[NPU]Add a full test pipeline on NPU, resolve issues in the NPU test architecture (#20751)

test/registered/ascend/basic_function/HiCache/test_npu_hicache_mha.py

@@ -3,6 +3,7 @@ from types import SimpleNamespace
 from urllib.parse import urlparse
 
 from sglang.srt.utils import kill_process_tree
+from sglang.test.ci.ci_register import register_npu_ci
 from sglang.test.few_shot_gsm8k import run_eval as run_eval_few_shot_gsm8k
 from sglang.test.test_utils import (
     DEFAULT_TIMEOUT_FOR_SERVER_LAUNCH,
@@ -11,6 +12,9 @@ from sglang.test.test_utils import (
     popen_launch_server,
 )
 
+register_npu_ci(est_time=400, suite="stage-b-test-1-npu-a2", nightly=False)
+register_npu_ci(est_time=400, suite="nightly-1-npu-a3", nightly=True)
+
 TEST_MODEL_MATRIX = {
     "/root/.cache/modelscope/hub/models/Qwen/Qwen2.5-7B-Instruct": {
         "accuracy": 0.85,

test/registered/ascend/basic_function/HiCache/test_npu_hicache_mla.py

@@ -3,6 +3,7 @@ from types import SimpleNamespace
 from urllib.parse import urlparse
 
 from sglang.srt.utils import kill_process_tree
+from sglang.test.ci.ci_register import register_npu_ci
 from sglang.test.few_shot_gsm8k import run_eval as run_eval_few_shot_gsm8k
 from sglang.test.test_utils import (
     DEFAULT_TIMEOUT_FOR_SERVER_LAUNCH,
@@ -11,6 +12,9 @@ from sglang.test.test_utils import (
     popen_launch_server,
 )
 
+register_npu_ci(est_time=400, suite="stage-b-test-4-npu-a3", nightly=False)
+register_npu_ci(est_time=400, suite="nightly-4-npu-a3", nightly=True)
+
 TEST_MODEL_MATRIX = {
     "/root/.cache/modelscope/hub/models/vllm-ascend/DeepSeek-V2-Lite-W8A8": {
         "accuracy": 0.34,

test/registered/ascend/basic_function/backends/test_npu_sampling_backend.py

@@ -4,6 +4,7 @@ from types import SimpleNamespace
 import requests
 
 from sglang.srt.utils import kill_process_tree
+from sglang.test.ci.ci_register import register_npu_ci
 from sglang.test.run_eval import run_eval
 from sglang.test.test_utils import (
     DEFAULT_TIMEOUT_FOR_SERVER_LAUNCH,
@@ -12,6 +13,9 @@ from sglang.test.test_utils import (
     popen_launch_server,
 )
 
+register_npu_ci(est_time=400, suite="stage-b-test-1-npu-a2", nightly=False)
+register_npu_ci(est_time=400, suite="nightly-1-npu-a3", nightly=True)
+
 
 class TestAscendSamplingBackend(CustomTestCase):
     @classmethod

test/registered/ascend/basic_function/dllm/test_npu_llada2_mini.py

@@ -3,6 +3,7 @@ import unittest
 from types import SimpleNamespace
 
 from sglang.srt.utils import kill_process_tree
+from sglang.test.ci.ci_register import register_npu_ci
 from sglang.test.few_shot_gsm8k import run_eval as run_eval_few_shot_gsm8k
 from sglang.test.send_one import BenchArgs, send_one_prompt
 from sglang.test.test_utils import (
@@ -14,6 +15,9 @@ from sglang.test.test_utils import (
     write_github_step_summary,
 )
 
+register_npu_ci(est_time=400, suite="stage-b-test-4-npu-a3", nightly=False)
+register_npu_ci(est_time=400, suite="nightly-1-npu-a3", nightly=True)
+
 
 class TestLLaDA2Mini(CustomTestCase):
     @classmethod

test/registered/ascend/basic_function/optimization_debug/test_npu_compile_graph_tp1_bf16.py

@@ -4,6 +4,7 @@ from types import SimpleNamespace
 from urllib.parse import urlparse
 
 from sglang.srt.utils import kill_process_tree
+from sglang.test.ci.ci_register import register_npu_ci
 from sglang.test.few_shot_gsm8k import run_eval as run_eval_few_shot_gsm8k
 from sglang.test.test_utils import (
     DEFAULT_TIMEOUT_FOR_SERVER_LAUNCH,
@@ -12,6 +13,9 @@ from sglang.test.test_utils import (
     popen_launch_server,
 )
 
+register_npu_ci(est_time=400, suite="stage-b-test-1-npu-a2", nightly=False)
+register_npu_ci(est_time=400, suite="nightly-1-npu-a3", nightly=True)
+
 TEST_MODEL_MATRIX = {
     "/root/.cache/modelscope/hub/models/Qwen/Qwen2.5-7B-Instruct": {
         "accuracy": 0.84,

test/registered/ascend/basic_function/optimization_debug/test_npu_graph_tp1_bf16.py

@@ -3,6 +3,7 @@ from types import SimpleNamespace
 from urllib.parse import urlparse
 
 from sglang.srt.utils import kill_process_tree
+from sglang.test.ci.ci_register import register_npu_ci
 from sglang.test.few_shot_gsm8k import run_eval as run_eval_few_shot_gsm8k
 from sglang.test.test_utils import (
     DEFAULT_TIMEOUT_FOR_SERVER_LAUNCH,
@@ -11,6 +12,9 @@ from sglang.test.test_utils import (
     popen_launch_server,
 )
 
+register_npu_ci(est_time=400, suite="stage-b-test-1-npu-a2", nightly=False)
+register_npu_ci(est_time=400, suite="nightly-1-npu-a3", nightly=True)
+
 TEST_MODEL_MATRIX = {
     "/root/.cache/modelscope/hub/models/Qwen/Qwen2.5-7B-Instruct": {
         "accuracy": 0.85,

test/registered/ascend/basic_function/optimization_debug/test_npu_graph_tp2_bf16.py

@@ -3,6 +3,7 @@ from types import SimpleNamespace
 from urllib.parse import urlparse
 
 from sglang.srt.utils import kill_process_tree
+from sglang.test.ci.ci_register import register_npu_ci
 from sglang.test.few_shot_gsm8k import run_eval as run_eval_few_shot_gsm8k
 from sglang.test.test_utils import (
     DEFAULT_TIMEOUT_FOR_SERVER_LAUNCH,
@@ -11,6 +12,9 @@ from sglang.test.test_utils import (
     popen_launch_server,
 )
 
+register_npu_ci(est_time=400, suite="stage-b-test-2-npu-a2", nightly=False)
+register_npu_ci(est_time=400, suite="nightly-2-npu-a3", nightly=True)
+
 TEST_MODEL_MATRIX = {
     "/root/.cache/modelscope/hub/models/Qwen/Qwen2.5-7B-Instruct": {
         "accuracy": 0.85,

test/registered/ascend/basic_function/optimization_debug/test_npu_piecewise_graph_prefill.py

@@ -2,6 +2,7 @@ import unittest
 from urllib.parse import urlparse
 
 from sglang.srt.utils import kill_process_tree
+from sglang.test.ci.ci_register import register_npu_ci
 from sglang.test.few_shot_gsm8k import run_eval as run_eval_few_shot_gsm8k
 from sglang.test.test_utils import (
     DEFAULT_TIMEOUT_FOR_SERVER_LAUNCH,
@@ -12,6 +13,9 @@ from sglang.test.test_utils import (
     run_bench_one_batch,
 )
 
+register_npu_ci(est_time=400, suite="stage-b-test-1-npu-a2", nightly=False)
+register_npu_ci(est_time=400, suite="nightly-1-npu-a3", nightly=True)
+
 MODEL = "/root/.cache/modelscope/hub/models/Qwen/Qwen2.5-7B-Instruct"
 GSM8K_EXP_ACCURACY = 0.84
 EXP_PREFILL_LATENCY = 0.045

test/registered/ascend/basic_function/parallel_strategy/expert_parallelism/test_npu_deepep.py

@@ -4,6 +4,7 @@ from types import SimpleNamespace
 from urllib.parse import urlparse
 
 from sglang.srt.utils import kill_process_tree
+from sglang.test.ci.ci_register import register_npu_ci
 from sglang.test.few_shot_gsm8k import run_eval as run_eval_few_shot_gsm8k
 from sglang.test.test_utils import (
     DEFAULT_URL_FOR_TEST,
@@ -11,6 +12,9 @@ from sglang.test.test_utils import (
     popen_launch_server,
 )
 
+register_npu_ci(est_time=400, suite="stage-b-test-16-npu-a3", nightly=False)
+register_npu_ci(est_time=400, suite="nightly-16-npu-a3", nightly=True)
+
 TEST_MODEL_MATRIX = {
     "/root/.cache/modelscope/hub/models/vllm-ascend/DeepSeek-R1-0528-W8A8": {
         "accuracy": 0.95,

test/registered/ascend/basic_function/quant/test_npu_autoround_dense.py

@@ -4,6 +4,8 @@ from types import SimpleNamespace
 from urllib.parse import urlparse
 
 from sglang.srt.utils import kill_process_tree
+from sglang.test.ascend.test_ascend_utils import QWEN3_8B_INT4_AUTOROUND_WEIGHTS_PATH
+from sglang.test.ci.ci_register import register_npu_ci
 from sglang.test.few_shot_gsm8k import run_eval as run_eval_few_shot_gsm8k
 from sglang.test.test_utils import (
     DEFAULT_TIMEOUT_FOR_SERVER_LAUNCH,
@@ -12,10 +14,13 @@ from sglang.test.test_utils import (
     popen_launch_server,
 )
 
+register_npu_ci(est_time=400, suite="stage-b-test-1-npu-a2", nightly=False)
+register_npu_ci(est_time=400, suite="nightly-1-npu-a3", nightly=True)
+
 logger = logging.getLogger(__name__)
 
 TEST_MODEL_MATRIX = {
-    "/root/.cache/modelscope/hub/models/Intel/Qwen3-8B-int4-AutoRound": {
+    QWEN3_8B_INT4_AUTOROUND_WEIGHTS_PATH: {
         "accuracy": 0.85,
     },
 }

test/registered/ascend/basic_function/quant/test_npu_autoround_moe.py

@@ -4,6 +4,10 @@ from types import SimpleNamespace
 from urllib.parse import urlparse
 
 from sglang.srt.utils import kill_process_tree
+from sglang.test.ascend.test_ascend_utils import (
+    QWEN3_30B_A3B_INSTRUCT_2507_INT4_AUTOROUND_WEIGHTS_PATH,
+)
+from sglang.test.ci.ci_register import register_npu_ci
 from sglang.test.few_shot_gsm8k import run_eval as run_eval_few_shot_gsm8k
 from sglang.test.test_utils import (
     DEFAULT_TIMEOUT_FOR_SERVER_LAUNCH,
@@ -12,10 +16,13 @@ from sglang.test.test_utils import (
     popen_launch_server,
 )
 
+register_npu_ci(est_time=400, suite="stage-b-test-1-npu-a2", nightly=False)
+register_npu_ci(est_time=400, suite="nightly-1-npu-a3", nightly=True)
+
 logger = logging.getLogger(__name__)
 
 TEST_MODEL_MATRIX = {
-    "/root/.cache/modelscope/hub/models/Intel/Qwen3-30B-A3B-Instruct-2507-int4-AutoRound": {
+    QWEN3_30B_A3B_INSTRUCT_2507_INT4_AUTOROUND_WEIGHTS_PATH: {
         "accuracy": 0.85,
     },
 }

test/registered/ascend/basic_function/quant/test_npu_gptq_moe.py

@@ -4,6 +4,10 @@ from types import SimpleNamespace
 from urllib.parse import urlparse
 
 from sglang.srt.utils import kill_process_tree
+from sglang.test.ascend.test_ascend_utils import (
+    QWEN3_30B_A3B_GPTQ_2507_INT4_WEIGHTS_PATH,
+)
+from sglang.test.ci.ci_register import register_npu_ci
 from sglang.test.few_shot_gsm8k import run_eval as run_eval_few_shot_gsm8k
 from sglang.test.test_utils import (
     DEFAULT_TIMEOUT_FOR_SERVER_LAUNCH,
@@ -12,10 +16,13 @@ from sglang.test.test_utils import (
     popen_launch_server,
 )
 
+register_npu_ci(est_time=400, suite="stage-b-test-1-npu-a2", nightly=False)
+register_npu_ci(est_time=400, suite="nightly-1-npu-a3", nightly=True)
+
 logger = logging.getLogger(__name__)
 
 TEST_MODEL_MATRIX = {
-    "/root/.cache/modelscope/hub/models/Qwen/Qwen3-30B-A3B-GPTQ-Int4": {
+    QWEN3_30B_A3B_GPTQ_2507_INT4_WEIGHTS_PATH: {
         "accuracy": 0.85,
     },
 }

test/registered/ascend/basic_function/quant/test_npu_w4a4_quantization.py

@@ -12,6 +12,7 @@ from urllib.parse import urlparse
 import requests
 
 from sglang.srt.utils import kill_process_tree
+from sglang.test.ci.ci_register import register_npu_ci
 from sglang.test.few_shot_gsm8k import run_eval
 from sglang.test.test_utils import (
     DEFAULT_TIMEOUT_FOR_SERVER_LAUNCH,
@@ -21,6 +22,9 @@ from sglang.test.test_utils import (
     popen_launch_server,
 )
 
+register_npu_ci(est_time=400, suite="stage-b-test-4-npu-a3", nightly=False)
+register_npu_ci(est_time=400, suite="nightly-4-npu-a3", nightly=True)
+
 if "ASCEND_RT_VISIBLE_DEVICES" not in os.environ:
     os.environ["ASCEND_RT_VISIBLE_DEVICES"] = "0,1,2,3"
 DEFAULT_PORT_FOR_SRT_TEST_RUNNER = (

test/registered/ascend/basic_function/quant/test_npu_w8a8_quantization.py

@@ -12,6 +12,7 @@ from urllib.parse import urlparse
 import requests
 
 from sglang.srt.utils import kill_process_tree
+from sglang.test.ci.ci_register import register_npu_ci
 from sglang.test.few_shot_gsm8k import run_eval
 from sglang.test.test_utils import (
     DEFAULT_TIMEOUT_FOR_SERVER_LAUNCH,
@@ -21,6 +22,9 @@ from sglang.test.test_utils import (
     popen_launch_server,
 )
 
+register_npu_ci(est_time=400, suite="stage-b-test-1-npu-a2", nightly=False)
+register_npu_ci(est_time=400, suite="nightly-1-npu-a3", nightly=True)
+
 if "ASCEND_RT_VISIBLE_DEVICES" not in os.environ:
     os.environ["ASCEND_RT_VISIBLE_DEVICES"] = "0,1"
 DEFAULT_PORT_FOR_SRT_TEST_RUNNER = (

test/registered/ascend/basic_function/runtime_opts/test_npu_mla_fia_w8a8int8.py

@@ -4,6 +4,7 @@ from types import SimpleNamespace
 from urllib.parse import urlparse
 
 from sglang.srt.utils import kill_process_tree
+from sglang.test.ci.ci_register import register_npu_ci
 from sglang.test.few_shot_gsm8k import run_eval as run_eval_few_shot_gsm8k
 from sglang.test.test_utils import (
     DEFAULT_TIMEOUT_FOR_SERVER_LAUNCH,
@@ -12,6 +13,9 @@ from sglang.test.test_utils import (
     popen_launch_server,
 )
 
+register_npu_ci(est_time=400, suite="stage-b-test-2-npu-a2", nightly=False)
+register_npu_ci(est_time=400, suite="nightly-2-npu-a3", nightly=True)
+
 TEST_MODEL_MATRIX = {
     "/root/.cache/modelscope/hub/models/vllm-ascend/DeepSeek-V2-Lite-W8A8": {
         "accuracy": 0.34,

test/registered/ascend/basic_function/runtime_opts/test_npu_mla_w8a8int8.py

@@ -3,6 +3,7 @@ from types import SimpleNamespace
 from urllib.parse import urlparse
 
 from sglang.srt.utils import kill_process_tree
+from sglang.test.ci.ci_register import register_npu_ci
 from sglang.test.few_shot_gsm8k import run_eval as run_eval_few_shot_gsm8k
 from sglang.test.test_utils import (
     DEFAULT_TIMEOUT_FOR_SERVER_LAUNCH,
@@ -11,6 +12,9 @@ from sglang.test.test_utils import (
     popen_launch_server,
 )
 
+register_npu_ci(est_time=400, suite="stage-b-test-4-npu-a3", nightly=False)
+register_npu_ci(est_time=400, suite="nightly-4-npu-a3", nightly=True)
+
 TEST_MODEL_MATRIX = {
     "/root/.cache/modelscope/hub/models/vllm-ascend/DeepSeek-V2-Lite-W8A8": {
         "accuracy": 0.34,

test/registered/ascend/basic_function/runtime_opts/test_npu_tp1_bf16.py

@@ -3,6 +3,7 @@ from types import SimpleNamespace
 from urllib.parse import urlparse
 
 from sglang.srt.utils import kill_process_tree
+from sglang.test.ci.ci_register import register_npu_ci
 from sglang.test.few_shot_gsm8k import run_eval as run_eval_few_shot_gsm8k
 from sglang.test.test_utils import (
     DEFAULT_TIMEOUT_FOR_SERVER_LAUNCH,
@@ -11,6 +12,9 @@ from sglang.test.test_utils import (
     popen_launch_server,
 )
 
+register_npu_ci(est_time=400, suite="stage-b-test-1-npu-a2", nightly=False)
+register_npu_ci(est_time=400, suite="nightly-1-npu-a3", nightly=True)
+
 TEST_MODEL_MATRIX = {
     "/root/.cache/modelscope/hub/models/Qwen/Qwen2.5-7B-Instruct": {
         "accuracy": 0.84,

test/registered/ascend/basic_function/runtime_opts/test_npu_tp2_bf16.py

@@ -3,6 +3,7 @@ from types import SimpleNamespace
 from urllib.parse import urlparse
 
 from sglang.srt.utils import kill_process_tree
+from sglang.test.ci.ci_register import register_npu_ci
 from sglang.test.few_shot_gsm8k import run_eval as run_eval_few_shot_gsm8k
 from sglang.test.test_utils import (
     DEFAULT_TIMEOUT_FOR_SERVER_LAUNCH,
@@ -11,6 +12,9 @@ from sglang.test.test_utils import (
     popen_launch_server,
 )
 
+register_npu_ci(est_time=400, suite="stage-b-test-2-npu-a2", nightly=False)
+register_npu_ci(est_time=400, suite="nightly-2-npu-a3", nightly=True)
+
 TEST_MODEL_MATRIX = {
     "/root/.cache/modelscope/hub/models/Qwen/Qwen2.5-7B-Instruct": {
         "accuracy": 0.85,

test/registered/ascend/basic_function/runtime_opts/test_npu_tp2_fia_bf16.py

@@ -4,6 +4,7 @@ from types import SimpleNamespace
 from urllib.parse import urlparse
 
 from sglang.srt.utils import kill_process_tree
+from sglang.test.ci.ci_register import register_npu_ci
 from sglang.test.few_shot_gsm8k import run_eval as run_eval_few_shot_gsm8k
 from sglang.test.test_utils import (
     DEFAULT_TIMEOUT_FOR_SERVER_LAUNCH,
@@ -12,6 +13,9 @@ from sglang.test.test_utils import (
     popen_launch_server,
 )
 
+register_npu_ci(est_time=400, suite="stage-b-test-2-npu-a2", nightly=False)
+register_npu_ci(est_time=400, suite="nightly-2-npu-a3", nightly=True)
+
 TEST_MODEL_MATRIX = {
     "/root/.cache/modelscope/hub/models/Qwen/Qwen2.5-7B-Instruct": {
         "accuracy": 0.85,

test/registered/ascend/basic_function/runtime_opts/test_npu_tp4_bf16.py

@@ -3,6 +3,7 @@ from types import SimpleNamespace
 from urllib.parse import urlparse
 
 from sglang.srt.utils import kill_process_tree
+from sglang.test.ci.ci_register import register_npu_ci
 from sglang.test.few_shot_gsm8k import run_eval as run_eval_few_shot_gsm8k
 from sglang.test.test_utils import (
     DEFAULT_URL_FOR_TEST,
@@ -10,6 +11,9 @@ from sglang.test.test_utils import (
     popen_launch_server,
 )
 
+register_npu_ci(est_time=400, suite="stage-b-test-4-npu-a3", nightly=False)
+register_npu_ci(est_time=400, suite="nightly-4-npu-a3", nightly=True)
+
 TEST_MODEL_MATRIX = {
     "Qwen/Qwen3-30B-A3B-Instruct-2507": {
         "accuracy": 0.90,

test/registered/ascend/embedding_models/test_npu_bge_large_en_v1_5.py

@@ -11,7 +11,7 @@ from sglang.test.test_utils import CustomTestCase, get_similarities
 
 register_npu_ci(
     est_time=400,
-    suite="nightly-1-npu-a3",
+    suite="full-1-npu-a3",
     nightly=True,
     disabled="embeddings are not all close",
 )

test/registered/ascend/llm_models/test_npu_afm_4_5b.py

@@ -5,7 +5,7 @@ from sglang.test.ascend.test_ascend_utils import AFM_4_5B_BASE_WEIGHTS_PATH
 from sglang.test.ci.ci_register import register_npu_ci
 from sglang.test.test_utils import CustomTestCase
 
-register_npu_ci(est_time=400, suite="nightly-1-npu-a3", nightly=True)
+register_npu_ci(est_time=400, suite="full-1-npu-a3", nightly=True)
 
 
 class TestAFM(GSM8KAscendMixin, CustomTestCase):

test/registered/ascend/llm_models/test_npu_c4ai_command_r_v01.py

@@ -8,7 +8,7 @@ from sglang.test.ascend.test_ascend_utils import (
 from sglang.test.ci.ci_register import register_npu_ci
 from sglang.test.test_utils import CustomTestCase
 
-register_npu_ci(est_time=400, suite="nightly-2-npu-a3", nightly=False)
+register_npu_ci(est_time=400, suite="full-2-npu-a3", nightly=False)
 
 
 class TestC4AI(GSM8KAscendMixin, CustomTestCase):

test/registered/ascend/llm_models/test_npu_exaone_3.py

@@ -5,7 +5,7 @@ from sglang.test.ascend.test_ascend_utils import EXAONE_3_5_7_8B_INSTRUCT_WEIGHT
 from sglang.test.ci.ci_register import register_npu_ci
 from sglang.test.test_utils import CustomTestCase
 
-register_npu_ci(est_time=400, suite="nightly-1-npu-a3", nightly=False)
+register_npu_ci(est_time=400, suite="full-1-npu-a3", nightly=False)
 
 
 class TestEXAONE(GSM8KAscendMixin, CustomTestCase):

test/registered/ascend/llm_models/test_npu_granite_3_0_3b_a800m.py

@@ -7,7 +7,7 @@ from sglang.test.ascend.test_ascend_utils import (
 from sglang.test.ci.ci_register import register_npu_ci
 from sglang.test.test_utils import CustomTestCase
 
-register_npu_ci(est_time=400, suite="nightly-1-npu-a3", nightly=True)
+register_npu_ci(est_time=400, suite="full-1-npu-a3", nightly=True)
 
 
 class TestGranite(GSM8KAscendMixin, CustomTestCase):

test/registered/ascend/llm_models/test_npu_granite_3_1_8b.py

@@ -5,7 +5,7 @@ from sglang.test.ascend.test_ascend_utils import GRANITE_3_1_8B_INSTRUCT_WEIGHTS
 from sglang.test.ci.ci_register import register_npu_ci
 from sglang.test.test_utils import CustomTestCase
 
-register_npu_ci(est_time=400, suite="nightly-1-npu-a3", nightly=True)
+register_npu_ci(est_time=400, suite="full-1-npu-a3", nightly=True)
 
 
 class TestGranite(GSM8KAscendMixin, CustomTestCase):

test/registered/ascend/llm_models/test_npu_grok_2.py

@@ -6,7 +6,7 @@ from sglang.test.test_utils import CustomTestCase
 
 register_npu_ci(
     est_time=400,
-    suite="nightly-16-npu-a3",
+    suite="full-16-npu-a3",
     nightly=False,
     disabled="https://github.com/Ascend/sglang/issues/25",
 )

test/registered/ascend/llm_models/test_npu_ling_lite.py

@@ -5,7 +5,7 @@ from sglang.test.ascend.test_ascend_utils import LING_LITE_WEIGHTS_PATH
 from sglang.test.ci.ci_register import register_npu_ci
 from sglang.test.test_utils import CustomTestCase
 
-register_npu_ci(est_time=400, suite="nightly-2-npu-a3", nightly=True)
+register_npu_ci(est_time=400, suite="full-2-npu-a3", nightly=True)
 
 
 class TestLingLite(GSM8KAscendMixin, CustomTestCase):

test/registered/ascend/llm_models/test_npu_mimo_7b_rl.py

@@ -5,7 +5,7 @@ from sglang.test.ascend.test_ascend_utils import MIMO_7B_RL_WEIGHTS_PATH
 from sglang.test.ci.ci_register import register_npu_ci
 from sglang.test.test_utils import CustomTestCase
 
-register_npu_ci(est_time=400, suite="nightly-1-npu-a3", nightly=True)
+register_npu_ci(est_time=400, suite="full-1-npu-a3", nightly=True)
 
 
 class TestMiMo7BRL(GSM8KAscendMixin, CustomTestCase):

test/registered/ascend/llm_models/test_npu_persimmon_8b_chat.py

@@ -8,7 +8,7 @@ from sglang.test.test_utils import CustomTestCase
 
 register_npu_ci(
     est_time=400,
-    suite="nightly-1-npu-a3",
+    suite="full-1-npu-a3",
     nightly=False,
 )
 

test/registered/ascend/llm_models/test_npu_smollm_1_7b.py

@@ -5,7 +5,7 @@ from sglang.test.ascend.test_ascend_utils import SMOLLM_1_7B_WEIGHTS_PATH
 from sglang.test.ci.ci_register import register_npu_ci
 from sglang.test.test_utils import CustomTestCase
 
-register_npu_ci(est_time=400, suite="nightly-1-npu-a3", nightly=True)
+register_npu_ci(est_time=400, suite="full-1-npu-a3", nightly=True)
 
 
 class TestSmolLM(GSM8KAscendMixin, CustomTestCase):

test/registered/ascend/llm_models/test_npu_stablelm_2_1_6b.py

@@ -5,7 +5,7 @@ from sglang.test.ascend.test_ascend_utils import STABLELM_2_1_6B_WEIGHTS_PATH
 from sglang.test.ci.ci_register import register_npu_ci
 from sglang.test.test_utils import CustomTestCase
 
-register_npu_ci(est_time=400, suite="nightly-1-npu-a3", nightly=True)
+register_npu_ci(est_time=400, suite="full-1-npu-a3", nightly=True)
 
 
 class TestStablelm(GSM8KAscendMixin, CustomTestCase):

test/registered/ascend/reward_models/test_npu_gemma_2_27b_v0_2.py

@@ -10,7 +10,7 @@ from sglang.test.runners import HFRunner, SRTRunner
 from sglang.test.test_utils import CustomTestCase
 
 logger = logging.getLogger(__name__)
-register_npu_ci(est_time=400, suite="nightly-1-npu-a3", nightly=True)
+register_npu_ci(est_time=400, suite="full-1-npu-a3", nightly=True)
 
 MODELS = [
     (

test/registered/ascend/reward_models/test_npu_internlm2_7b_reward.py

@@ -13,8 +13,8 @@ from sglang.test.test_utils import CustomTestCase
 
 register_npu_ci(
     est_time=400,
-    suite="nightly-4-npu-a3",
-    nightly=False,
+    suite="full-4-npu-a3",
+    nightly=True,
 )
 
 PROMPT = (

test/registered/ascend/reward_models/test_npu_llama_3_1_8b_v0_2.py

@@ -10,7 +10,7 @@ from sglang.test.ci.ci_register import register_npu_ci
 from sglang.test.runners import HFRunner, SRTRunner
 from sglang.test.test_utils import CustomTestCase
 
-register_npu_ci(est_time=400, suite="nightly-1-npu-a3", nightly=False)
+register_npu_ci(est_time=400, suite="full-1-npu-a3", nightly=True)
 
 MODELS = [
     (SKYWORK_REWARD_LLAMA_3_1_8B_V0_2_WEIGHTS_PATH, 1, 4e-2),

test/registered/ascend/vlm_models/test_npu_qwen3_vl_4b_instruct.py

@@ -4,7 +4,7 @@ from sglang.test.ascend.test_ascend_utils import QWEN3_VL_4B_INSTRUCT_WEIGHTS_PA
 from sglang.test.ascend.vlm_utils import TestVLMModels
 from sglang.test.ci.ci_register import register_npu_ci
 
-register_npu_ci(est_time=400, suite="nightly-4-npu-a3", nightly=True)
+register_npu_ci(est_time=400, suite="full-4-npu-a3", nightly=True)
 
 
 class TestQwen3VL4B(TestVLMModels):

test/run_suite.py

@@ -105,6 +105,11 @@ NIGHTLY_SUITES = {
         "nightly-4-npu-a3",
         "nightly-8-npu-a3",
         "nightly-16-npu-a3",
+        "full-1-npu-a3",
+        "full-2-npu-a3",
+        "full-4-npu-a3",
+        "full-8-npu-a3",
+        "full-16-npu-a3",
     ],
 }