Skip to content

[TRTLLM-16186][feat] Define backend contract for KV cache transfers #20117

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

Merged
Shixiaowei02 merged 3 commits into from
Oct 21, 2026
Merged

[TRTLLM-16186][feat] Define backend contract for KV cache transfers #20117

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
3 commits
Select commit Hold shift + click to select a range
409f727
[TRTLLM-16186][feat] Define backend contract for KV cache transfers
Shixiaowei02 Oct 17, 2026
2b24846
[None][fix] Keep read-only bool delivery masks in Outcome
Shixiaowei02 Oct 20, 2026
8ac96cf
[None][fix] Promise a new mask from every View.row_masks call
Shixiaowei02 Oct 20, 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
42 changes: 25 additions & 17 deletions tensorrt_llm/_torch/disaggregation/base/__init__.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -12,36 +12,44 @@
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
"""The cache-transfer contract.
"""The KV cache transfer backend contract.

What this package exports is the contract; what it does not is backend-internal and has to be
imported by file path.
The package exports the contract, plus ``TokenRange`` and ``CacheKind``, which belong to native and
the page table. All of them come from ``backend.py`` except ``MemoryKind``, which ``memory.py``
defines. The package's other modules are not part of the contract and are imported by their own
module paths, such as ``base.transfer``.
"""

from .backend import (
Attempt,
CacheExtent,
Backend,
CacheKind,
Cancelled,
Chunk,
Delivered,
Failed,
Fetches,
Capabilities,
Hint,
InvalidSubmission,
LayerGroupRows,
Outcome,
Publishes,
SlotRegion,
SubmissionRejected,
TokenRange,
View,
WakeFn,
)
from .memory import MemoryKind

__all__ = [
"Attempt",
"CacheExtent",
"Backend",
"CacheKind",
"Cancelled",
"Chunk",
"Delivered",
"Failed",
"Fetches",
"Capabilities",
"Hint",
"InvalidSubmission",
"LayerGroupRows",
"MemoryKind",
"Outcome",
"Publishes",
"SlotRegion",
"SubmissionRejected",
"TokenRange",
"View",
"WakeFn",
]
Loading
Loading