[None][chore] Remove the disagg transfer admission controller

tensorrt_llm/_torch/disaggregation/executor/coordinator.py

@@ -8,7 +8,7 @@
 """
 
 from dataclasses import dataclass, fields
-from typing import TYPE_CHECKING, Callable, List, Tuple
+from typing import TYPE_CHECKING, Callable, List
 
 from tensorrt_llm._torch.pyexecutor.llm_request import LlmRequest
 
@@ -28,7 +28,6 @@ class DisaggLoopDelegates:
     prepare_context_schedulable: Callable[[List[LlmRequest]], None]
     poll_gen_transfers: Callable[[], None]
     check_transfer_timeouts: Callable[[], None]
-    admit: Callable[[List[LlmRequest]], Tuple[List[LlmRequest], bool]]
     revert_deferred_gen_init: Callable[[List[LlmRequest], List[LlmRequest]], None]
     receive_gen_init: Callable[[List[LlmRequest]], None]
     poll_progress_when_idle: Callable[[], None]
@@ -71,13 +70,6 @@ def check_transfer_timeouts(self) -> None:
 
     # -- scheduling ----------------------------------------------------------
 
-    def admit(self, fitting_gen_init: List[LlmRequest]) -> Tuple[List[LlmRequest], bool]:
-        """Select the gen-init requests that may start receiving this iteration.
-
-        Returns ``(admitted, blocked_by_active_transfers)``.
-        """
-        return self._d.admit(fitting_gen_init)
-
     def revert_deferred_gen_init(
         self, candidates: List[LlmRequest], admitted: List[LlmRequest]
     ) -> None:
@@ -121,9 +113,6 @@ def __init__(self) -> None:
             DisaggLoopDelegates(**{f.name: _noop for f in fields(DisaggLoopDelegates)})
         )
 
-    def admit(self, fitting_gen_init: List[LlmRequest]) -> Tuple[List[LlmRequest], bool]:
-        return fitting_gen_init, False
-
 
 def _noop(*_args, **_kwargs) -> None:
     return None

tensorrt_llm/_torch/disaggregation/kv_cache_transceiver.py

@@ -302,11 +302,6 @@ def has_retired_send_session(self, req: LlmRequest) -> bool:
         """Whether the send session closed before its final slice."""
         return False
 
-    @property
-    def consumes_transfer_buffer(self) -> bool:
-        """Return whether this runtime consumes the C++ CacheTransBuffer budget."""
-        return True
-
     @abstractmethod
     def respond_and_send_async(self, req: LlmRequest) -> None:
         """Start sending ``req``'s KV cache to the requesting instance.

tensorrt_llm/_torch/disaggregation/transceiver.py

@@ -82,10 +82,6 @@ def _find_consensus_request_ids(request_ids_all_ranks, sync_size):
 
 
 class KvCacheTransceiverV2(KvCacheTransceiver):
-    @property
-    def consumes_transfer_buffer(self) -> bool:
-        return False
-
     def __init__(
         self,
         mapping: Mapping,