From f5d72b2fc6771de19c351945f1fbbb0198d53b8e Mon Sep 17 00:00:00 2001
From: sroy745 <142070531+sroy745@users.noreply.github.com>
Date: Thu, 3 Oct 2024 09:44:21 -0700
Subject: [PATCH] [Core] Make BlockSpaceManagerV2 the default BlockManager to
 use. (#8678)

---
 vllm/config.py           |  2 +-
 vllm/engine/arg_utils.py | 11 +++++++----
 2 files changed, 8 insertions(+), 5 deletions(-)

diff --git a/vllm/config.py b/vllm/config.py
index 1310c07ade482..05d5f4998d74d 100644
--- a/vllm/config.py
+++ b/vllm/config.py
@@ -970,7 +970,7 @@ class SchedulerConfig:
                  max_num_batched_tokens: Optional[int],
                  max_num_seqs: int,
                  max_model_len: int,
-                 use_v2_block_manager: bool = False,
+                 use_v2_block_manager: bool = True,
                  num_lookahead_slots: int = 0,
                  delay_factor: float = 0.0,
                  enable_chunked_prefill: bool = False,
diff --git a/vllm/engine/arg_utils.py b/vllm/engine/arg_utils.py
index c97b6ffb093f7..097fe7c02444c 100644
--- a/vllm/engine/arg_utils.py
+++ b/vllm/engine/arg_utils.py
@@ -107,7 +107,7 @@ class EngineArgs:
     block_size: int = 16
     enable_prefix_caching: bool = False
     disable_sliding_window: bool = False
-    use_v2_block_manager: bool = False
+    use_v2_block_manager: bool = True
     swap_space: float = 4  # GiB
     cpu_offload_gb: float = 0  # GiB
     gpu_memory_utilization: float = 0.90
@@ -369,9 +369,12 @@ class EngineArgs:
                             action='store_true',
                             help='Disables sliding window, '
                             'capping to sliding window size')
-        parser.add_argument('--use-v2-block-manager',
-                            action='store_true',
-                            help='Use BlockSpaceMangerV2.')
+        parser.add_argument(
+            '--use-v2-block-manager',
+            default=EngineArgs.use_v2_block_manager,
+            action='store_true',
+            help='Use BlockSpaceMangerV2. By default this is set to True. '
+            'Set to False to use BlockSpaceManagerV1')
         parser.add_argument(
             '--num-lookahead-slots',
             type=int,