ci: migrate 2-GPU tests to test/registered/ (#16529)

test/registered/disaggregation/test_disaggregation_basic.py

@@ -7,6 +7,7 @@ import openai
 import requests
 from transformers import AutoTokenizer
 
+from sglang.test.ci.ci_register import register_cuda_ci
 from sglang.test.few_shot_gsm8k import run_eval as run_eval_few_shot_gsm8k
 from sglang.test.server_fixtures.disaggregation_fixture import (
     PDDisaggregationServerBase,
@@ -19,6 +20,8 @@ from sglang.test.test_utils import (
     popen_launch_pd_server,
 )
 
+register_cuda_ci(est_time=400, suite="stage-b-test-large-2-gpu")
+
 
 class TestDisaggregationAccuracy(PDDisaggregationServerBase):
     @classmethod

test/registered/distributed/test_data_parallelism.py

@@ -5,6 +5,7 @@ from types import SimpleNamespace
 import requests
 
 from sglang.srt.utils import kill_process_tree
+from sglang.test.ci.ci_register import register_amd_ci, register_cuda_ci
 from sglang.test.run_eval import run_eval
 from sglang.test.test_utils import (
     DEFAULT_MODEL_NAME_FOR_TEST,
@@ -14,6 +15,9 @@ from sglang.test.test_utils import (
     popen_launch_server,
 )
 
+register_cuda_ci(est_time=73, suite="stage-b-test-large-2-gpu")
+register_amd_ci(est_time=73, suite="stage-b-test-large-2-gpu-amd")
+
 
 class TestDataParallelism(CustomTestCase):
     @classmethod

test/registered/distributed/test_dp_attention.py

@@ -5,6 +5,7 @@ import requests
 
 from sglang.srt.environ import envs
 from sglang.srt.utils import kill_process_tree
+from sglang.test.ci.ci_register import register_cuda_ci
 from sglang.test.few_shot_gsm8k import run_eval as run_eval_few_shot_gsm8k
 from sglang.test.kits.radix_cache_server_kit import run_radix_attention_test
 from sglang.test.run_eval import run_eval
@@ -19,6 +20,8 @@ from sglang.test.test_utils import (
     popen_launch_server,
 )
 
+register_cuda_ci(est_time=350, suite="stage-b-test-large-2-gpu")
+
 
 class TestDPAttentionDP2TP2(CustomTestCase):
     @classmethod

test/registered/distributed/test_load_weights_from_remote_instance.py

@@ -24,6 +24,7 @@ import torch
 import torch.multiprocessing as mp
 
 import sglang as sgl
+from sglang.test.ci.ci_register import register_amd_ci, register_cuda_ci
 from sglang.test.test_utils import (
     DEFAULT_PORT_FOR_SRT_TEST_RUNNER,
     DEFAULT_SMALL_MODEL_NAME_FOR_TEST,
@@ -37,6 +38,9 @@ from sglang.utils import terminate_process
 
 mp.set_start_method("spawn", force=True)
 
+register_cuda_ci(est_time=72, suite="stage-b-test-large-2-gpu")
+register_amd_ci(est_time=72, suite="stage-b-test-large-2-gpu-amd")
+
 
 def verify_params_close(params1, params2, error_msg):
     """Verify if two parameter arrays are close enough."""

test/registered/hicache/test_hicache_storage_3fs_backend.py

@@ -1,7 +1,7 @@
 """
 Benchmark tests for HiCache Storage with 3FS backend.
 Usage:
-    python3 -m pytest test/srt/hicache/test_hicache_storage_3fs_backend.py -v
+    python3 -m pytest test/registered/hicache/test_hicache_storage_3fs_backend.py -v
 """
 
 import json
@@ -10,8 +10,11 @@ import unittest
 
 from test_hicache_storage_file_backend import HiCacheStorageBaseMixin
 
+from sglang.test.ci.ci_register import register_cuda_ci
 from sglang.test.test_utils import CustomTestCase
 
+register_cuda_ci(est_time=200, suite="stage-b-test-large-2-gpu")
+
 
 class HiCacheStorage3FSBackendBaseMixin(HiCacheStorageBaseMixin):
     """Base mixin class with common setup and utilities"""

test/registered/hicache/test_hicache_storage_file_backend.py

@@ -1,7 +1,7 @@
 """
 E2E tests for HiCache Storage functionality.
 Usage:
-    python3 -m pytest test/srt/hicache/test_hicache_storage_e2e.py -v
+    python3 -m pytest test/registered/hicache/test_hicache_storage_file_backend.py -v
 """
 
 import json
@@ -18,6 +18,7 @@ import requests
 
 from sglang.bench_serving import get_tokenizer
 from sglang.srt.utils import kill_process_tree
+from sglang.test.ci.ci_register import register_cuda_ci
 from sglang.test.few_shot_gsm8k import run_eval as run_eval_few_shot_gsm8k
 from sglang.test.test_utils import (
     DEFAULT_MLA_MODEL_NAME_FOR_TEST,
@@ -29,6 +30,8 @@ from sglang.test.test_utils import (
     popen_launch_server,
 )
 
+register_cuda_ci(est_time=200, suite="stage-b-test-large-2-gpu")
+
 
 class HiCacheStorageBaseMixin:
     """Base mixin class with common setup and utilities"""

test/registered/hicache/test_hicache_storage_mooncake_backend.py

@@ -1,7 +1,7 @@
 """
 Benchmark tests for HiCache Storage with Mooncake backend.
 Usage:
-    python3.10 -m pytest test/srt/hicache/test_hicache_storage_mooncake_backend.py -v
+    python3.10 -m pytest test/registered/hicache/test_hicache_storage_mooncake_backend.py -v
 """
 
 import os
@@ -12,6 +12,7 @@ import unittest
 import requests
 from test_hicache_storage_file_backend import HiCacheStorageBaseMixin
 
+from sglang.test.ci.ci_register import register_cuda_ci
 from sglang.test.test_utils import (
     DEFAULT_MLA_MODEL_NAME_FOR_TEST,
     CustomTestCase,
@@ -19,6 +20,8 @@ from sglang.test.test_utils import (
     is_in_ci,
 )
 
+register_cuda_ci(est_time=300, suite="stage-b-test-large-2-gpu")
+
 
 class HiCacheStorageMooncakeBackendBaseMixin(HiCacheStorageBaseMixin):
     """Base mixin class with common setup and utilities"""

test/registered/models/test_kimi_linear_models.py

@@ -2,6 +2,7 @@ import unittest
 from types import SimpleNamespace
 
 from sglang.srt.utils import kill_process_tree
+from sglang.test.ci.ci_register import register_cuda_ci
 from sglang.test.few_shot_gsm8k import run_eval
 from sglang.test.test_utils import (
     DEFAULT_TIMEOUT_FOR_SERVER_LAUNCH,
@@ -10,6 +11,8 @@ from sglang.test.test_utils import (
     popen_launch_server,
 )
 
+register_cuda_ci(est_time=90, suite="stage-b-test-large-2-gpu")
+
 
 class TestKimiLinear(CustomTestCase):
     @classmethod

test/registered/models/test_nvidia_nemotron_nano_v2.py

@@ -1,9 +1,12 @@
 import unittest
 
 from sglang.srt.utils import is_blackwell
+from sglang.test.ci.ci_register import register_cuda_ci
 from sglang.test.kits.gsm8k_accuracy_kit import GSM8KMixin
 from sglang.test.server_fixtures.default_fixture import DefaultServerBase
 
+register_cuda_ci(est_time=132, suite="stage-b-test-large-2-gpu")
+
 
 class TestNvidiaNemotronNanoV2BF16(GSM8KMixin, DefaultServerBase):
     model = "nvidia/NVIDIA-Nemotron-Nano-9B-v2"

test/srt/run_suite.py

@@ -16,17 +16,6 @@ suites = {
         TestFile("test_video_utils.py", 5),
         TestFile("test_modelopt_export.py", 9),
     ],
-    "per-commit-2-gpu": [
-        TestFile("hicache/test_hicache_storage_3fs_backend.py", 200),
-        TestFile("hicache/test_hicache_storage_file_backend.py", 200),
-        TestFile("hicache/test_hicache_storage_mooncake_backend.py", 300),
-        TestFile("models/test_kimi_linear_models.py", 90),
-        TestFile("models/test_nvidia_nemotron_nano_v2.py", 132),
-        TestFile("test_data_parallelism.py", 73),
-        TestFile("test_disaggregation_basic.py", 400),
-        TestFile("test_dp_attention.py", 350),
-        TestFile("test_load_weights_from_remote_instance.py", 72),
-    ],
     "per-commit-4-gpu": [
         TestFile("models/test_qwen3_next_models.py", 650),
         TestFile("test_gpt_oss_4gpu.py", 300),
@@ -116,11 +105,6 @@ suite_amd = {
         # TestFile("test_vision_chunked_prefill.py", 175), # Disabled temporarily and track in #7701
         # TestFile("test_wave_attention_backend.py", 150), # Disabled temporarily, see https://github.com/sgl-project/sglang/issues/11127
     ],
-    "per-commit-amd-mi35x": [],
-    "per-commit-2-gpu-amd": [
-        TestFile("test_data_parallelism.py", 73),
-        TestFile("test_load_weights_from_remote_instance.py", 72),
-    ],
     "per-commit-4-gpu-amd": [
         TestFile("test_pp_single_node.py", 150),
     ],