chore: update CI test est_time values (#38238)

Co-authored-by: sglang-bot <sglang-bot@users.noreply.github.com>

test/registered/model_loading/test_external_models.py

@@ -9,7 +9,7 @@ from sglang.test.ci.ci_register import (
 )
 from sglang.test.test_utils import CustomTestCase
 
-register_cuda_ci(est_time=29, stage="base-b", runner_config="1-gpu-small")
+register_cuda_ci(est_time=34, stage="base-b", runner_config="1-gpu-small")
 register_amd_ci(est_time=45, suite="stage-b-test-1-gpu-small-amd")
 register_cpu_ci(est_time=32, suite="stage-b-test-cpu-intel")
 

test/registered/model_loading/test_load_weights_from_remote_instance.py

@@ -38,7 +38,7 @@ from sglang.utils import terminate_process
 
 mp.set_start_method("spawn", force=True)
 
-register_cuda_ci(est_time=145, stage="extra-a", runner_config="2-gpu-large")
+register_cuda_ci(est_time=131, stage="extra-a", runner_config="2-gpu-large")
 register_amd_ci(est_time=72, stage="extra-a", runner_config="2-gpu-large-amd")
 
 

test/registered/model_loading/test_startup_weight_load.py

@@ -9,7 +9,7 @@ from sglang.test.test_utils import (
     CustomTestCase,
 )
 
-register_cuda_ci(est_time=120, stage="base-b", runner_config="1-gpu-small")
+register_cuda_ci(est_time=63, stage="base-b", runner_config="1-gpu-small")
 
 
 class TestStartupWeightLoad(CustomTestCase):

test/registered/model_loading/test_utils_update_weights.py

@@ -12,7 +12,7 @@ from sglang.srt.weight_sync.utils import update_weights
 from sglang.test.ci.ci_register import register_cuda_ci
 from sglang.test.test_utils import DEFAULT_SMALL_MODEL_NAME_FOR_TEST
 
-register_cuda_ci(est_time=32, stage="base-b", runner_config="1-gpu-large")
+register_cuda_ci(est_time=42, stage="base-b", runner_config="1-gpu-large")
 
 
 class AsyncEngine(Engine):

test/registered/model_loading/test_weight_cache_daemon.py

@@ -29,8 +29,8 @@ DEFAULT_MODEL = "Qwen/Qwen3-0.6B"
 # IPC handoff is exercised on every PR. Since the CI runner executes the whole
 # file per suite, TestWeightCacheDaemonTP2 self-skips when fewer than 2 GPUs are
 # visible (i.e. on the 1-gpu runner).
-register_cuda_ci(est_time=280, stage="extra-a", runner_config="2-gpu-large")
-register_cuda_ci(est_time=45, stage="base-b", runner_config="1-gpu-small")
+register_cuda_ci(est_time=313, stage="extra-a", runner_config="2-gpu-large")
+register_cuda_ci(est_time=68, stage="base-b", runner_config="1-gpu-small")
 
 # Capture the client server's logs so test_loaded_via_ipc can assert the IPC
 # load path actually ran (and did not silently fall back to disk).

test/registered/model_loading/test_weight_loader_v2_e2e.py

@@ -14,7 +14,7 @@
 
 from sglang.test.ci.ci_register import register_cuda_ci
 
-register_cuda_ci(est_time=120, stage="base-b", runner_config="1-gpu-small")
+register_cuda_ci(est_time=62, stage="base-b", runner_config="1-gpu-small")
 
 import multiprocessing as mp