chenchenghao/sglang
1
0
Fork 0
Code Issues Pull Requests Actions 11 Packages Projects Releases Wiki Activity

ci: move metrics scripts under scripts/ci/utils (#20986)

Browse Source
This commit is contained in:
Lianmin Zheng
2026-03-19 23:47:57 -07:00
committed by GitHub
parent 0be88ac44c
commit 712a48c5d2
9 changed files with 26 additions and 27 deletions
Download Patch File Download Diff File Expand all files Collapse all files

8
.github/workflows/amd-ci-job-monitor.yml vendored
View File

@@ -6,7 +6,7 @@ on:
pull_request:
paths:
- '.github/workflows/amd-ci-job-monitor.yml'
- 'scripts/ci/query_job_status.py'
- 'scripts/ci/utils/query_job_status.py'
workflow_dispatch:
inputs:
hours:
@@ -42,7 +42,7 @@ jobs:
- name: Generate Custom Job Report
timeout-minutes: 30
run: |
python scripts/ci/query_job_status.py \
python scripts/ci/utils/query_job_status.py \
--repo ${{ github.repository }} \
--job "${{ inputs.job_filter }}" \
--workflow "pr-test-amd.yml" \
@@ -107,7 +107,7 @@ jobs:
- name: Generate Report
timeout-minutes: 15
run: |
python scripts/ci/query_job_status.py \
python scripts/ci/utils/query_job_status.py \
--repo ${{ github.repository }} \
--job "${{ matrix.job_name }}" \
--workflow "pr-test-amd.yml" \
@@ -141,7 +141,7 @@ jobs:
- name: Generate Nightly Report
timeout-minutes: 15
run: |
python scripts/ci/query_job_status.py \
python scripts/ci/utils/query_job_status.py \
--repo ${{ github.repository }} \
--job "${{ matrix.job_name }}" \
--workflow "nightly-test-amd.yml" \

10
.github/workflows/nightly-test-nvidia.yml vendored
View File

@@ -157,7 +157,7 @@ jobs:
- name: Collect performance metrics
if: always()
run: |
python3 scripts/ci/save_metrics.py \
python3 scripts/ci/utils/save_metrics.py \
--gpu-config 8-gpu-h200 \
--partition ${{ matrix.partition }} \
--run-id ${{ github.run_id }} \
@@ -257,7 +257,7 @@ jobs:
- name: Collect performance metrics
if: always()
run: |
python3 scripts/ci/save_metrics.py \
python3 scripts/ci/utils/save_metrics.py \
--gpu-config 8-gpu-b200 \
--partition ${{ matrix.partition }} \
--run-id ${{ github.run_id }} \
@@ -434,7 +434,7 @@ jobs:
- name: Collect diffusion performance metrics
if: always()
run: |
python3 scripts/ci/save_diffusion_metrics.py \
python3 scripts/ci/utils/save_diffusion_metrics.py \
--gpu-config 1-gpu-runner \
--run-id ${{ github.run_id }} \
--output python/diffusion-metrics-1gpu-partition-${{ matrix.part }}.json \
@@ -490,7 +490,7 @@ jobs:
- name: Collect diffusion performance metrics
if: always()
run: |
python3 scripts/ci/save_diffusion_metrics.py \
python3 scripts/ci/utils/save_diffusion_metrics.py \
--gpu-config 2-gpu-runner \
--run-id ${{ github.run_id }} \
--output python/diffusion-metrics-2gpu-partition-${{ matrix.part }}.json \
@@ -589,7 +589,7 @@ jobs:
- name: Merge metrics
run: |
python3 scripts/ci/merge_metrics.py \
python3 scripts/ci/utils/merge_metrics.py \
--input-dir metrics/ \
--output consolidated-metrics-${{ github.run_id }}.json \
--run-id ${{ github.run_id }} \

18
.github/workflows/pr-test.yml vendored
View File

@@ -114,24 +114,24 @@ jobs:
with:
filters: |
main_package:
- "python/sglang/!(multimodal_gen)/**"
- ".github/workflows/pr-test.yml"
- "python/pyproject.toml"
- "python/sglang/!(multimodal_gen)/**"
- "scripts/ci/cuda/*"
- "scripts/ci/utils/*"
- "test/**"
- ".github/workflows/pr-test.yml"
sgl_kernel:
- "sgl-kernel/**"
jit_kernel:
- "python/sglang/jit_kernel/**"
- "python/pyproject.toml"
- ".github/workflows/pr-test.yml"
multimodal_gen:
- ".github/workflows/pr-test.yml"
- "python/pyproject.toml"
- "python/sglang/multimodal_gen/**"
- "python/sglang/jit_kernel/**"
- "python/sglang/cli/**"
- "python/pyproject.toml"
jit_kernel:
- ".github/workflows/pr-test.yml"
- "python/pyproject.toml"
- "python/sglang/jit_kernel/**"
sgl_kernel:
- "sgl-kernel/**"
# For /rerun-stage (workflow_dispatch with target_stage), dorny/paths-filter doesn't work
# correctly because it falls back to "last commit" detection which breaks for merge commits.

2
python/sglang/test/nightly_utils.py
View File

@@ -209,7 +209,7 @@ class NightlyBenchmarkRunner:
)
# Note: JSON files are preserved for metrics collection by CI scripts
# They will be collected by scripts/ci/save_metrics.py
# They will be collected by scripts/ci/utils/save_metrics.py
return benchmark_results, True

2
scripts/ci/merge_metrics.py → scripts/ci/utils/merge_metrics.py
View File

@@ -5,7 +5,7 @@ This script reads all per-partition metric JSON files and consolidates them
into a single JSON file with run-level metadata.
Usage:
python3 scripts/ci/merge_metrics.py \
python3 scripts/ci/utils/merge_metrics.py \
--input-dir metrics/ \
--output consolidated-metrics-12345678.json \
--run-id 12345678 \

6
scripts/ci/query_job_status.py → scripts/ci/utils/query_job_status.py
View File

@@ -3,9 +3,9 @@
Query GitHub Actions job status for specific jobs.
Usage:
python scripts/ci/query_job_status.py --job "stage-c-test-large-8-gpu-amd-mi35x"
python scripts/ci/query_job_status.py --job "stage-c-test-large-8-gpu-amd-mi35x" --hours 48
python scripts/ci/query_job_status.py --job "AMD" --workflow pr-test-amd.yml
python scripts/ci/utils/query_job_status.py --job "stage-c-test-large-8-gpu-amd-mi35x"
python scripts/ci/utils/query_job_status.py --job "stage-c-test-large-8-gpu-amd-mi35x" --hours 48
python scripts/ci/utils/query_job_status.py --job "AMD" --workflow pr-test-amd.yml
Requirements:
pip install tabulate

2
scripts/ci/save_diffusion_metrics.py → scripts/ci/utils/save_diffusion_metrics.py
View File

@@ -5,7 +5,7 @@ This script reads diffusion test results from the pytest stash and saves them
with metadata for the performance dashboard.
Usage:
python3 scripts/ci/save_diffusion_metrics.py \
python3 scripts/ci/utils/save_diffusion_metrics.py \
--gpu-config 1-gpu-runner \
--run-id 12345678 \
--output test/diffusion-metrics-1gpu.json \

2
scripts/ci/save_metrics.py → scripts/ci/utils/save_metrics.py
View File

@@ -5,7 +5,7 @@ This script reads benchmark result JSON files from performance profile directori
and saves them with metadata for artifact collection in CI.
Usage:
python3 scripts/ci/save_metrics.py \
python3 scripts/ci/utils/save_metrics.py \
--gpu-config 8-gpu-h200 \
--partition 0 \
--run-id 12345678 \

3
test/show_partitions.py
View File

@@ -22,7 +22,6 @@ _spec = importlib.util.spec_from_file_location("ci_register", _ci_register_path)
_ci_register = importlib.util.module_from_spec(_spec)
_spec.loader.exec_module(_ci_register)
CIRegistry = _ci_register.CIRegistry
HWBackend = _ci_register.HWBackend
auto_partition = _ci_register.auto_partition
collect_tests = _ci_register.collect_tests
@@ -110,7 +109,7 @@ def main():
total_time = sum(t.est_time for t in enabled)
out(
f"### {suite} (hw={hw_str}, {partition_count} partition(s), "
f"### {suite} (hw={hw_str}, {partition_count} partitions, "
f"{len(enabled)} enabled, {len(disabled)} disabled, est {total_time:.0f}s total)"
)
out()