stash

iotdb-core/ainode/ainode/core/inference/pool_scheduler/basic_pool_scheduler.py

diff --git a/iotdb-core/ainode/ainode/core/inference/pool_scheduler/basic_pool_scheduler.py b/iotdb-core/ainode/ainode/core/inference/pool_scheduler/basic_pool_scheduler.py
index 8994a9e..0343c2a 100644
--- a/iotdb-core/ainode/ainode/core/inference/pool_scheduler/basic_pool_scheduler.py
+++ b/iotdb-core/ainode/ainode/core/inference/pool_scheduler/basic_pool_scheduler.py
@@ -51,7 +51,7 @@
         Schedule a scaling action for the given model_id.
         """
         if model_id not in self._request_pool_map:
-            pool_num = estimate_pool_size(self.DEFAULT_DEVICE, model_id)
+            pool_num = 2
             if pool_num <= 0:
                 raise InferenceModelInternalError(
                     f"Not enough memory to run model {model_id}."

iotdb-core/ainode/ainode/core/manager/inference_manager.py

diff --git a/iotdb-core/ainode/ainode/core/manager/inference_manager.py b/iotdb-core/ainode/ainode/core/manager/inference_manager.py
index 2d4e208..a189274 100644
--- a/iotdb-core/ainode/ainode/core/manager/inference_manager.py
+++ b/iotdb-core/ainode/ainode/core/manager/inference_manager.py
@@ -242,9 +242,7 @@
                     predict_length,
                 )
 
-            if model_id in self.ACCELERATE_MODEL_ID and "cuda" in str(
-                self.DEFAULT_DEVICE
-            ):
+            if model_id in self.ACCELERATE_MODEL_ID:
                 # TODO: Logic in this branch shall handle all LTSM inferences
                 # TODO: TSBlock -> Tensor codes should be unified
                 data = full_data[1][0]