minke.yu/sglang
1
0
Fork 0
Code Issues Pull Requests Actions 12 Packages Projects Releases Wiki Activity

Revert "[ci] split stage-c-test-4-gpu-b200 to enable a low-disk runner pool" (#24163)

Browse Source 
Co-authored-by: Alison Shao <alisonshao@radixark.ai>
This commit is contained in:
Alison Shao
2026-04-30 15:57:19 -07:00
committed by GitHub
co-authored by Alison Shao
parent cdc4078815
commit 694ef516cb
17 changed files with 29 additions and 99 deletions
Download Patch File Download Diff File Expand all files Collapse all files
.github/workflows/pr-test.yml
+15 -82
View File
@@ -101,7 +101,6 @@ jobs:
max_parallel_small: ${{ steps.set-parallel.outputs.max_parallel_small }}
max_parallel_2gpu: ${{ steps.set-parallel.outputs.max_parallel_2gpu }}
b200_runner: ${{ steps.set-runner.outputs.b200_runner }}
b200_low_disk_runner: ${{ steps.set-runner.outputs.b200_low_disk_runner }}
enable_retry: ${{ steps.set-retry.outputs.enable_retry }}
continue_on_error: ${{ steps.set-continue-on-error.outputs.continue_on_error }}
steps:
@@ -283,10 +282,8 @@ jobs:
target_stage="${{ inputs.target_stage }}"
if [[ "$sgl_kernel" == "true" && -z "$target_stage" ]]; then
echo "b200_runner=4-gpu-b200-kernel" >> $GITHUB_OUTPUT
echo "b200_low_disk_runner=4-gpu-b200-kernel-low-disk" >> $GITHUB_OUTPUT
else
echo "b200_runner=4-gpu-b200" >> $GITHUB_OUTPUT
echo "b200_low_disk_runner=4-gpu-b200-low-disk" >> $GITHUB_OUTPUT
fi
- name: Enable retry for CI
@@ -337,20 +334,19 @@ jobs:
{
echo "## Change Detection"
echo ""
echo "| Component | Changed |"
echo "|----------------------|---------|"
echo "| main_package | ${{ steps.filter-api.outputs.main_package || steps.filter.outputs.main_package || steps.run-mode.outputs.run_all_tests }} |"
echo "| sgl_kernel (raw) | ${{ steps.filter-api.outputs.sgl_kernel || steps.filter.outputs.sgl_kernel }} |"
echo "| sgl_kernel (used) | ${{ (!inputs.target_stage || inputs.include_wheel_build) && (steps.filter-api.outputs.sgl_kernel || steps.filter.outputs.sgl_kernel) }} |"
echo "| jit_kernel | ${{ steps.filter-api.outputs.jit_kernel || steps.filter.outputs.jit_kernel || steps.run-mode.outputs.run_all_tests }} |"
echo "| multimodal_gen | ${{ steps.filter-api.outputs.multimodal_gen || steps.filter.outputs.multimodal_gen || steps.run-mode.outputs.run_all_tests }} |"
echo "| target_stage | ${{ inputs.target_stage || '(none)' }} |"
echo "| detection_method | ${{ inputs.target_stage && 'GitHub API' || 'dorny/paths-filter' }} |"
echo "| max_parallel | ${{ steps.set-parallel.outputs.parallel_level }} (h100=${{ steps.set-parallel.outputs.max_parallel }}, 5090=${{ steps.set-parallel.outputs.max_parallel_small }}, 2gpu=${{ steps.set-parallel.outputs.max_parallel_2gpu }}) |"
echo "| b200_runner | ${{ steps.set-runner.outputs.b200_runner }} |"
echo "| b200_low_disk_runner | ${{ steps.set-runner.outputs.b200_low_disk_runner }} |"
echo "| enable_retry | ${{ steps.set-retry.outputs.enable_retry }} |"
echo "| continue_on_error | ${{ steps.set-continue-on-error.outputs.continue_on_error }} |"
echo "| Component | Changed |"
echo "|-------------------|---------|"
echo "| main_package | ${{ steps.filter-api.outputs.main_package || steps.filter.outputs.main_package || steps.run-mode.outputs.run_all_tests }} |"
echo "| sgl_kernel (raw) | ${{ steps.filter-api.outputs.sgl_kernel || steps.filter.outputs.sgl_kernel }} |"
echo "| sgl_kernel (used) | ${{ (!inputs.target_stage || inputs.include_wheel_build) && (steps.filter-api.outputs.sgl_kernel || steps.filter.outputs.sgl_kernel) }} |"
echo "| jit_kernel | ${{ steps.filter-api.outputs.jit_kernel || steps.filter.outputs.jit_kernel || steps.run-mode.outputs.run_all_tests }} |"
echo "| multimodal_gen | ${{ steps.filter-api.outputs.multimodal_gen || steps.filter.outputs.multimodal_gen || steps.run-mode.outputs.run_all_tests }} |"
echo "| target_stage | ${{ inputs.target_stage || '(none)' }} |"
echo "| detection_method | ${{ inputs.target_stage && 'GitHub API' || 'dorny/paths-filter' }} |"
echo "| max_parallel | ${{ steps.set-parallel.outputs.parallel_level }} (h100=${{ steps.set-parallel.outputs.max_parallel }}, 5090=${{ steps.set-parallel.outputs.max_parallel_small }}, 2gpu=${{ steps.set-parallel.outputs.max_parallel_2gpu }}) |"
echo "| b200_runner | ${{ steps.set-runner.outputs.b200_runner }} |"
echo "| enable_retry | ${{ steps.set-retry.outputs.enable_retry }} |"
echo "| continue_on_error | ${{ steps.set-continue-on-error.outputs.continue_on_error }} |"
} >> $GITHUB_STEP_SUMMARY
# =============================================== Wait Jobs for Sequential PR Execution ====================================================
@@ -1382,7 +1378,7 @@ jobs:
strategy:
fail-fast: false
matrix:
part: [0, 1, 2]
part: [0, 1, 2, 3, 4, 5]
steps:
- name: Checkout code
@@ -1413,69 +1409,7 @@ jobs:
CONTINUE_ON_ERROR_FLAG: ${{ needs.check-changes.outputs.continue_on_error == 'true' && '--continue-on-error' || '' }}
run: |
cd test
python3 run_suite.py --hw cuda --suite stage-c-test-4-gpu-b200 --auto-partition-id ${{ matrix.part }} --auto-partition-size 3 --timeout-per-file 1800 $CONTINUE_ON_ERROR_FLAG
- uses: ./.github/actions/upload-cuda-coredumps
if: failure()
with:
artifact-suffix: ${{ matrix.part }}
- name: Cleanup venv
if: always()
run: bash scripts/ci/cuda/ci_cleanup_venv.sh
stage-c-test-4-gpu-b200-small:
needs: [check-changes, call-gate, wait-for-stage-b, sgl-kernel-build-wheels]
if: |
always() &&
(
(inputs.target_stage == 'stage-c-test-4-gpu-b200-small') ||
(
!inputs.target_stage &&
((github.event_name == 'schedule' || inputs.test_parallel_dispatch == true) || (!failure() && !cancelled())) &&
((needs.check-changes.outputs.main_package == 'true') || (needs.check-changes.outputs.sgl_kernel == 'true'))
)
)
# The `*-low-disk` label (resolved by `set-runner` to `4-gpu-b200-low-disk` or
# `4-gpu-b200-kernel-low-disk`) is advertised by both the existing large-disk B200
# runners and the new low-disk runner, so this job can land on either pool.
runs-on: ${{ needs.check-changes.outputs.b200_low_disk_runner }}
timeout-minutes: 240
strategy:
fail-fast: false
matrix:
part: [0, 1, 2]
steps:
- name: Checkout code
uses: actions/checkout@v4
with:
ref: ${{ inputs.pr_head_sha || inputs.git_ref || github.sha }}
- uses: ./.github/actions/check-stage-health
- uses: ./.github/actions/check-maintenance
- name: Download artifacts
if: needs.check-changes.outputs.sgl_kernel == 'true'
uses: actions/download-artifact@v6
with:
path: sgl-kernel/dist/
merge-multiple: true
pattern: wheel-python3.10-cuda*
- name: Install dependencies
timeout-minutes: 20
run: |
CUSTOM_BUILD_SGL_KERNEL=${{needs.check-changes.outputs.sgl_kernel}} bash scripts/ci/cuda/ci_install_dependency.sh
- name: Run test
timeout-minutes: 30
env:
CONTINUE_ON_ERROR_FLAG: ${{ needs.check-changes.outputs.continue_on_error == 'true' && '--continue-on-error' || '' }}
run: |
cd test
python3 run_suite.py --hw cuda --suite stage-c-test-4-gpu-b200-small --auto-partition-id ${{ matrix.part }} --auto-partition-size 3 --timeout-per-file 1800 $CONTINUE_ON_ERROR_FLAG
python3 run_suite.py --hw cuda --suite stage-c-test-4-gpu-b200 --auto-partition-id ${{ matrix.part }} --auto-partition-size 6 --timeout-per-file 1800 $CONTINUE_ON_ERROR_FLAG
- uses: ./.github/actions/upload-cuda-coredumps
if: failure()
@@ -1567,7 +1501,6 @@ jobs:
stage-c-test-deepep-4-gpu-h100,
stage-c-test-deepep-8-gpu-h200,
stage-c-test-4-gpu-b200,
stage-c-test-4-gpu-b200-small,
# stage-c-test-4-gpu-gb200, # Temporarily disabled — no GB200 runner
]
if: always()
scripts/ci/utils/slash_command_handler.py
-2
View File
@@ -303,7 +303,6 @@ def handle_rerun_stage(
"stage-c-test-8-gpu-h200",
"stage-c-test-8-gpu-h20",
"stage-c-test-4-gpu-b200",
"stage-c-test-4-gpu-b200-small",
"stage-c-test-4-gpu-gb200",
"stage-c-test-deepep-4-gpu-h100",
"stage-c-test-deepep-8-gpu-h200",
@@ -486,7 +485,6 @@ CUDA_SUITE_TO_RUNNER = {
"stage-c-test-8-gpu-h200": "8-gpu-h200",
"stage-c-test-8-gpu-h20": "8-gpu-h20",
"stage-c-test-4-gpu-b200": "4-gpu-b200",
"stage-c-test-4-gpu-b200-small": "4-gpu-b200-low-disk",
"stage-c-test-deepep-4-gpu-h100": "4-gpu-h100",
"stage-c-test-deepep-8-gpu-h200": "8-gpu-h200",
# Nightly test suites (NVIDIA)
test/registered/4-gpu-models/test_gpt_oss_4gpu.py
+1 -1
View File
@@ -4,7 +4,7 @@ from sglang.test.ci.ci_register import register_cuda_ci
from sglang.test.gpt_oss_common import BaseTestGptOss
register_cuda_ci(est_time=392, suite="stage-c-test-4-gpu-h100")
register_cuda_ci(est_time=584, suite="stage-c-test-4-gpu-b200-small")
register_cuda_ci(est_time=740, suite="stage-c-test-4-gpu-b200")
class TestGptOss4Gpu(BaseTestGptOss):
test/registered/4-gpu-models/test_nvidia_nemotron_3_super_nvfp4.py
+1 -1
View File
@@ -11,7 +11,7 @@ from sglang.test.test_utils import (
popen_launch_server,
)
register_cuda_ci(est_time=710, suite="nightly-4-gpu-b200", nightly=True)
register_cuda_ci(est_time=710, suite="stage-c-test-4-gpu-b200")
NEMOTRON_3_SUPER_NVFP4_MODEL = "nvidia/NVIDIA-Nemotron-3-Super-120B-A12B-NVFP4"
test/registered/4-gpu-models/test_qwen35_fp4_mtp_v2.py
+1 -1
View File
@@ -15,7 +15,7 @@ from sglang.test.test_utils import (
popen_launch_server,
)
register_cuda_ci(est_time=422, suite="stage-c-test-4-gpu-b200-small")
register_cuda_ci(est_time=540, suite="stage-c-test-4-gpu-b200")
QWEN35_FP4_MODEL = "nvidia/Qwen3.5-397B-A17B-NVFP4"
ACC_THRESHOLDS = {QWEN35_FP4_MODEL: {"gsm8k": 0.95}}
test/registered/4-gpu-models/test_qwen35_fp4_triton.py
+1 -1
View File
@@ -11,7 +11,7 @@ from sglang.test.test_utils import (
ModelLaunchSettings,
)
register_cuda_ci(est_time=563, suite="stage-c-test-4-gpu-b200-small")
register_cuda_ci(est_time=720, suite="stage-c-test-4-gpu-b200")
QWEN35_FP4_MODEL = "nvidia/Qwen3.5-397B-A17B-NVFP4"
ACC_THRESHOLDS = {QWEN35_FP4_MODEL: {"gsm8k": 0.95}}
test/registered/lora/test_lora_gpt_oss_20b_logprob_diff.py
+1 -1
View File
@@ -36,7 +36,7 @@ from sglang.test.test_utils import CustomTestCase
register_cuda_ci(
est_time=300,
suite="stage-c-test-4-gpu-b200-small",
suite="stage-c-test-4-gpu-b200",
)
BASE_MODEL = "lmsys/gpt-oss-20b-bf16"
test/registered/lora/test_lora_nemotron_3_super_120b_a12b_logprob_diff.py
+1 -1
View File
@@ -36,7 +36,7 @@ from sglang.test.test_utils import CustomTestCase
register_cuda_ci(
est_time=300,
suite="stage-c-test-4-gpu-b200-small",
suite="stage-c-test-4-gpu-b200",
)
BASE_MODEL = "nvidia/NVIDIA-Nemotron-3-Super-120B-A12B-BF16"
test/registered/lora/test_lora_qwen3_30b_a3b_instruct_2507_logprob_diff.py
+1 -1
View File
@@ -36,7 +36,7 @@ from sglang.test.test_utils import CustomTestCase
register_cuda_ci(
est_time=160,
suite="stage-c-test-4-gpu-b200-small",
suite="stage-c-test-4-gpu-b200",
)
BASE_MODEL = "Qwen/Qwen3-30B-A3B-Instruct-2507"
test/registered/lora/test_lora_qwen3_5_35b_a3b_logprob_diff.py
+1 -1
View File
@@ -36,7 +36,7 @@ from sglang.test.test_utils import CustomTestCase
register_cuda_ci(
est_time=160,
suite="stage-c-test-4-gpu-b200-small",
suite="stage-c-test-4-gpu-b200",
)
BASE_MODEL = "Qwen/Qwen3.5-35B-A3B"
test/registered/lora/test_lora_qwen3_vl_30b_a3b_instruct_logprob_diff.py
+1 -1
View File
@@ -36,7 +36,7 @@ from sglang.test.test_utils import CustomTestCase
register_cuda_ci(
est_time=160,
suite="stage-c-test-4-gpu-b200-small",
suite="stage-c-test-4-gpu-b200",
)
BASE_MODEL = "Qwen/Qwen3-VL-30B-A3B-Instruct"
test/registered/moe/test_cutedsl_moe.py
+1 -1
View File
@@ -16,7 +16,7 @@ except ImportError:
CuteDslMoEWrapper = None
convert_sf_to_mma_layout = None
register_cuda_ci(est_time=427, suite="stage-c-test-4-gpu-b200-small")
register_cuda_ci(est_time=590, suite="stage-c-test-4-gpu-b200")
SKIP_TEST = torch.cuda.get_device_capability() < (10, 0)
SKIP_REASON = "Nvfp4 Requires compute capability of 10 or above."
test/registered/quant/test_fp8_blockwise_gemm.py
+1 -1
View File
@@ -12,7 +12,7 @@ from sglang.test.test_utils import (
try_cached_model,
)
register_cuda_ci(est_time=630, suite="nightly-4-gpu-b200", nightly=True)
register_cuda_ci(est_time=630, suite="stage-c-test-4-gpu-b200")
MODEL_PATH = "Qwen/Qwen3-4B-Instruct-2507-FP8"
MXFP8_MODEL_PATH = "zianglih/Qwen3-4B-Instruct-2507-MXFP8"
test/registered/quant/test_nvfp4_gemm.py
+1 -1
View File
@@ -12,7 +12,7 @@ from sglang.test.test_utils import (
try_cached_model,
)
register_cuda_ci(est_time=550, suite="nightly-4-gpu-b200", nightly=True)
register_cuda_ci(est_time=550, suite="stage-c-test-4-gpu-b200")
MODEL_PATH = "nvidia/Llama-3.1-8B-Instruct-NVFP4"
test/registered/rl/test_update_weights_from_disk_blackwell.py
+1 -1
View File
@@ -1,6 +1,6 @@
from sglang.test.ci.ci_register import register_cuda_ci
register_cuda_ci(est_time=400, suite="stage-c-test-4-gpu-b200-small")
register_cuda_ci(est_time=400, suite="stage-c-test-4-gpu-b200")
import unittest
test/registered/spec/eagle/test_eagle_infer_beta_dp_attention.py
+1 -1
View File
@@ -17,7 +17,7 @@ from sglang.test.test_utils import (
)
# EAGLE with DP attention on B200 (tp=2, dp=2, requires 4 B200 GPUs)
register_cuda_ci(est_time=123, suite="stage-c-test-4-gpu-b200-small")
register_cuda_ci(est_time=123, suite="stage-c-test-4-gpu-b200")
def test_gsm8k(base_url: str, model: str):
test/run_suite.py
-1
View File
@@ -48,7 +48,6 @@ PER_COMMIT_SUITES = {
"stage-b-kernel-benchmark-1-gpu-large",
"stage-c-test-4-gpu-h100",
"stage-c-test-4-gpu-b200",
"stage-c-test-4-gpu-b200-small",
"stage-c-test-4-gpu-gb200",
"stage-c-test-8-gpu-h20",
"stage-c-test-8-gpu-h200",