mirror of
https://github.com/kvcache-ai/ktransformers.git
synced 2026-10-18 14:43:29 +08:00
yyj
7c35863f62
* ci: add isolated model E2E acceptance for PR wheels and PyPI * ci: recheck GPU occupancy before each model starts * ci: disable inherited container healthchecks during acceptance * ci: wait for free GPUs before native import validation * docs: record qj5090 acceptance infrastructure validation * ci: add immutable candidate and public PyPI release acceptance gates
74 lines
2.5 KiB
YAML
74 lines
2.5 KiB
YAML
name: Model E2E acceptance (release gate)
|
|
|
|
# Reuses the PR/PyPI harness; cannot publish. Called twice by the release workflow.
|
|
on:
|
|
workflow_call:
|
|
inputs:
|
|
stage:
|
|
type: string
|
|
required: true
|
|
artifact:
|
|
type: string
|
|
required: true
|
|
manifest_sha256:
|
|
type: string
|
|
required: true
|
|
build_attempt:
|
|
type: string
|
|
required: true
|
|
outputs:
|
|
evidence:
|
|
value: ${{ jobs.gpu.outputs.evidence }}
|
|
|
|
permissions:
|
|
contents: read
|
|
|
|
jobs:
|
|
gpu:
|
|
if: github.repository == 'kvcache-ai/ktransformers' && github.ref == 'refs/heads/main'
|
|
environment: qj5090-model-e2e
|
|
runs-on: [self-hosted, linux, x64, gpu, kt-model-e2e]
|
|
timeout-minutes: 1100
|
|
outputs:
|
|
evidence: ${{ steps.names.outputs.evidence }}
|
|
steps:
|
|
- uses: actions/checkout@v4
|
|
with:
|
|
persist-credentials: false
|
|
path: harness
|
|
- uses: actions/download-artifact@v4
|
|
with:
|
|
name: ${{ inputs.artifact }}
|
|
path: input/release
|
|
- name: Check caller, run identity and exact wheelhouse
|
|
env:
|
|
ENABLED: ${{ vars.KT_MODEL_E2E_ENABLED }}
|
|
STAGE: ${{ inputs.stage }}
|
|
MANIFEST_SHA256: ${{ inputs.manifest_sha256 }}
|
|
BUILD_ATTEMPT: ${{ inputs.build_attempt }}
|
|
CONTROLLER_PYTHON: ${{ vars.KT_MODEL_E2E_PYTHON || 'python3.12' }}
|
|
run: '"$CONTROLLER_PYTHON" harness/.github/model-e2e/release_prepare.py'
|
|
- name: Name stage evidence
|
|
id: names
|
|
env:
|
|
STAGE: ${{ inputs.stage }}
|
|
run: echo "evidence=release-e2e-$STAGE-${GITHUB_RUN_ID}-${GITHUB_RUN_ATTEMPT}" >> "$GITHUB_OUTPUT"
|
|
- name: Queue, install into clean environments, and run all three models
|
|
env:
|
|
HOST_CONFIG: ${{ vars.KT_MODEL_E2E_HOST_CONFIG || '/etc/ktransformers/model-e2e/qj5090.json' }}
|
|
CONTROLLER_PYTHON: ${{ vars.KT_MODEL_E2E_PYTHON || 'python3.12' }}
|
|
run: |
|
|
set -euo pipefail
|
|
test_root="$(mktemp -d "$RUNNER_TEMP/kt-release-e2e.XXXXXXXX")"
|
|
echo "E2E_OUTPUT=$test_root/output" >> "$GITHUB_ENV"
|
|
"$CONTROLLER_PYTHON" harness/.github/model-e2e/host.py \
|
|
--config "$HOST_CONFIG" --input "$GITHUB_WORKSPACE/input" \
|
|
--output "$test_root/output"
|
|
- uses: actions/upload-artifact@v4
|
|
if: always() && env.E2E_OUTPUT != ''
|
|
with:
|
|
name: ${{ steps.names.outputs.evidence }}
|
|
path: ${{ env.E2E_OUTPUT }}/evidence/
|
|
if-no-files-found: error
|
|
retention-days: 90
|