chore: moved asyncio import to top of the file for planner test (#3001)

Signed-off-by: Michael Shin <michaelshin@users.noreply.github.com>

Author: michaelshin

tests/planner/test_replica_calculation.py

@@ -9,6 +9,7 @@
 """
 
 import argparse
+import asyncio
 import math
 import os
 
@@ -152,8 +153,6 @@ async def mock_get_workers_info():
         planner.decode_interpolator.interpolate_itl.return_value = 10.0
 
         # Run the calculation
-        import asyncio
-
         asyncio.run(planner.make_adjustments())
 
         # Extract the calculated values from the log calls or by checking the mock calls
@@ -217,8 +216,6 @@ async def mock_get_workers_info():
         planner.decode_interpolator.interpolate_itl.return_value = 10.0
 
         # Run the calculation
-        import asyncio
-
         asyncio.run(planner.make_adjustments())
 
         # Check the results
@@ -290,8 +287,6 @@ async def mock_get_workers_info():
         planner.connector.reset_mock()
 
         # Run calculation
-        import asyncio
-
         asyncio.run(planner.make_adjustments())
 
         # Verify results
@@ -344,8 +339,6 @@ async def mock_get_workers_info():
         planner.decode_interpolator.interpolate_itl.return_value = 10.0
 
         # Run calculation
-        import asyncio
-
         asyncio.run(planner.make_adjustments())
 
         # Verify that total GPU usage doesn't exceed budget
@@ -401,8 +394,6 @@ async def mock_get_workers_info():
         planner.decode_interpolator.interpolate_itl.return_value = 10.0
 
         # Run calculation
-        import asyncio
-
         asyncio.run(planner.make_adjustments())
 
         # Verify minimum constraints are respected
@@ -465,8 +456,6 @@ async def mock_get_workers_info():
         )
 
         # Run calculation
-        import asyncio
-
         asyncio.run(planner.make_adjustments())
 
         # Verify that correction factor was effectively clamped
@@ -526,8 +515,6 @@ async def mock_get_workers_info():
                 planner.decode_interpolator.interpolate_itl.return_value = 10.0
 
                 # Run calculation
-                import asyncio
-
                 asyncio.run(planner.make_adjustments())
 
                 # Should handle gracefully without crashing
@@ -590,8 +577,6 @@ async def mock_get_workers_info():
         )  # 4 GPUs per engine
 
         # Run calculation
-        import asyncio
-
         asyncio.run(planner.make_adjustments())
 
         if planner.connector.set_component_replicas.called:
@@ -649,8 +634,6 @@ async def mock_get_workers_info():
         planner.decode_interpolator.interpolate_itl.return_value = 10.0
 
         # Run calculation
-        import asyncio
-
         asyncio.run(planner.make_adjustments())
 
         if planner.connector.set_component_replicas.called: