[None][chore] Remove is_disaggregated param in executor request queue (#9049)

Signed-off-by: Patrice Castonguay <[email protected]>

Author: pcastonguay

tensorrt_llm/_torch/pyexecutor/executor_request_queue.py

@@ -50,7 +50,7 @@ class ExecutorRequestQueue:
     def __init__(self, dist: Distributed, enable_attention_dp: bool,
                  max_batch_size: int, max_beam_width: int,
                  max_num_active_requests: int, enable_iter_perf_stats: bool,
-                 batch_wait_timeout_ms: float, is_disaggregated: bool):
+                 batch_wait_timeout_ms: float):
         self.dist = dist
         self.request_queue: queue.Queue[RequestQueueItem] = queue.Queue()
         self.waiting_queue: deque[RequestQueueItem] = deque()
@@ -59,7 +59,6 @@ def __init__(self, dist: Distributed, enable_attention_dp: bool,
         self.max_batch_size = max_batch_size
         self.max_beam_width = max_beam_width
         self.max_num_active_requests = max_num_active_requests
-        self.is_disaggregated = is_disaggregated
         self.enqueue_lock = threading.Lock()
         self.next_request_id = max_batch_size
         self.enable_iter_perf_stats = enable_iter_perf_stats

tensorrt_llm/_torch/pyexecutor/py_executor.py

@@ -254,7 +254,6 @@ def __init__(self,
             max_num_active_requests=self.max_num_active_requests,
             enable_iter_perf_stats=self.enable_iter_perf_stats,
             batch_wait_timeout_ms=self.batch_wait_timeout_ms,
-            is_disaggregated=kv_cache_transceiver is not None,
         )
         self.executor_request_queue.set_exclude_last_generation_logits(
             self.disable_overlap_scheduler, self.dist.pp_size)

tests/unittest/_torch/executor/test_executor_request_queue.py

@@ -42,8 +42,7 @@ def executor_queue(mock_dist):
                                 max_beam_width=1,
                                 max_num_active_requests=16,
                                 enable_iter_perf_stats=True,
-                                batch_wait_timeout_ms=0.0,
-                                is_disaggregated=False)
+                                batch_wait_timeout_ms=0.0)
 
 
 @pytest.fixture
@@ -55,8 +54,7 @@ def integration_queue(mock_dist):
                                 max_beam_width=2,
                                 max_num_active_requests=8,
                                 enable_iter_perf_stats=True,
-                                batch_wait_timeout_ms=0.0,
-                                is_disaggregated=False)
+                                batch_wait_timeout_ms=0.0)
 
 
 def test_executor_queue_init(executor_queue, mock_dist):
@@ -65,7 +63,6 @@ def test_executor_queue_init(executor_queue, mock_dist):
     assert not executor_queue.enable_attention_dp
     assert executor_queue.max_beam_width == 1
     assert executor_queue.max_num_active_requests == 16
-    assert not executor_queue.is_disaggregated
     assert executor_queue.next_request_id == 8
     assert executor_queue.enable_iter_perf_stats
     assert executor_queue.active
@@ -124,8 +121,7 @@ def test_merge_helix_requests_with_padding(mock_dist):
                                               max_beam_width=1,
                                               max_num_active_requests=16,
                                               enable_iter_perf_stats=True,
-                                              batch_wait_timeout_ms=0.0,
-                                              is_disaggregated=True)
+                                              batch_wait_timeout_ms=0.0)
 
         # Mock _should_exclude_last_generation_logits.
         with patch.object(executor_queue,
@@ -181,8 +177,7 @@ def test_merge_helix_requests_without_padding(mock_dist):
                                               max_beam_width=1,
                                               max_num_active_requests=16,
                                               enable_iter_perf_stats=True,
-                                              batch_wait_timeout_ms=0.0,
-                                              is_disaggregated=True)
+                                              batch_wait_timeout_ms=0.0)
 
         # Mock _should_exclude_last_generation_logits.
         with patch.object(executor_queue,
@@ -235,8 +230,7 @@ def test_merge_helix_requests_insufficient_blocks_error(mock_dist):
                                               max_beam_width=1,
                                               max_num_active_requests=16,
                                               enable_iter_perf_stats=True,
-                                              batch_wait_timeout_ms=0.0,
-                                              is_disaggregated=True)
+                                              batch_wait_timeout_ms=0.0)
 
         with pytest.raises(
                 ValueError,
@@ -598,8 +592,7 @@ def attention_dp_queue(mock_dist_attention_dp):
                                  max_beam_width=2,
                                  max_num_active_requests=8,
                                  enable_iter_perf_stats=True,
-                                 batch_wait_timeout_ms=0.0,
-                                 is_disaggregated=False)
+                                 batch_wait_timeout_ms=0.0)
     # Initialize all_ranks_num_active_requests
     return queue