Files
sglang/.github/workflows/nightly-amd-mi355x-disagg.yml

296 lines
12 KiB
YAML

name: Nightly Test (AMD MI355X 2N 1P1D Disagg)
# Self-hosted runners on the amd-sglang login node salloc two or four MI355X
# nodes and run the Docker harness (prefill + decode over MORI, a standalone
# load balancer, then an sglang.bench_serving concurrency sweep) via
# scripts/ci/slurm/launch_mi355x.sh.
on:
# Temporarily paused while the DI CI nodes are unavailable. Uncomment the
# schedule below when the nodes return.
# schedule:
# - cron: '0 7 * * *'
workflow_dispatch:
inputs:
image:
description: 'Optional. ROCm SGLang Docker image to benchmark. Leave empty for the recipe default.'
required: false
default: ''
configs:
description: 'Optional. Comma-separated names to run only a subset. Leave empty to run all. Available names are listed in the setup job log.'
required: false
default: ''
concurrency:
group: nightly-test-mi355x
cancel-in-progress: false
env:
SGLANG_IS_IN_CI: true
jobs:
# ---------------------------------------------------------------------------
# Reads scripts/ci/slurm/nightly-configs.yaml and generates one matrix entry
# per recipe YAML for runner=mi355x. To add/remove configs, edit
# nightly-configs.yaml only.
# ---------------------------------------------------------------------------
setup:
if: github.repository == 'sgl-project/sglang'
runs-on: ubuntu-latest
outputs:
matrix-2n: ${{ steps.generate.outputs.matrix-2n }}
matrix-4n: ${{ steps.generate.outputs.matrix-4n }}
image: ${{ steps.image.outputs.image }}
steps:
- name: Checkout code
uses: actions/checkout@v4
- name: Resolve latest MI35x nightly image
id: image
# Track the newest published image so the nightly actually catches
# regressions instead of re-testing one pinned build. We resolve on the
# GitHub-hosted runner (clean Docker Hub access, no rate limits) and pass
# the tag to the self-hosted runner. Scope is deliberately tight: only
# the clean main MI35x ROCm 7.2 line
# `lmsysorg/sglang-rocm:v*-rocm720-mi35x-YYYYMMDD` (excludes -test-,
# miles-, mori-, vllm-, rocm7_14 variants). On any failure we emit an
# empty string and the launcher falls back to the recipe's pinned
# `runtime.image`.
run: |
set -o pipefail
RESOLVED=$(curl -fsSL "https://hub.docker.com/v2/repositories/lmsysorg/sglang-rocm/tags?page_size=100&ordering=last_updated" \
| python3 -c "import json,re,sys;p=re.compile(r'^v[0-9][0-9A-Za-z.]*-rocm720-mi35x-([0-9]{8})\$');c=sorted((m.group(1),t['name']) for t in json.load(sys.stdin).get('results',[]) for m in [p.match(t.get('name',''))] if m);print(c[-1][1] if c else '')" || true)
if [ -n "$RESOLVED" ]; then
echo "Resolved latest MI35x image: lmsysorg/sglang-rocm:$RESOLVED"
echo "image=lmsysorg/sglang-rocm:$RESOLVED" >> $GITHUB_OUTPUT
else
echo "Could not resolve a latest image; launcher will use recipe default."
echo "image=" >> $GITHUB_OUTPUT
fi
- name: Generate benchmark matrix
id: generate
env:
CONFIGS_FILTER: ${{ inputs.configs }}
run: |
pip install pyyaml -q
ALL_MATRIX=$(python3 scripts/ci/slurm/generate_matrix.py \
scripts/ci/slurm/nightly-configs.yaml --runner mi355x)
echo "Available config names for runner mi355x:"
echo "$ALL_MATRIX" | python3 -c "import json,sys; [print(f' - {e[\"name\"]}') for e in json.load(sys.stdin)]"
FILTER_ARG=()
if [ -n "$CONFIGS_FILTER" ]; then
echo ""
echo "Filtering to: $CONFIGS_FILTER"
FILTER_ARG=(--filter "$CONFIGS_FILTER")
fi
MATRIX=$(python3 scripts/ci/slurm/generate_matrix.py \
scripts/ci/slurm/nightly-configs.yaml --runner mi355x "${FILTER_ARG[@]}")
MATRIX="$MATRIX" python3 - <<'PY'
import json
import os
matrix = json.loads(os.environ["MATRIX"])
groups = {
"matrix-2n": [entry for entry in matrix if "-1p1d" in entry["name"]],
"matrix-4n": [
entry for entry in matrix if "-2p1d-ep16" in entry["name"]
],
}
matched = {entry["name"] for entries in groups.values() for entry in entries}
unmatched = [entry["name"] for entry in matrix if entry["name"] not in matched]
if unmatched:
raise SystemExit(
"Unrecognized MI355X topology for config(s): " + ", ".join(unmatched)
)
with open(os.environ["GITHUB_OUTPUT"], "a") as output:
for name, entries in groups.items():
value = {"include": [{"config": entry} for entry in entries]}
print(f"{name}={json.dumps(value, separators=(',', ':'))}", file=output)
PY
nightly-mi355x-2n:
needs: setup
if: |
always() && github.repository == 'sgl-project/sglang'
&& needs.setup.result == 'success'
&& needs.setup.outputs.matrix-2n != '{"include":[]}'
runs-on: linux-sglang-mi35x-di
strategy:
fail-fast: false
max-parallel: 2
matrix: ${{ fromJson(needs.setup.outputs.matrix-2n) }}
env: &mi355x-env
FRAMEWORK: sglang
HW: mi355x
MODEL: ${{ matrix.config.model }}
MODEL_PREFIX: ${{ matrix.config.model_prefix }}
PRECISION: ${{ matrix.config.precision }}
ISL: ${{ matrix.config.isl }}
OSL: ${{ matrix.config.osl }}
CONFIG_FILE: ${{ matrix.config.config_file }}
RESULT_FILENAME: mi355x-${{ matrix.config.name }}
MATRIX_CONFIG_NAME: ${{ matrix.config.name }}
SLURM_EXCLUDE: mia1-p01-g20,mia1-p02-g09
# NOTE: RUNNER_NAME is a built-in default env var on every runner, so the
# launch + cleanup steps read it directly. Do NOT set it here from
# ${{ runner.name }} -- the `runner` context is not available in job-level
# env and makes the whole workflow file invalid.
# Shared-NFS HuggingFace cache dir; the launcher resolves the live snapshot
# via refs/main. Per-model value comes from nightly-configs.yaml model_path.
MODEL_PATH: ${{ matrix.config.model_path }}
steps: &mi355x-steps
- name: Checkout code
uses: actions/checkout@v4
with:
fetch-depth: 0
persist-credentials: false
- name: Clean up prior Slurm jobs from this runner
continue-on-error: true
run: |
# launch_mi355x.sh names the allocation
# mi355x-ci-<RUNNER_NAME>-<GITHUB_RUN_ID>-<config>
# Here we clear THIS runner's leftovers from a crashed previous run, so
# we match the RUNNER_NAME prefix (older runs have a different run id).
# Never a blanket match on all jobs, which would kill a concurrent leg.
# %200j: squeue truncates the job name (%j) by default -- widen it so
# the grep sees the full name.
# No --me: Spur's squeue does not accept it. The RUNNER_NAME prefix is
# what makes this selective, and that is unique per runner, so the
# filter is equally precise without it.
if [ -z "${RUNNER_NAME:-}" ]; then echo "RUNNER_NAME unset; skipping"; exit 0; fi
STALE_JOBS=$(squeue --noheader --format="%i %200j" | grep -F "mi355x-ci-${RUNNER_NAME}-" | awk '{print $1}' || true)
if [ -n "$STALE_JOBS" ]; then
echo "Cancelling stale jobs for ${RUNNER_NAME}: $STALE_JOBS"
scancel $STALE_JOBS
fi
- name: Launch MI355X benchmark
timeout-minutes: 180
env:
# Manual dispatch input wins; otherwise use the latest image resolved
# by the setup job; otherwise the launcher falls back to the recipe default.
IMAGE_OVERRIDE: ${{ inputs.image != '' && inputs.image || needs.setup.outputs.image }}
run: bash scripts/ci/slurm/launch_mi355x.sh
- name: Pack logs
if: always()
run: |
LOG_DIR="$HOME/.mi355x_ci/${MATRIX_CONFIG_NAME}"
LOG_BUNDLE="${GITHUB_WORKSPACE}/${RESULT_FILENAME}_logs.tar.gz"
if [ ! -d "$LOG_DIR" ]; then
echo "WARN: log directory not found: $LOG_DIR"
exit 0
fi
mapfile -d '' LOG_FILES < <(
find "$LOG_DIR" -maxdepth 1 -type f \
\( -name '*.log' -o -name 'server_exit_*' -o -name 'bench_exit' \) \
-printf '%P\0' | sort -z
)
if [ "${#LOG_FILES[@]}" -eq 0 ]; then
echo "WARN: no log files found in $LOG_DIR"
exit 0
fi
STAGING_DIR="$(mktemp -d)"
trap 'rm -rf "$STAGING_DIR"' EXIT
for log_file in "${LOG_FILES[@]}"; do
cp -a "$LOG_DIR/$log_file" "$STAGING_DIR/$log_file" || true
done
tar czf "$LOG_BUNDLE" -C "$STAGING_DIR" .
echo "Packed ${#LOG_FILES[@]} log file(s) -> $LOG_BUNDLE"
- name: Process results
if: always()
run: |
pip install tabulate pyyaml -q
for result_file in ${{ github.workspace }}/${RESULT_FILENAME}_*.json; do
[ -f "$result_file" ] || continue
basename_file=$(basename "$result_file")
ctx=$(echo "$basename_file" | sed -n 's/.*_ctx_\([0-9]*\)_gen.*/\1/p')
gen=$(echo "$basename_file" | sed -n 's/.*_gen_\([0-9]*\)\.json/\1/p')
[ -n "$ctx" ] && [ -n "$gen" ] || continue
RESULT_FILENAME="${result_file%.json}" PREFILL_GPUS="$ctx" DECODE_GPUS="$gen" \
RECIPE_FILE="${{ github.workspace }}/$CONFIG_FILE" \
python3 scripts/ci/slurm/process_result.py
done
- name: Upload results
if: always()
uses: actions/upload-artifact@v4
with:
name: mi355x-${{ matrix.config.name }}-${{ github.run_id }}
path: |
${{ github.workspace }}/*.json
${{ github.workspace }}/*_logs.tar.gz
retention-days: 30
if-no-files-found: warn
- name: Clean up Slurm jobs on failure/cancel
if: failure() || cancelled()
continue-on-error: true
run: |
# Cancel only THIS leg's allocation -- match the full job name
# mi355x-ci-<RUNNER_NAME>-<GITHUB_RUN_ID>-<config>. GITHUB_RUN_ID +
# config are unique per matrix leg, so a concurrent leg is never hit.
if [ -z "${RUNNER_NAME:-}" ]; then echo "RUNNER_NAME unset; skipping"; exit 0; fi
JOB_TAG="mi355x-ci-${RUNNER_NAME}-${GITHUB_RUN_ID}-${MATRIX_CONFIG_NAME}"
ACTIVE_JOBS=$(squeue --noheader --format="%i %200j" | grep -F "$JOB_TAG" | awk '{print $1}' || true)
if [ -n "$ACTIVE_JOBS" ]; then
echo "Cancelling jobs for ${JOB_TAG}: $ACTIVE_JOBS"
scancel $ACTIVE_JOBS
fi
- name: Clean up MI355X scratch
if: always()
continue-on-error: true
run: |
LOG_DIR="$HOME/.mi355x_ci/${MATRIX_CONFIG_NAME}"
if [ -d "$LOG_DIR" ]; then
echo "Removing MI355X scratch directory: $LOG_DIR"
rm -rf "$LOG_DIR"
fi
nightly-mi355x-4n:
needs:
- setup
- nightly-mi355x-2n
if: |
false
&& !cancelled() && github.repository == 'sgl-project/sglang'
&& needs.setup.result == 'success'
&& needs.setup.outputs.matrix-4n != '{"include":[]}'
runs-on: linux-sglang-mi35x-di
strategy:
fail-fast: false
max-parallel: 1
matrix: ${{ fromJson(needs.setup.outputs.matrix-4n) }}
env: *mi355x-env
steps: *mi355x-steps
collect-results:
needs:
- nightly-mi355x-2n
- nightly-mi355x-4n
if: github.repository == 'sgl-project/sglang' && always()
runs-on: ubuntu-latest
steps:
- name: Checkout code
uses: actions/checkout@v4
- name: Download artifacts
uses: actions/download-artifact@v4
with:
path: results/
pattern: mi355x-*
- name: Print summary
run: |
pip install tabulate -q
python3 scripts/ci/slurm/summarize.py results/ >> $GITHUB_STEP_SUMMARY