Pass peft_cache_manager=None to BindCapacityScheduler in create_autodeploy_executor function in ad_executor.py

Signed-off-by: Amit Zuker <[email protected]>

Author: amitz-nv

tensorrt_llm/_torch/auto_deploy/shim/ad_executor.py

@@ -286,7 +286,9 @@ def create_autodeploy_executor(executor_config: ExecutorConfig, checkpoint_dir:
     resource_manager.resource_managers.move_to_end(ResourceManagerType.KV_CACHE_MANAGER, last=True)
 
     # scheduling
-    capacitor_scheduler = BindCapacityScheduler(ad_config.max_batch_size, kv_cache_manager.impl)
+    capacitor_scheduler = BindCapacityScheduler(
+        ad_config.max_batch_size, kv_cache_manager.impl, peft_cache_manager=None
+    )
     mb_scheduler = BindMicroBatchScheduler(
         ad_config.max_batch_size, engine.cache_seq_interface.info.max_num_tokens
     )

tensorrt_llm/_torch/pyexecutor/scheduler.py

@@ -73,7 +73,7 @@ def __init__(
         self,
         max_num_requests: int,
         kv_cache_manager,
-        peft_cache_manager: tb_internal.batch_manager.PeftCacheManager,
+        peft_cache_manager: tb_internal.batch_manager.PeftCacheManager | None,
         scheduler_policy: tb_executor.CapacitySchedulerPolicy = tb_executor.
         CapacitySchedulerPolicy.GUARANTEED_NO_EVICT,
         two_step_lookahead: bool = False,