chore: update CI test est_time values (#38238)

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

test/registered/unit/layers/quantization/test_compressed_tensors_kv_cache.py

@@ -2,7 +2,7 @@
 
 from sglang.test.ci.ci_register import register_cpu_ci
 
-register_cpu_ci(est_time=5, suite="base-a-test-cpu")
+register_cpu_ci(est_time=10, suite="base-a-test-cpu")
 
 import unittest
 

test/registered/unit/layers/quantization/test_compressed_tensors_lm_head.py

@@ -2,7 +2,7 @@
 
 from sglang.test.ci.ci_register import register_cpu_ci
 
-register_cpu_ci(est_time=5, suite="base-a-test-cpu")
+register_cpu_ci(est_time=11, suite="base-a-test-cpu")
 
 import unittest
 from unittest.mock import patch

test/registered/unit/layers/quantization/test_compressed_tensors_mixed_precision.py

@@ -9,7 +9,7 @@ for every group. Both are config-parsing paths, so these tests run on CPU.
 
 from sglang.test.ci.ci_register import register_cpu_ci
 
-register_cpu_ci(est_time=5, suite="base-a-test-cpu")
+register_cpu_ci(est_time=10, suite="base-a-test-cpu")
 
 import unittest
 from unittest import mock

test/registered/unit/layers/quantization/test_compressed_tensors_wna16_moe_no_linear.py

@@ -16,7 +16,7 @@ from sglang.srt.runtime_context import override_platform
 from sglang.test.ci.ci_register import register_cpu_ci
 from sglang.test.test_utils import CustomTestCase
 
-register_cpu_ci(est_time=5, suite="base-a-test-cpu")
+register_cpu_ci(est_time=10, suite="base-a-test-cpu")
 
 _WNA16_MOE_SCHEMES = (CompressedTensorsWNA16MoE, CompressedTensorsWNA16TritonMoE)
 EXPERTS_LAYER = "model.layers.0.mlp.experts"

test/registered/unit/layers/quantization/test_deepgemm_ue8m0_requant.py

@@ -2,7 +2,7 @@
 
 from sglang.test.ci.ci_register import register_cpu_ci
 
-register_cpu_ci(est_time=5, suite="base-a-test-cpu")
+register_cpu_ci(est_time=10, suite="base-a-test-cpu")
 
 import unittest
 from unittest.mock import call, patch

test/registered/unit/layers/quantization/test_flashinfer_pr4266_bf16_gemm.py

@@ -9,7 +9,7 @@ from sglang.srt.layers.quantization.unquant import (
 )
 from sglang.test.ci.ci_register import register_cpu_ci
 
-register_cpu_ci(est_time=5, suite="base-a-test-cpu")
+register_cpu_ci(est_time=11, suite="base-a-test-cpu")
 
 
 @pytest.mark.parametrize("m,n,k", _FLASHINFER_PR4266_TUNED_TACTICS)

test/registered/unit/layers/quantization/test_flashinfer_trtllm_fp8_fallback.py

@@ -20,7 +20,7 @@ backend selector and the two GEMM implementations so they run on CPU CI.
 
 from sglang.test.ci.ci_register import register_cpu_ci
 
-register_cpu_ci(est_time=5, suite="base-a-test-cpu")
+register_cpu_ci(est_time=10, suite="base-a-test-cpu")
 
 import unittest
 from unittest.mock import MagicMock, patch

test/registered/unit/layers/quantization/test_fp4_kv_cache_quant_method.py

@@ -2,7 +2,7 @@
 
 from sglang.test.ci.ci_register import register_cpu_ci
 
-register_cpu_ci(est_time=5, suite="base-a-test-cpu")
+register_cpu_ci(est_time=35, suite="base-a-test-cpu")
 
 import unittest
 

test/registered/unit/layers/quantization/test_fp8_blockwise_linear_backends.py

@@ -24,9 +24,9 @@ from sglang.test.layer_ut_utils import (
 )
 from sglang.test.test_utils import CustomTestCase
 
-register_cuda_ci(est_time=120, stage="base-b", runner_config="4-gpu-b200")
-register_cuda_ci(est_time=60, stage="base-b", runner_config="1-gpu-small")
-register_cuda_ci(est_time=60, stage="base-b", runner_config="1-gpu-large")
+register_cuda_ci(est_time=15, stage="base-b", runner_config="4-gpu-b200")
+register_cuda_ci(est_time=11, stage="base-b", runner_config="1-gpu-small")
+register_cuda_ci(est_time=12, stage="base-b", runner_config="1-gpu-large")
 
 FP8_MAX = 448.0
 

test/registered/unit/layers/quantization/test_fp8_moe_runner_ownership.py

@@ -20,7 +20,7 @@ from sglang.srt.runtime_context import get_flags
 from sglang.test.ci.ci_register import register_cpu_ci
 from sglang.test.test_utils import CustomTestCase
 
-register_cpu_ci(est_time=2, suite="base-a-test-cpu")
+register_cpu_ci(est_time=11, suite="base-a-test-cpu")
 
 _ACTIVATION_PARAMS = ("gemm1_alpha", "gemm1_beta", "gemm1_clamp_limit")
 

test/registered/unit/layers/quantization/test_fp8_moe_weight_gating.py

@@ -9,7 +9,7 @@ follow the same shard count, or the two describe different tensors.
 
 from sglang.test.ci.ci_register import register_cpu_ci
 
-register_cpu_ci(est_time=5, suite="base-a-test-cpu")
+register_cpu_ci(est_time=10, suite="base-a-test-cpu")
 
 import unittest
 from unittest.mock import MagicMock, patch

test/registered/unit/layers/quantization/test_fp8_utils_mxfp4.py

@@ -11,7 +11,7 @@ from sglang.srt.layers.quantization.fp8_utils import (
 from sglang.test.ci.ci_register import register_cpu_ci
 from sglang.test.test_utils import CustomTestCase
 
-register_cpu_ci(est_time=4, suite="base-a-test-cpu")
+register_cpu_ci(est_time=10, suite="base-a-test-cpu")
 
 
 class TestFp8UtilsMxfp4(CustomTestCase):

test/registered/unit/layers/quantization/test_gptq_scheme_attach.py

@@ -18,7 +18,7 @@ from sglang.srt.layers.vocab_parallel_embedding import VocabParallelEmbedding
 from sglang.test.ci.ci_register import register_cpu_ci
 from sglang.test.test_utils import CustomTestCase
 
-register_cpu_ci(est_time=5, suite="base-a-test-cpu")
+register_cpu_ci(est_time=11, suite="base-a-test-cpu")
 
 _GPTQ_CHECKPOINT_CONFIG = {
     "bits": 4,

test/registered/unit/layers/quantization/test_humming_w4afp8_schemas.py

@@ -19,7 +19,7 @@ from sglang.srt.layers.quantization.humming import (  # noqa: E402
     _W4AFp8CheckpointWeightSchema,
 )
 
-register_cpu_ci(est_time=3, suite="base-a-test-cpu")
+register_cpu_ci(est_time=12, suite="base-a-test-cpu")
 
 
 class TestW4AFp8CheckpointSchema(CustomTestCase):

test/registered/unit/layers/quantization/test_int8_linear_methods.py

@@ -23,7 +23,7 @@ from sglang.test.layer_ut_utils import (
 )
 from sglang.test.test_utils import CustomTestCase
 
-register_cuda_ci(est_time=60, stage="base-b", runner_config="1-gpu-large")
+register_cuda_ci(est_time=10, stage="base-b", runner_config="1-gpu-large")
 
 INT8_MAX = 127.0
 

test/registered/unit/layers/quantization/test_marlin_utils_fp8.py

@@ -9,7 +9,7 @@ from sglang.srt.layers.quantization import marlin_utils_fp8
 from sglang.test.ci.ci_register import register_cpu_ci
 from sglang.test.test_utils import CustomTestCase
 
-register_cpu_ci(est_time=5, suite="base-a-test-cpu")
+register_cpu_ci(est_time=11, suite="base-a-test-cpu")
 
 
 class TestFp8MarlinBias(CustomTestCase):

test/registered/unit/layers/quantization/test_modelopt_nvfp4.py

@@ -14,7 +14,7 @@ from sglang.srt.model_loader.weight_utils import default_weight_loader
 from sglang.test.ci.ci_register import register_cpu_ci
 from sglang.test.test_utils import CustomTestCase
 
-register_cpu_ci(est_time=5, suite="base-a-test-cpu")
+register_cpu_ci(est_time=11, suite="base-a-test-cpu")
 
 
 class TestModelOptNvfp4(CustomTestCase):

test/registered/unit/layers/quantization/test_modelopt_nvfp4_moe_scales.py

@@ -18,7 +18,7 @@ sizes, so a failure looks like one a real checkpoint would hit.
 
 from sglang.test.ci.ci_register import register_cpu_ci
 
-register_cpu_ci(est_time=5, suite="base-a-test-cpu")
+register_cpu_ci(est_time=11, suite="base-a-test-cpu")
 
 import unittest
 

test/registered/unit/layers/quantization/test_mxfp4_flashinfer_activation_prep.py

@@ -12,7 +12,7 @@ from sglang.srt.layers.quantization.mxfp4 import (
 from sglang.test.ci.ci_register import register_cpu_ci
 from sglang.test.test_utils import CustomTestCase
 
-register_cpu_ci(est_time=2, suite="base-a-test-cpu")
+register_cpu_ci(est_time=10, suite="base-a-test-cpu")
 
 per_token_group_quant_module = importlib.import_module(
     "sglang.kernels.ops.quantization.per_token_group_quant"

test/registered/unit/layers/quantization/test_mxfp4_sm120_cutlass.py

@@ -14,7 +14,7 @@ import torch
 from sglang.srt.runtime_context import override_platform
 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=14, stage="base-b", runner_config="1-gpu-small")
 
 
 def _random_weights(num_experts: int, hidden: int, intermediate: int):

test/registered/unit/layers/quantization/test_mxfp4_sm90_cutlass.py

@@ -20,7 +20,7 @@ import torch
 
 from sglang.test.ci.ci_register import register_cuda_ci
 
-register_cuda_ci(est_time=120, stage="base-b", runner_config="1-gpu-large")
+register_cuda_ci(est_time=11, stage="base-b", runner_config="1-gpu-large")
 
 flashinfer_fused_moe = pytest.importorskip("flashinfer.fused_moe")
 

test/registered/unit/layers/quantization/test_nvfp4_linear_backends.py

@@ -30,7 +30,7 @@ from sglang.test.quant_ref_utils import (
 )
 from sglang.test.test_utils import CustomTestCase
 
-register_cuda_ci(est_time=120, stage="base-b", runner_config="4-gpu-b200")
+register_cuda_ci(est_time=11, stage="base-b", runner_config="4-gpu-b200")
 
 # (M, N, K). The second shape hits the padding paths: N=160 is not a multiple
 # of 128 (TRTLLM shuffle pad) and K=336 is neither a multiple of 32 (CUTLASS

test/registered/unit/layers/quantization/test_nvfp4_moe_backends.py

@@ -23,7 +23,7 @@ from sglang.test.quant_ref_utils import (
 )
 from sglang.test.test_utils import CustomTestCase
 
-register_cuda_ci(est_time=120, stage="base-b", runner_config="4-gpu-b200")
+register_cuda_ci(est_time=14, stage="base-b", runner_config="4-gpu-b200")
 
 E, H, I, TOPK, M = 8, 1024, 1024, 2, 32
 

test/registered/unit/layers/quantization/test_quark_config.py

@@ -2,7 +2,7 @@
 
 from sglang.test.ci.ci_register import register_cpu_ci
 
-register_cpu_ci(est_time=5, suite="base-a-test-cpu")
+register_cpu_ci(est_time=12, suite="base-a-test-cpu")
 
 import unittest
 from unittest.mock import patch

test/registered/unit/layers/quantization/test_quark_utils.py

@@ -2,7 +2,7 @@
 
 from sglang.test.ci.ci_register import register_cpu_ci
 
-register_cpu_ci(est_time=5, suite="base-a-test-cpu")
+register_cpu_ci(est_time=10, suite="base-a-test-cpu")
 
 import unittest