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

[CI] Re-enable GB300 jobs (#37522)

Browse Source
This commit is contained in:
Alison Shao
2026-09-02 16:15:40 -07:00
committed by GitHub
parent 718bd39fe0
commit a6da3a4922
3 changed files with 5 additions and 10 deletions
Download Patch File Download Diff File Expand all files Collapse all files
.github/workflows/_pr-test-stage.yml
-2
View File
@@ -100,14 +100,12 @@ jobs:
run: run:
name: ${{ inputs.self_name }} (${{ matrix.partition }}) name: ${{ inputs.self_name }} (${{ matrix.partition }})
# Runs on schedule / parallel-dispatch / non-failed PR with main_package or sgl_kernel changes. # Runs on schedule / parallel-dispatch / non-failed PR with main_package or sgl_kernel changes.
# Temporarily skip the broken GB300 runner before GitHub tries to allocate it.
# `runner_filter` is matched here rather than in each caller's job `if`: a # `runner_filter` is matched here rather than in each caller's job `if`: a
# caller that declares its runners as a matrix cannot, since job-level `if` # caller that declares its runners as a matrix cannot, since job-level `if`
# sees github/needs/vars/inputs but not `matrix`. Callers that never declare # sees github/needs/vars/inputs but not `matrix`. Callers that never declare
# the input read back null and match every runner_config. # the input read back null and match every runner_config.
if: | if: |
always() && always() &&
inputs.runner_config != '4-gpu-gb300' &&
(fromJson(inputs.caller_inputs).runner_filter == null || (fromJson(inputs.caller_inputs).runner_filter == null ||
fromJson(inputs.caller_inputs).runner_filter == '' || fromJson(inputs.caller_inputs).runner_filter == '' ||
fromJson(inputs.caller_inputs).runner_filter == 'all' || fromJson(inputs.caller_inputs).runner_filter == 'all' ||
.github/workflows/release-whl-deepgemm.yml
+3 -4
View File
@@ -194,10 +194,9 @@ jobs:
# - arch_label: sm120 # - arch_label: sm120
# runner: 1-gpu-5090 # runner: 1-gpu-5090
# wheel_arch: x86_64 # wheel_arch: x86_64
# Temporarily disabled while the GB300 runner is unavailable. - arch_label: sm100-aarch64
# - arch_label: sm100-aarch64 runner: 4-gpu-gb300
# runner: 4-gpu-gb300 wheel_arch: aarch64
# wheel_arch: aarch64
runs-on: ${{ matrix.runner }} runs-on: ${{ matrix.runner }}
timeout-minutes: 120 timeout-minutes: 120
steps: steps:
.github/workflows/rerun-test.yml
+2 -4
View File
@@ -87,8 +87,7 @@ permissions:
jobs: jobs:
rerun-test-cuda: rerun-test-cuda:
# Temporarily reject reruns targeting the broken GB300 runner. if: inputs.mode == 'cuda'
if: inputs.mode == 'cuda' && inputs.runs_on != '4-gpu-gb300'
runs-on: ${{ inputs.runs_on }} runs-on: ${{ inputs.runs_on }}
timeout-minutes: 120 timeout-minutes: 120
permissions: permissions:
@@ -200,8 +199,7 @@ jobs:
if: failure() if: failure()
rerun-test-multimodal-gen: rerun-test-multimodal-gen:
# Temporarily reject reruns targeting the broken GB300 runner. if: inputs.mode == 'multimodal_gen'
if: inputs.mode == 'multimodal_gen' && inputs.runs_on != '4-gpu-gb300'
runs-on: ${{ inputs.runs_on }} runs-on: ${{ inputs.runs_on }}
timeout-minutes: 120 timeout-minutes: 120
permissions: permissions: