gpu_block_size,
cpu_block_size=gpu_block_size * self.block_size_factor,
Comment on lines 362 to +368
num_cpu_blocks=self.num_blocks,
gpu_caches=kv_caches,
)
Expand Down
Loading