1725 lines
87 KiB
YAML
1725 lines
87 KiB
YAML
name: Nightly Test (AMD ROCm 7.2)
|
|
|
|
on:
|
|
schedule:
|
|
- cron: '30 17 * * *'
|
|
push:
|
|
branches:
|
|
- main
|
|
paths:
|
|
- "python/sglang/version.py"
|
|
workflow_dispatch:
|
|
inputs:
|
|
aiter_ref:
|
|
description: 'Override AITER commit (optional, leave empty to use Dockerfile default)'
|
|
required: false
|
|
type: string
|
|
default: ''
|
|
continue_on_error:
|
|
description: 'Continue on error (do not fail the workflow on test failures)'
|
|
required: false
|
|
type: boolean
|
|
default: true
|
|
job_select:
|
|
description: 'Select a job to run from dropdown (choose "all" to run all jobs)'
|
|
required: false
|
|
type: choice
|
|
default: 'all'
|
|
options:
|
|
- 'all'
|
|
- nightly-test-1-gpu-unit-rocm720
|
|
- nightly-accuracy-2-gpu-rocm720
|
|
- nightly-accuracy-2-gpu-vlm-rocm720
|
|
- nightly-perf-2-gpu-text-rocm720
|
|
- nightly-perf-2-gpu-vlm-rocm720
|
|
- nightly-4-gpu-rocm720
|
|
- nightly-accuracy-8-gpu-rocm720
|
|
- nightly-8-gpu-grok1-int4-rocm720
|
|
- nightly-8-gpu-grok2-rocm720
|
|
- nightly-8-gpu-deepseek-v31-rocm720
|
|
- nightly-8-gpu-deepseek-v32-rocm720
|
|
- nightly-8-gpu-deepseek-v32-mtp-rocm720
|
|
- nightly-8-gpu-deepseek-v3-kv-fp8-rocm720
|
|
- nightly-8-gpu-kimi-k26-rocm720
|
|
- nightly-8-gpu-qwen3-235b-rocm720
|
|
- nightly-8-gpu-qwen35-rocm720
|
|
- nightly-8-gpu-glm51-rocm720
|
|
- nightly-8-gpu-minimax-m27-rocm720
|
|
- nightly-1-gpu-zimage-turbo-rocm720
|
|
- nightly-test-1-gpu-mi35x-rocm720
|
|
- nightly-accuracy-8-gpu-mi35x-rocm720
|
|
- nightly-8-gpu-mi35x-grok1-int4-rocm720
|
|
- nightly-8-gpu-mi35x-grok2-rocm720
|
|
- nightly-8-gpu-mi35x-deepseek-r1-mxfp4-rocm720
|
|
- nightly-8-gpu-mi35x-deepseek-r1-mxfp4-kv-fp8-rocm720
|
|
- nightly-8-gpu-mi35x-deepseek-r1-mxfp4-ar-fusion-rocm720
|
|
- nightly-accuracy-8-gpu-mi35x-deepseek-v32-rocm720
|
|
- nightly-accuracy-8-gpu-mi35x-deepseek-v32-mtp-rocm720
|
|
- nightly-perf-8-gpu-mi35x-deepseek-v32-basic-rocm720
|
|
- nightly-perf-8-gpu-mi35x-deepseek-v32-mtp-rocm720
|
|
- nightly-8-gpu-mi35x-deepseek-v4-flash-rocm720
|
|
- nightly-8-gpu-mi35x-deepseek-v4-pro-rocm720
|
|
- nightly-8-gpu-mi35x-kimi-k26-rocm720
|
|
- nightly-8-gpu-mi35x-qwen3-235b-mxfp4-rocm720
|
|
- nightly-8-gpu-mi35x-qwen35-rocm720
|
|
- nightly-8-gpu-mi35x-glm51-rocm720
|
|
- nightly-8-gpu-mi35x-glm5-mxfp4-rocm720
|
|
job_filter:
|
|
description: 'Or type comma-separated job names (overrides dropdown if non-empty)'
|
|
required: false
|
|
type: string
|
|
default: ''
|
|
workflow_call:
|
|
inputs:
|
|
ref:
|
|
description: 'Git ref (branch, tag, or SHA) to test. If not provided, uses the default branch.'
|
|
required: false
|
|
type: string
|
|
default: ''
|
|
aiter_ref:
|
|
description: 'Override AITER commit (optional, leave empty to use Dockerfile default)'
|
|
required: false
|
|
type: string
|
|
default: ''
|
|
job_filter:
|
|
description: 'Select which job to run (leave empty or "all" to run all jobs)'
|
|
required: false
|
|
type: string
|
|
default: 'all'
|
|
continue_on_error:
|
|
description: 'Continue on error (do not fail the workflow on test failures)'
|
|
required: false
|
|
type: boolean
|
|
default: true
|
|
|
|
env:
|
|
AITER_COMMIT_OVERRIDE: ${{ inputs.aiter_ref }}
|
|
DOCKERHUB_AMD_USERNAME: ${{ secrets.DOCKERHUB_AMD_USERNAME }}
|
|
DOCKERHUB_AMD_TOKEN: ${{ secrets.DOCKERHUB_AMD_TOKEN }}
|
|
|
|
concurrency:
|
|
# When called via workflow_call with ref set, use a unique group per caller run to avoid
|
|
# collisions with direct schedule/push triggers. We use inputs.ref (not github.event_name)
|
|
# to detect this, because github.event_name inherits from the caller in workflow_call.
|
|
# Manual dispatch runs also get unique groups so they never cancel each other.
|
|
group: nightly-test-amd-rocm720-${{ github.event_name == 'workflow_dispatch' && format('manual-{0}', github.run_id) || inputs.ref && format('caller-{0}', github.run_id) || github.ref }}
|
|
cancel-in-progress: ${{ !inputs.ref && github.event_name != 'workflow_call' && github.event_name != 'workflow_dispatch' }}
|
|
|
|
jobs:
|
|
# ============================================== MI30x ROCm 7.2 Unit Tests ==============================================
|
|
# 1-GPU Unit Tests - LoRA, debug utils, scheduler, etc. (MI30x ROCm 7.2)
|
|
nightly-test-1-gpu-unit-rocm720:
|
|
if: (github.repository == 'sgl-project/sglang' || github.event_name == 'pull_request') && (!(inputs.job_filter || inputs.job_select) || (inputs.job_filter || inputs.job_select) == 'all' || contains(format(',{0},', inputs.job_filter || inputs.job_select), ',nightly-test-1-gpu-unit-rocm720,'))
|
|
runs-on: linux-mi325-1gpu-sglang
|
|
steps:
|
|
- name: Checkout code
|
|
uses: actions/checkout@v4
|
|
with:
|
|
ref: ${{ inputs.ref || github.sha }}
|
|
|
|
- name: Ensure VRAM is clear
|
|
run: bash scripts/ci/amd/ensure_vram_clear.sh rocm
|
|
|
|
- name: Setup docker (ROCm 7.2)
|
|
run: |
|
|
touch github_summary.md
|
|
bash scripts/ci/amd/amd_ci_start_container.sh --rocm-version rocm720
|
|
env:
|
|
GITHUB_WORKSPACE: ${{ github.workspace }}
|
|
|
|
- name: Install dependencies
|
|
run: bash scripts/ci/amd/amd_ci_install_dependency.sh
|
|
- name: Nightly Unit Test ROCm 7.2 (1-GPU)
|
|
timeout-minutes: 90
|
|
run: |
|
|
bash scripts/ci/amd/amd_ci_exec.sh -w /sglang-checkout/test \
|
|
-e GITHUB_STEP_SUMMARY="/sglang-checkout/github_summary.md" \
|
|
python3 run_suite.py --hw amd --suite nightly-amd-1-gpu --nightly --timeout-per-file 900 ${{ (github.event_name == 'schedule' || inputs.continue_on_error) && '--continue-on-error' || '' }} || TEST_EXIT_CODE=$?
|
|
echo "$(<github_summary.md )" >> $GITHUB_STEP_SUMMARY || true
|
|
exit ${TEST_EXIT_CODE:-0}
|
|
|
|
# ============================================== MI30x ROCm 7.2 Accuracy Tests ==============================================
|
|
# 2-GPU Accuracy Tests - GSM8K eval (MI30x ROCm 7.2)
|
|
nightly-accuracy-2-gpu-rocm720:
|
|
if: (github.repository == 'sgl-project/sglang' || github.event_name == 'pull_request') && (!(inputs.job_filter || inputs.job_select) || (inputs.job_filter || inputs.job_select) == 'all' || contains(format(',{0},', inputs.job_filter || inputs.job_select), ',nightly-accuracy-2-gpu-rocm720,'))
|
|
runs-on: linux-mi325-2gpu-sglang
|
|
steps:
|
|
- name: Checkout code
|
|
uses: actions/checkout@v4
|
|
with:
|
|
ref: ${{ inputs.ref || github.sha }}
|
|
|
|
- name: Ensure VRAM is clear
|
|
run: bash scripts/ci/amd/ensure_vram_clear.sh rocm
|
|
|
|
- name: Setup docker (ROCm 7.2)
|
|
run: |
|
|
touch github_summary.md
|
|
bash scripts/ci/amd/amd_ci_start_container.sh --rocm-version rocm720
|
|
env:
|
|
GITHUB_WORKSPACE: ${{ github.workspace }}
|
|
|
|
- name: Install dependencies
|
|
run: bash scripts/ci/amd/amd_ci_install_dependency.sh
|
|
- name: Nightly Test ROCm 7.2 (2-GPU)
|
|
run: |
|
|
> github_summary.md # Clear summary file
|
|
bash scripts/ci/amd/amd_ci_exec.sh -w /sglang-checkout/test \
|
|
-e GITHUB_STEP_SUMMARY="/sglang-checkout/github_summary.md" \
|
|
python3 run_suite.py --hw amd --suite nightly-amd --nightly --timeout-per-file 7200 ${{ (github.event_name == 'schedule' || inputs.continue_on_error) && '--continue-on-error' || '' }} || TEST_EXIT_CODE=$?
|
|
echo "$(<github_summary.md )" >> $GITHUB_STEP_SUMMARY || true
|
|
exit ${TEST_EXIT_CODE:-0}
|
|
|
|
# 2-GPU VLM Accuracy Tests - Vision-Language Models MMMU evaluation (ROCm 7.2)
|
|
nightly-accuracy-2-gpu-vlm-rocm720:
|
|
if: (github.repository == 'sgl-project/sglang' || github.event_name == 'pull_request') && (!(inputs.job_filter || inputs.job_select) || (inputs.job_filter || inputs.job_select) == 'all' || contains(format(',{0},', inputs.job_filter || inputs.job_select), ',nightly-accuracy-2-gpu-vlm-rocm720,'))
|
|
runs-on: linux-mi325-2gpu-sglang
|
|
steps:
|
|
- name: Checkout code
|
|
uses: actions/checkout@v4
|
|
with:
|
|
ref: ${{ inputs.ref || github.sha }}
|
|
|
|
- name: Ensure VRAM is clear
|
|
run: bash scripts/ci/amd/ensure_vram_clear.sh rocm
|
|
|
|
- name: Setup docker (ROCm 7.2)
|
|
run: |
|
|
touch github_summary.md
|
|
bash scripts/ci/amd/amd_ci_start_container.sh --rocm-version rocm720
|
|
env:
|
|
GITHUB_WORKSPACE: ${{ github.workspace }}
|
|
|
|
- name: Install dependencies
|
|
run: bash scripts/ci/amd/amd_ci_install_dependency.sh
|
|
- name: Nightly Accuracy Test ROCm 7.2 (2-GPU VLM MMMU)
|
|
timeout-minutes: 180
|
|
run: |
|
|
> github_summary.md # Clear summary file
|
|
bash scripts/ci/amd/amd_ci_exec.sh -w /sglang-checkout/test \
|
|
-e GITHUB_STEP_SUMMARY="/sglang-checkout/github_summary.md" \
|
|
python3 run_suite.py --hw amd --suite nightly-amd-accuracy-2-gpu-vlm --nightly --timeout-per-file 7200 ${{ (github.event_name == 'schedule' || inputs.continue_on_error) && '--continue-on-error' || '' }} || TEST_EXIT_CODE=$?
|
|
echo "$(<github_summary.md )" >> $GITHUB_STEP_SUMMARY || true
|
|
exit ${TEST_EXIT_CODE:-0}
|
|
|
|
# 2-GPU Text Models Performance Tests (ROCm 7.2)
|
|
nightly-perf-2-gpu-text-rocm720:
|
|
if: (github.repository == 'sgl-project/sglang' || github.event_name == 'pull_request') && (!(inputs.job_filter || inputs.job_select) || (inputs.job_filter || inputs.job_select) == 'all' || contains(format(',{0},', inputs.job_filter || inputs.job_select), ',nightly-perf-2-gpu-text-rocm720,'))
|
|
runs-on: linux-mi325-2gpu-sglang
|
|
steps:
|
|
- name: Checkout code
|
|
uses: actions/checkout@v4
|
|
with:
|
|
ref: ${{ inputs.ref || github.sha }}
|
|
|
|
- name: Ensure VRAM is clear
|
|
run: bash scripts/ci/amd/ensure_vram_clear.sh rocm
|
|
|
|
- name: Setup docker (ROCm 7.2)
|
|
run: |
|
|
touch github_summary.md
|
|
bash scripts/ci/amd/amd_ci_start_container.sh --rocm-version rocm720
|
|
env:
|
|
GITHUB_WORKSPACE: ${{ github.workspace }}
|
|
|
|
- name: Install dependencies
|
|
run: bash scripts/ci/amd/amd_ci_install_dependency.sh
|
|
- name: Performance Test ROCm 7.2 (2-GPU Text Models)
|
|
timeout-minutes: 120
|
|
run: |
|
|
> github_summary.md # Clear summary file
|
|
bash scripts/ci/amd/amd_ci_exec.sh -w /sglang-checkout/test \
|
|
-e SGLANG_USE_AITER=1 \
|
|
-e GITHUB_STEP_SUMMARY="/sglang-checkout/github_summary.md" \
|
|
python3 run_suite.py --hw amd --suite nightly-amd-perf-text-2-gpu --nightly --timeout-per-file 3600 ${{ (github.event_name == 'schedule' || inputs.continue_on_error) && '--continue-on-error' || '' }} || TEST_EXIT_CODE=$?
|
|
echo "$(<github_summary.md )" >> $GITHUB_STEP_SUMMARY || true
|
|
exit ${TEST_EXIT_CODE:-0}
|
|
|
|
# 2-GPU VLM Performance Tests (ROCm 7.2)
|
|
nightly-perf-2-gpu-vlm-rocm720:
|
|
if: (github.repository == 'sgl-project/sglang' || github.event_name == 'pull_request') && (!(inputs.job_filter || inputs.job_select) || (inputs.job_filter || inputs.job_select) == 'all' || contains(format(',{0},', inputs.job_filter || inputs.job_select), ',nightly-perf-2-gpu-vlm-rocm720,'))
|
|
runs-on: linux-mi325-2gpu-sglang
|
|
steps:
|
|
- name: Checkout code
|
|
uses: actions/checkout@v4
|
|
with:
|
|
ref: ${{ inputs.ref || github.sha }}
|
|
|
|
- name: Ensure VRAM is clear
|
|
run: bash scripts/ci/amd/ensure_vram_clear.sh rocm
|
|
|
|
- name: Setup docker (ROCm 7.2)
|
|
run: |
|
|
touch github_summary.md
|
|
bash scripts/ci/amd/amd_ci_start_container.sh --rocm-version rocm720
|
|
env:
|
|
GITHUB_WORKSPACE: ${{ github.workspace }}
|
|
|
|
- name: Install dependencies
|
|
run: bash scripts/ci/amd/amd_ci_install_dependency.sh
|
|
- name: Performance Test ROCm 7.2 (2-GPU VLM Models)
|
|
timeout-minutes: 180
|
|
run: |
|
|
> github_summary.md # Clear summary file
|
|
bash scripts/ci/amd/amd_ci_exec.sh -w /sglang-checkout/test \
|
|
-e SGLANG_USE_AITER=1 \
|
|
-e GITHUB_STEP_SUMMARY="/sglang-checkout/github_summary.md" \
|
|
python3 run_suite.py --hw amd --suite nightly-amd-perf-vlm-2-gpu --nightly --timeout-per-file 7200 ${{ (github.event_name == 'schedule' || inputs.continue_on_error) && '--continue-on-error' || '' }} || TEST_EXIT_CODE=$?
|
|
echo "$(<github_summary.md )" >> $GITHUB_STEP_SUMMARY || true
|
|
exit ${TEST_EXIT_CODE:-0}
|
|
|
|
# ============================================== MI30x ROCm 7.2 4-GPU Tests ==============================================
|
|
# 4-GPU Nightly Tests - Dumper/Comparator E2E, VLM Encoder DP (ROCm 7.2)
|
|
nightly-4-gpu-rocm720:
|
|
if: (github.repository == 'sgl-project/sglang' || github.event_name == 'pull_request') && (!(inputs.job_filter || inputs.job_select) || (inputs.job_filter || inputs.job_select) == 'all' || contains(format(',{0},', inputs.job_filter || inputs.job_select), ',nightly-4-gpu-rocm720,'))
|
|
runs-on: linux-mi325-4gpu-sglang
|
|
steps:
|
|
- name: Checkout code
|
|
uses: actions/checkout@v4
|
|
with:
|
|
ref: ${{ inputs.ref || github.sha }}
|
|
|
|
- name: Ensure VRAM is clear
|
|
run: bash scripts/ci/amd/ensure_vram_clear.sh rocm
|
|
|
|
- name: Setup docker (ROCm 7.2)
|
|
run: |
|
|
touch github_summary.md
|
|
bash scripts/ci/amd/amd_ci_start_container.sh --rocm-version rocm720
|
|
env:
|
|
GITHUB_WORKSPACE: ${{ github.workspace }}
|
|
|
|
- name: Install dependencies
|
|
run: bash scripts/ci/amd/amd_ci_install_dependency.sh
|
|
|
|
- name: Nightly Test ROCm 7.2 (4-GPU)
|
|
timeout-minutes: 120
|
|
run: |
|
|
> github_summary.md
|
|
bash scripts/ci/amd/amd_ci_exec.sh -w /sglang-checkout/test \
|
|
-e GITHUB_STEP_SUMMARY="/sglang-checkout/github_summary.md" \
|
|
python3 run_suite.py --hw amd --suite nightly-amd-4-gpu --nightly --timeout-per-file 3600 ${{ (github.event_name == 'schedule' || inputs.continue_on_error) && '--continue-on-error' || '' }} || TEST_EXIT_CODE=$?
|
|
echo "$(<github_summary.md )" >> $GITHUB_STEP_SUMMARY || true
|
|
exit ${TEST_EXIT_CODE:-0}
|
|
|
|
# 8-GPU Accuracy Tests - GPT-OSS, Grok1-FP8 (ROCm 7.2)
|
|
nightly-accuracy-8-gpu-rocm720:
|
|
if: (github.repository == 'sgl-project/sglang' || github.event_name == 'pull_request') && (!(inputs.job_filter || inputs.job_select) || (inputs.job_filter || inputs.job_select) == 'all' || contains(format(',{0},', inputs.job_filter || inputs.job_select), ',nightly-accuracy-8-gpu-rocm720,'))
|
|
runs-on: linux-mi325-8gpu-sglang
|
|
steps:
|
|
- name: Checkout code
|
|
uses: actions/checkout@v4
|
|
with:
|
|
ref: ${{ inputs.ref || github.sha }}
|
|
|
|
- name: Ensure VRAM is clear
|
|
run: bash scripts/ci/amd/ensure_vram_clear.sh rocm
|
|
|
|
- name: Setup docker (ROCm 7.2)
|
|
run: |
|
|
touch github_summary.md
|
|
bash scripts/ci/amd/amd_ci_start_container.sh --rocm-version rocm720
|
|
env:
|
|
GITHUB_WORKSPACE: ${{ github.workspace }}
|
|
|
|
- name: Install dependencies
|
|
run: bash scripts/ci/amd/amd_ci_install_dependency.sh --skip-test-time-deps
|
|
|
|
- name: Accuracy Test ROCm 7.2 (8-GPU GPT-OSS)
|
|
timeout-minutes: 180
|
|
run: |
|
|
bash scripts/ci/amd/amd_ci_exec.sh -w /sglang-checkout/test \
|
|
-e GITHUB_STEP_SUMMARY="/sglang-checkout/github_summary.md" \
|
|
python3 run_suite.py --hw amd --suite nightly-amd-accuracy-8-gpu-gpt-oss --nightly --timeout-per-file 7200 ${{ (github.event_name == 'schedule' || inputs.continue_on_error) && '--continue-on-error' || '' }} || TEST_EXIT_CODE=$?
|
|
echo "$(<github_summary.md )" >> $GITHUB_STEP_SUMMARY || true
|
|
exit ${TEST_EXIT_CODE:-0}
|
|
|
|
- name: Accuracy Test ROCm 7.2 (8-GPU Grok1-FP8)
|
|
timeout-minutes: 60
|
|
run: |
|
|
bash scripts/ci/amd/amd_ci_exec.sh -w /sglang-checkout/test \
|
|
-e RCCL_MSCCL_ENABLE=0 \
|
|
-e GITHUB_STEP_SUMMARY="/sglang-checkout/github_summary.md" \
|
|
python3 run_suite.py --hw amd --suite nightly-amd-accuracy-8-gpu-grok1-fp8 --nightly --timeout-per-file 3600 ${{ (github.event_name == 'schedule' || inputs.continue_on_error) && '--continue-on-error' || '' }} || TEST_EXIT_CODE=$?
|
|
echo "$(<github_summary.md )" >> $GITHUB_STEP_SUMMARY || true
|
|
exit ${TEST_EXIT_CODE:-0}
|
|
|
|
# ============================================== MI30x ROCm 7.2 Combined Accuracy + Performance Tests ==============================================
|
|
# 8-GPU Grok1-INT4 (Accuracy + Performance) ROCm 7.2
|
|
nightly-8-gpu-grok1-int4-rocm720:
|
|
if: (github.repository == 'sgl-project/sglang' || github.event_name == 'pull_request') && (!(inputs.job_filter || inputs.job_select) || (inputs.job_filter || inputs.job_select) == 'all' || contains(format(',{0},', inputs.job_filter || inputs.job_select), ',nightly-8-gpu-grok1-int4-rocm720,'))
|
|
runs-on: linux-mi325-8gpu-sglang
|
|
steps:
|
|
- name: Checkout code
|
|
uses: actions/checkout@v4
|
|
with:
|
|
ref: ${{ inputs.ref || github.sha }}
|
|
|
|
- name: Ensure VRAM is clear
|
|
run: bash scripts/ci/amd/ensure_vram_clear.sh rocm
|
|
|
|
- name: Setup docker (ROCm 7.2)
|
|
run: |
|
|
touch github_summary.md
|
|
bash scripts/ci/amd/amd_ci_start_container.sh --rocm-version rocm720
|
|
env:
|
|
GITHUB_WORKSPACE: ${{ github.workspace }}
|
|
|
|
- name: Install dependencies
|
|
run: bash scripts/ci/amd/amd_ci_install_dependency.sh --skip-test-time-deps
|
|
|
|
- name: Accuracy Test ROCm 7.2 (8-GPU Grok1-INT4)
|
|
timeout-minutes: 60
|
|
run: |
|
|
> github_summary.md # Clear summary file
|
|
bash scripts/ci/amd/amd_ci_exec.sh -w /sglang-checkout/test \
|
|
-e RCCL_MSCCL_ENABLE=0 \
|
|
-e GITHUB_STEP_SUMMARY="/sglang-checkout/github_summary.md" \
|
|
python3 run_suite.py --hw amd --suite nightly-amd-accuracy-8-gpu-grok1-int4 --nightly --timeout-per-file 3600 ${{ (github.event_name == 'schedule' || inputs.continue_on_error) && '--continue-on-error' || '' }} || TEST_EXIT_CODE=$?
|
|
echo "$(<github_summary.md )" >> $GITHUB_STEP_SUMMARY || true
|
|
exit ${TEST_EXIT_CODE:-0}
|
|
|
|
- name: Performance Test ROCm 7.2 (8-GPU Grok1-INT4)
|
|
timeout-minutes: 60
|
|
continue-on-error: true
|
|
run: |
|
|
> github_summary.md # Clear summary file
|
|
bash scripts/ci/amd/amd_ci_exec.sh -w /sglang-checkout/test \
|
|
-e RCCL_MSCCL_ENABLE=0 \
|
|
-e GITHUB_STEP_SUMMARY="/sglang-checkout/github_summary.md" \
|
|
python3 run_suite.py --hw amd --suite nightly-perf-8-gpu-grok1-int4 --nightly --timeout-per-file 3600 ${{ (github.event_name == 'schedule' || inputs.continue_on_error) && '--continue-on-error' || '' }} || TEST_EXIT_CODE=$?
|
|
echo "$(<github_summary.md )" >> $GITHUB_STEP_SUMMARY || true
|
|
exit ${TEST_EXIT_CODE:-0}
|
|
|
|
# 8-GPU Grok2 (Accuracy + Performance) ROCm 7.2
|
|
nightly-8-gpu-grok2-rocm720:
|
|
if: (github.repository == 'sgl-project/sglang' || github.event_name == 'pull_request') && (!(inputs.job_filter || inputs.job_select) || (inputs.job_filter || inputs.job_select) == 'all' || contains(format(',{0},', inputs.job_filter || inputs.job_select), ',nightly-8-gpu-grok2-rocm720,'))
|
|
runs-on: linux-mi325-8gpu-sglang
|
|
steps:
|
|
- name: Checkout code
|
|
uses: actions/checkout@v4
|
|
with:
|
|
ref: ${{ inputs.ref || github.sha }}
|
|
|
|
- name: Ensure VRAM is clear
|
|
run: bash scripts/ci/amd/ensure_vram_clear.sh rocm
|
|
|
|
- name: Setup docker (ROCm 7.2)
|
|
run: |
|
|
touch github_summary.md
|
|
bash scripts/ci/amd/amd_ci_start_container.sh --rocm-version rocm720
|
|
env:
|
|
GITHUB_WORKSPACE: ${{ github.workspace }}
|
|
|
|
- name: Install dependencies
|
|
run: bash scripts/ci/amd/amd_ci_install_dependency.sh --skip-test-time-deps
|
|
|
|
- name: Accuracy Test ROCm 7.2 (8-GPU Grok2)
|
|
timeout-minutes: 60
|
|
run: |
|
|
> github_summary.md # Clear summary file
|
|
bash scripts/ci/amd/amd_ci_exec.sh -w /sglang-checkout/test \
|
|
-e RCCL_MSCCL_ENABLE=0 \
|
|
-e GITHUB_STEP_SUMMARY="/sglang-checkout/github_summary.md" \
|
|
python3 run_suite.py --hw amd --suite nightly-amd-accuracy-8-gpu-grok2 --nightly --timeout-per-file 3600 ${{ (github.event_name == 'schedule' || inputs.continue_on_error) && '--continue-on-error' || '' }} || TEST_EXIT_CODE=$?
|
|
echo "$(<github_summary.md )" >> $GITHUB_STEP_SUMMARY || true
|
|
exit ${TEST_EXIT_CODE:-0}
|
|
|
|
- name: Performance Test ROCm 7.2 (8-GPU Grok2)
|
|
timeout-minutes: 60
|
|
continue-on-error: true
|
|
run: |
|
|
> github_summary.md # Clear summary file
|
|
bash scripts/ci/amd/amd_ci_exec.sh -w /sglang-checkout/test \
|
|
-e RCCL_MSCCL_ENABLE=0 \
|
|
-e GITHUB_STEP_SUMMARY="/sglang-checkout/github_summary.md" \
|
|
python3 run_suite.py --hw amd --suite nightly-perf-8-gpu-grok2 --nightly --timeout-per-file 3600 ${{ (github.event_name == 'schedule' || inputs.continue_on_error) && '--continue-on-error' || '' }} || TEST_EXIT_CODE=$?
|
|
echo "$(<github_summary.md )" >> $GITHUB_STEP_SUMMARY || true
|
|
exit ${TEST_EXIT_CODE:-0}
|
|
|
|
# 8-GPU DeepSeek-V3.1 (Accuracy + Performance) ROCm 7.2
|
|
nightly-8-gpu-deepseek-v31-rocm720:
|
|
if: (github.repository == 'sgl-project/sglang' || github.event_name == 'pull_request') && (!(inputs.job_filter || inputs.job_select) || (inputs.job_filter || inputs.job_select) == 'all' || contains(format(',{0},', inputs.job_filter || inputs.job_select), ',nightly-8-gpu-deepseek-v31-rocm720,'))
|
|
runs-on: linux-mi325-8gpu-sglang
|
|
steps:
|
|
- name: Checkout code
|
|
uses: actions/checkout@v4
|
|
with:
|
|
ref: ${{ inputs.ref || github.sha }}
|
|
|
|
- name: Ensure VRAM is clear
|
|
run: bash scripts/ci/amd/ensure_vram_clear.sh rocm
|
|
|
|
- name: Setup docker (ROCm 7.2)
|
|
run: |
|
|
touch github_summary.md
|
|
bash scripts/ci/amd/amd_ci_start_container.sh --rocm-version rocm720
|
|
env:
|
|
GITHUB_WORKSPACE: ${{ github.workspace }}
|
|
|
|
- name: Install dependencies
|
|
run: bash scripts/ci/amd/amd_ci_install_dependency.sh --skip-test-time-deps
|
|
|
|
- name: Accuracy Test ROCm 7.2 (8-GPU DeepSeek-V3.1)
|
|
timeout-minutes: 120
|
|
run: |
|
|
> github_summary.md # Clear summary file
|
|
bash scripts/ci/amd/amd_ci_exec.sh -w /sglang-checkout/test \
|
|
-e SGLANG_USE_AITER=1 \
|
|
-e GITHUB_STEP_SUMMARY="/sglang-checkout/github_summary.md" \
|
|
python3 run_suite.py --hw amd --suite nightly-amd-accuracy-8-gpu-deepseek-v31 --nightly --timeout-per-file 3600 ${{ (github.event_name == 'schedule' || inputs.continue_on_error) && '--continue-on-error' || '' }} || TEST_EXIT_CODE=$?
|
|
echo "$(<github_summary.md )" >> $GITHUB_STEP_SUMMARY || true
|
|
exit ${TEST_EXIT_CODE:-0}
|
|
|
|
- name: Performance Test ROCm 7.2 (8-GPU DeepSeek-V3.1)
|
|
timeout-minutes: 300
|
|
continue-on-error: true
|
|
run: |
|
|
> github_summary.md # Clear summary file
|
|
bash scripts/ci/amd/amd_ci_exec.sh -w /sglang-checkout/test \
|
|
-e SGLANG_USE_ROCM700A=1 \
|
|
-e GITHUB_STEP_SUMMARY="/sglang-checkout/github_summary.md" \
|
|
python3 run_suite.py --hw amd --suite nightly-perf-8-gpu-deepseek-v31 --nightly --timeout-per-file 18000 ${{ (github.event_name == 'schedule' || inputs.continue_on_error) && '--continue-on-error' || '' }} || TEST_EXIT_CODE=$?
|
|
echo "$(<github_summary.md )" >> $GITHUB_STEP_SUMMARY || true
|
|
exit ${TEST_EXIT_CODE:-0}
|
|
|
|
# 8-GPU DeepSeek-V3.2 (Basic Accuracy + Perf) ROCm 7.2
|
|
nightly-8-gpu-deepseek-v32-rocm720:
|
|
if: (github.repository == 'sgl-project/sglang' || github.event_name == 'pull_request') && (!(inputs.job_filter || inputs.job_select) || (inputs.job_filter || inputs.job_select) == 'all' || contains(format(',{0},', inputs.job_filter || inputs.job_select), ',nightly-8-gpu-deepseek-v32-rocm720,'))
|
|
runs-on: linux-mi325-8gpu-sglang
|
|
steps:
|
|
- name: Checkout code
|
|
uses: actions/checkout@v4
|
|
with:
|
|
ref: ${{ inputs.ref || github.sha }}
|
|
|
|
- name: Ensure VRAM is clear
|
|
run: bash scripts/ci/amd/ensure_vram_clear.sh rocm
|
|
|
|
- name: Setup docker (ROCm 7.2)
|
|
run: |
|
|
touch github_summary.md
|
|
bash scripts/ci/amd/amd_ci_start_container.sh --rocm-version rocm720
|
|
env:
|
|
GITHUB_WORKSPACE: ${{ github.workspace }}
|
|
|
|
- name: Install dependencies
|
|
run: bash scripts/ci/amd/amd_ci_install_dependency.sh --skip-test-time-deps
|
|
|
|
- name: Accuracy Test ROCm 7.2 (8-GPU DeepSeek-V3.2 Basic)
|
|
timeout-minutes: 120
|
|
run: |
|
|
> github_summary.md # Clear summary file
|
|
bash scripts/ci/amd/amd_ci_exec.sh -w /sglang-checkout/test \
|
|
-e GITHUB_STEP_SUMMARY="/sglang-checkout/github_summary.md" \
|
|
python3 run_suite.py --hw amd --suite nightly-amd-accuracy-8-gpu-deepseek-v32 --nightly --timeout-per-file 3600 ${{ (github.event_name == 'schedule' || inputs.continue_on_error) && '--continue-on-error' || '' }} || TEST_EXIT_CODE=$?
|
|
echo "$(<github_summary.md )" >> $GITHUB_STEP_SUMMARY || true
|
|
exit ${TEST_EXIT_CODE:-0}
|
|
|
|
- name: Performance Test ROCm 7.2 (8-GPU DeepSeek-V3.2 Basic)
|
|
timeout-minutes: 150
|
|
continue-on-error: true
|
|
run: |
|
|
> github_summary.md # Clear summary file
|
|
bash scripts/ci/amd/amd_ci_exec.sh -w /sglang-checkout/test \
|
|
-e GITHUB_STEP_SUMMARY="/sglang-checkout/github_summary.md" \
|
|
python3 run_suite.py --hw amd --suite nightly-perf-8-gpu-deepseek-v32-basic --nightly --timeout-per-file 5400 ${{ (github.event_name == 'schedule' || inputs.continue_on_error) && '--continue-on-error' || '' }} || TEST_EXIT_CODE=$?
|
|
echo "$(<github_summary.md )" >> $GITHUB_STEP_SUMMARY || true
|
|
exit ${TEST_EXIT_CODE:-0}
|
|
|
|
# 8-GPU DeepSeek-V3.2 MTP (MTP Accuracy + Perf) ROCm 7.2
|
|
nightly-8-gpu-deepseek-v32-mtp-rocm720:
|
|
if: (github.repository == 'sgl-project/sglang' || github.event_name == 'pull_request') && (!(inputs.job_filter || inputs.job_select) || (inputs.job_filter || inputs.job_select) == 'all' || contains(format(',{0},', inputs.job_filter || inputs.job_select), ',nightly-8-gpu-deepseek-v32-mtp-rocm720,'))
|
|
runs-on: linux-mi325-8gpu-sglang
|
|
steps:
|
|
- name: Checkout code
|
|
uses: actions/checkout@v4
|
|
with:
|
|
ref: ${{ inputs.ref || github.sha }}
|
|
|
|
- name: Ensure VRAM is clear
|
|
run: bash scripts/ci/amd/ensure_vram_clear.sh rocm
|
|
|
|
- name: Setup docker (ROCm 7.2)
|
|
run: |
|
|
touch github_summary.md
|
|
bash scripts/ci/amd/amd_ci_start_container.sh --rocm-version rocm720
|
|
env:
|
|
GITHUB_WORKSPACE: ${{ github.workspace }}
|
|
|
|
- name: Install dependencies
|
|
run: bash scripts/ci/amd/amd_ci_install_dependency.sh --skip-test-time-deps
|
|
|
|
- name: Accuracy Test ROCm 7.2 (8-GPU DeepSeek-V3.2 MTP)
|
|
timeout-minutes: 120
|
|
run: |
|
|
> github_summary.md # Clear summary file
|
|
bash scripts/ci/amd/amd_ci_exec.sh -w /sglang-checkout/test \
|
|
-e GITHUB_STEP_SUMMARY="/sglang-checkout/github_summary.md" \
|
|
python3 run_suite.py --hw amd --suite nightly-amd-accuracy-8-gpu-deepseek-v32-mtp --nightly --timeout-per-file 3600 ${{ (github.event_name == 'schedule' || inputs.continue_on_error) && '--continue-on-error' || '' }} || TEST_EXIT_CODE=$?
|
|
echo "$(<github_summary.md )" >> $GITHUB_STEP_SUMMARY || true
|
|
exit ${TEST_EXIT_CODE:-0}
|
|
|
|
- name: Performance Test ROCm 7.2 (8-GPU DeepSeek-V3.2 MTP)
|
|
timeout-minutes: 180
|
|
continue-on-error: true
|
|
run: |
|
|
> github_summary.md # Clear summary file
|
|
bash scripts/ci/amd/amd_ci_exec.sh -w /sglang-checkout/test \
|
|
-e GITHUB_STEP_SUMMARY="/sglang-checkout/github_summary.md" \
|
|
python3 run_suite.py --hw amd --suite nightly-perf-8-gpu-deepseek-v32-mtp --nightly --timeout-per-file 7200 ${{ (github.event_name == 'schedule' || inputs.continue_on_error) && '--continue-on-error' || '' }} || TEST_EXIT_CODE=$?
|
|
echo "$(<github_summary.md )" >> $GITHUB_STEP_SUMMARY || true
|
|
exit ${TEST_EXIT_CODE:-0}
|
|
|
|
# 8-GPU DeepSeek-V3 KV FP8 (Basic + MTP with --kv-cache-dtype fp8_e4m3) ROCm 7.2
|
|
nightly-8-gpu-deepseek-v3-kv-fp8-rocm720:
|
|
if: (github.repository == 'sgl-project/sglang' || github.event_name == 'pull_request') && (!(inputs.job_filter || inputs.job_select) || (inputs.job_filter || inputs.job_select) == 'all' || contains(format(',{0},', inputs.job_filter || inputs.job_select), ',nightly-8-gpu-deepseek-v3-kv-fp8-rocm720,'))
|
|
runs-on: linux-mi325-8gpu-sglang
|
|
steps:
|
|
- name: Checkout code
|
|
uses: actions/checkout@v4
|
|
with:
|
|
ref: ${{ inputs.ref || github.sha }}
|
|
|
|
- name: Ensure VRAM is clear
|
|
run: bash scripts/ci/amd/ensure_vram_clear.sh rocm
|
|
|
|
- name: Setup docker (ROCm 7.2)
|
|
run: |
|
|
touch github_summary.md
|
|
bash scripts/ci/amd/amd_ci_start_container.sh --rocm-version rocm720
|
|
env:
|
|
GITHUB_WORKSPACE: ${{ github.workspace }}
|
|
|
|
- name: Install dependencies
|
|
run: bash scripts/ci/amd/amd_ci_install_dependency.sh --skip-test-time-deps
|
|
|
|
- name: DeepSeek-V3 KV FP8 Test ROCm 7.2 (8-GPU Basic + MTP)
|
|
timeout-minutes: 120
|
|
run: |
|
|
> github_summary.md # Clear summary file
|
|
bash scripts/ci/amd/amd_ci_exec.sh -w /sglang-checkout/test \
|
|
-e GITHUB_STEP_SUMMARY="/sglang-checkout/github_summary.md" \
|
|
python3 run_suite.py --hw amd --suite nightly-amd-8-gpu-deepseek-v3-kv-fp8 --nightly --timeout-per-file 3600 ${{ (github.event_name == 'schedule' || inputs.continue_on_error) && '--continue-on-error' || '' }} || TEST_EXIT_CODE=$?
|
|
echo "$(<github_summary.md )" >> $GITHUB_STEP_SUMMARY || true
|
|
exit ${TEST_EXIT_CODE:-0}
|
|
|
|
# 8-GPU Kimi-K2.6 (Accuracy) ROCm 7.2
|
|
nightly-8-gpu-kimi-k26-rocm720:
|
|
if: (github.repository == 'sgl-project/sglang' || github.event_name == 'pull_request') && (!(inputs.job_filter || inputs.job_select) || (inputs.job_filter || inputs.job_select) == 'all' || contains(format(',{0},', inputs.job_filter || inputs.job_select), ',nightly-8-gpu-kimi-k26-rocm720,'))
|
|
runs-on: linux-mi325-8gpu-sglang
|
|
steps:
|
|
- name: Checkout code
|
|
uses: actions/checkout@v4
|
|
with:
|
|
ref: ${{ inputs.ref || github.sha }}
|
|
|
|
- name: Ensure VRAM is clear
|
|
run: bash scripts/ci/amd/ensure_vram_clear.sh rocm
|
|
|
|
- name: Setup docker (ROCm 7.2)
|
|
run: |
|
|
touch github_summary.md
|
|
bash scripts/ci/amd/amd_ci_start_container.sh --rocm-version rocm720
|
|
env:
|
|
GITHUB_WORKSPACE: ${{ github.workspace }}
|
|
|
|
- name: Install dependencies
|
|
run: bash scripts/ci/amd/amd_ci_install_dependency.sh --skip-test-time-deps
|
|
|
|
- name: Accuracy Test ROCm 7.2 (8-GPU Kimi-K2.6)
|
|
timeout-minutes: 120
|
|
run: |
|
|
> github_summary.md # Clear summary file
|
|
bash scripts/ci/amd/amd_ci_exec.sh -w /sglang-checkout/test \
|
|
-e GITHUB_STEP_SUMMARY="/sglang-checkout/github_summary.md" \
|
|
python3 run_suite.py --hw amd --suite nightly-amd-accuracy-8-gpu-kimi-k26 --nightly --timeout-per-file 3600 ${{ (github.event_name == 'schedule' || inputs.continue_on_error) && '--continue-on-error' || '' }} || TEST_EXIT_CODE=$?
|
|
echo "$(<github_summary.md )" >> $GITHUB_STEP_SUMMARY || true
|
|
exit ${TEST_EXIT_CODE:-0}
|
|
|
|
# 8-GPU Qwen3-235B (Accuracy + Performance) ROCm 7.2
|
|
nightly-8-gpu-qwen3-235b-rocm720:
|
|
if: (github.repository == 'sgl-project/sglang' || github.event_name == 'pull_request') && (!(inputs.job_filter || inputs.job_select) || (inputs.job_filter || inputs.job_select) == 'all' || contains(format(',{0},', inputs.job_filter || inputs.job_select), ',nightly-8-gpu-qwen3-235b-rocm720,'))
|
|
runs-on: linux-mi325-8gpu-sglang
|
|
steps:
|
|
- name: Checkout code
|
|
uses: actions/checkout@v4
|
|
with:
|
|
ref: ${{ inputs.ref || github.sha }}
|
|
|
|
- name: Ensure VRAM is clear
|
|
run: bash scripts/ci/amd/ensure_vram_clear.sh rocm
|
|
|
|
- name: Setup docker (ROCm 7.2)
|
|
run: |
|
|
touch github_summary.md
|
|
bash scripts/ci/amd/amd_ci_start_container.sh --rocm-version rocm720
|
|
env:
|
|
GITHUB_WORKSPACE: ${{ github.workspace }}
|
|
|
|
- name: Install dependencies
|
|
run: bash scripts/ci/amd/amd_ci_install_dependency.sh --skip-test-time-deps
|
|
|
|
- name: Accuracy Test + Performance Test ROCm 7.2 (8-GPU Qwen3)
|
|
timeout-minutes: 120
|
|
run: |
|
|
> github_summary.md # Clear summary file
|
|
bash scripts/ci/amd/amd_ci_exec.sh -w /sglang-checkout/test \
|
|
-e GITHUB_STEP_SUMMARY="/sglang-checkout/github_summary.md" \
|
|
python3 run_suite.py --hw amd --suite nightly-8-gpu-qwen3-235b --nightly --timeout-per-file 3600 ${{ (github.event_name == 'schedule' || inputs.continue_on_error) && '--continue-on-error' || '' }} || TEST_EXIT_CODE=$?
|
|
echo "$(<github_summary.md )" >> $GITHUB_STEP_SUMMARY || true
|
|
exit ${TEST_EXIT_CODE:-0}
|
|
|
|
# 8-GPU Qwen 3.5 (Accuracy + Performance combined) ROCm 7.2
|
|
nightly-8-gpu-qwen35-rocm720:
|
|
if: (github.repository == 'sgl-project/sglang' || github.event_name == 'pull_request') && (!(inputs.job_filter || inputs.job_select) || (inputs.job_filter || inputs.job_select) == 'all' || contains(format(',{0},', inputs.job_filter || inputs.job_select), ',nightly-8-gpu-qwen35-rocm720,'))
|
|
runs-on: linux-mi325-8gpu-sglang
|
|
steps:
|
|
- name: Checkout code
|
|
uses: actions/checkout@v4
|
|
with:
|
|
ref: ${{ inputs.ref || github.sha }}
|
|
|
|
- name: Ensure VRAM is clear
|
|
run: bash scripts/ci/amd/ensure_vram_clear.sh rocm
|
|
|
|
- name: Setup docker (ROCm 7.2)
|
|
run: |
|
|
touch github_summary.md
|
|
bash scripts/ci/amd/amd_ci_start_container.sh --rocm-version rocm720
|
|
env:
|
|
GITHUB_WORKSPACE: ${{ github.workspace }}
|
|
|
|
- name: Install dependencies
|
|
run: |
|
|
bash scripts/ci/amd/amd_ci_install_dependency.sh --skip-aiter-build --skip-test-time-deps
|
|
bash scripts/ci/amd/amd_ci_exec.sh pip install mistral-common "lm-eval[api]"
|
|
|
|
- name: Accuracy Test ROCm 7.2 (8-GPU Qwen 3.5)
|
|
timeout-minutes: 120
|
|
run: |
|
|
> github_summary.md # Clear summary file
|
|
bash scripts/ci/amd/amd_ci_exec.sh -w /sglang-checkout/test \
|
|
-e GITHUB_STEP_SUMMARY="/sglang-checkout/github_summary.md" \
|
|
python3 run_suite.py --hw amd --suite nightly-amd-accuracy-8-gpu-qwen35 --nightly --timeout-per-file 3600 ${{ (github.event_name == 'schedule' || inputs.continue_on_error) && '--continue-on-error' || '' }} || TEST_EXIT_CODE=$?
|
|
echo "$(<github_summary.md )" >> $GITHUB_STEP_SUMMARY || true
|
|
exit ${TEST_EXIT_CODE:-0}
|
|
|
|
- name: Performance Test ROCm 7.2 (8-GPU Qwen 3.5 FP8)
|
|
timeout-minutes: 120
|
|
continue-on-error: true
|
|
run: |
|
|
> github_summary.md # Clear summary file
|
|
bash scripts/ci/amd/amd_ci_exec.sh -w /sglang-checkout/test \
|
|
-e SGLANG_USE_AITER=1 \
|
|
-e GITHUB_STEP_SUMMARY="/sglang-checkout/github_summary.md" \
|
|
python3 run_suite.py --hw amd --suite nightly-perf-8-gpu-qwen35-fp8 --nightly --timeout-per-file 5400 ${{ (github.event_name == 'schedule' || inputs.continue_on_error) && '--continue-on-error' || '' }} || TEST_EXIT_CODE=$?
|
|
echo "$(<github_summary.md )" >> $GITHUB_STEP_SUMMARY || true
|
|
exit ${TEST_EXIT_CODE:-0}
|
|
|
|
# 8-GPU GLM-5.1 (Accuracy + Performance combined) ROCm 7.2
|
|
nightly-8-gpu-glm51-rocm720:
|
|
if: (github.repository == 'sgl-project/sglang' || github.event_name == 'pull_request') && (!(inputs.job_filter || inputs.job_select) || (inputs.job_filter || inputs.job_select) == 'all' || contains(format(',{0},', inputs.job_filter || inputs.job_select), ',nightly-8-gpu-glm51-rocm720,'))
|
|
runs-on: linux-mi325-8gpu-sglang
|
|
steps:
|
|
- name: Checkout code
|
|
uses: actions/checkout@v4
|
|
with:
|
|
ref: ${{ inputs.ref || github.sha }}
|
|
|
|
- name: Ensure VRAM is clear
|
|
run: bash scripts/ci/amd/ensure_vram_clear.sh rocm
|
|
|
|
- name: Setup docker (ROCm 7.2)
|
|
run: |
|
|
touch github_summary.md
|
|
bash scripts/ci/amd/amd_ci_start_container.sh --rocm-version rocm720
|
|
env:
|
|
GITHUB_WORKSPACE: ${{ github.workspace }}
|
|
|
|
- name: Install dependencies
|
|
run: |
|
|
bash scripts/ci/amd/amd_ci_install_dependency.sh --skip-test-time-deps
|
|
bash scripts/ci/amd/amd_ci_exec.sh pip install git+https://github.com/huggingface/transformers.git@96f807a33b75
|
|
|
|
- name: Accuracy Test ROCm 7.2 (8-GPU GLM-5.1 NSA)
|
|
timeout-minutes: 120
|
|
run: |
|
|
> github_summary.md # Clear summary file
|
|
bash scripts/ci/amd/amd_ci_exec.sh -w /sglang-checkout/test \
|
|
-e GITHUB_STEP_SUMMARY="/sglang-checkout/github_summary.md" \
|
|
python3 run_suite.py --hw amd --suite nightly-amd-accuracy-8-gpu-glm51 --nightly --timeout-per-file 3600 ${{ (github.event_name == 'schedule' || inputs.continue_on_error) && '--continue-on-error' || '' }} || TEST_EXIT_CODE=$?
|
|
echo "$(<github_summary.md )" >> $GITHUB_STEP_SUMMARY || true
|
|
exit ${TEST_EXIT_CODE:-0}
|
|
|
|
- name: Performance Test ROCm 7.2 (8-GPU GLM-5.1)
|
|
timeout-minutes: 120
|
|
continue-on-error: true
|
|
run: |
|
|
> github_summary.md # Clear summary file
|
|
bash scripts/ci/amd/amd_ci_exec.sh -w /sglang-checkout/test \
|
|
-e SGLANG_USE_AITER=1 \
|
|
-e GITHUB_STEP_SUMMARY="/sglang-checkout/github_summary.md" \
|
|
python3 run_suite.py --hw amd --suite nightly-perf-8-gpu-glm51 --nightly --timeout-per-file 5400 ${{ (github.event_name == 'schedule' || inputs.continue_on_error) && '--continue-on-error' || '' }} || TEST_EXIT_CODE=$?
|
|
echo "$(<github_summary.md )" >> $GITHUB_STEP_SUMMARY || true
|
|
exit ${TEST_EXIT_CODE:-0}
|
|
|
|
# 8-GPU MiniMax-M2.7 (Accuracy + Performance combined, replaces M2.5) ROCm 7.2
|
|
nightly-8-gpu-minimax-m27-rocm720:
|
|
if: (github.repository == 'sgl-project/sglang' || github.event_name == 'pull_request') && (!(inputs.job_filter || inputs.job_select) || (inputs.job_filter || inputs.job_select) == 'all' || contains(format(',{0},', inputs.job_filter || inputs.job_select), ',nightly-8-gpu-minimax-m27-rocm720,'))
|
|
runs-on: linux-mi325-8gpu-sglang
|
|
steps:
|
|
- name: Checkout code
|
|
uses: actions/checkout@v4
|
|
with:
|
|
ref: ${{ inputs.ref || github.sha }}
|
|
|
|
- name: Ensure VRAM is clear
|
|
run: bash scripts/ci/amd/ensure_vram_clear.sh rocm
|
|
|
|
- name: Setup docker (ROCm 7.2)
|
|
run: |
|
|
touch github_summary.md
|
|
bash scripts/ci/amd/amd_ci_start_container.sh --rocm-version rocm720
|
|
env:
|
|
GITHUB_WORKSPACE: ${{ github.workspace }}
|
|
|
|
- name: Install dependencies
|
|
run: bash scripts/ci/amd/amd_ci_install_dependency.sh --skip-test-time-deps
|
|
|
|
- name: Accuracy Test ROCm 7.2 (8-GPU MiniMax-M2.7)
|
|
timeout-minutes: 120
|
|
run: |
|
|
> github_summary.md # Clear summary file
|
|
bash scripts/ci/amd/amd_ci_exec.sh -w /sglang-checkout/test \
|
|
-e SGLANG_USE_AITER=1 \
|
|
-e GITHUB_STEP_SUMMARY="/sglang-checkout/github_summary.md" \
|
|
python3 run_suite.py --hw amd --suite nightly-amd-accuracy-8-gpu-minimax-m27 --nightly --timeout-per-file 3600 ${{ (github.event_name == 'schedule' || inputs.continue_on_error) && '--continue-on-error' || '' }} || TEST_EXIT_CODE=$?
|
|
echo "$(<github_summary.md )" >> $GITHUB_STEP_SUMMARY || true
|
|
exit ${TEST_EXIT_CODE:-0}
|
|
|
|
- name: Performance Test ROCm 7.2 (8-GPU MiniMax-M2.7)
|
|
timeout-minutes: 120
|
|
continue-on-error: true # Perf test failure doesn't fail the job if accuracy passed
|
|
run: |
|
|
> github_summary.md # Clear summary file
|
|
bash scripts/ci/amd/amd_ci_exec.sh -w /sglang-checkout/test \
|
|
-e SGLANG_USE_AITER=1 \
|
|
-e GITHUB_STEP_SUMMARY="/sglang-checkout/github_summary.md" \
|
|
python3 run_suite.py --hw amd --suite nightly-perf-8-gpu-minimax-m27 --nightly --timeout-per-file 5400 ${{ (github.event_name == 'schedule' || inputs.continue_on_error) && '--continue-on-error' || '' }} || TEST_EXIT_CODE=$?
|
|
echo "$(<github_summary.md )" >> $GITHUB_STEP_SUMMARY || true
|
|
exit ${TEST_EXIT_CODE:-0}
|
|
|
|
# ============================================== MI30x ROCm 7.2 Diffusion Tests ==============================================
|
|
# 1-GPU Z-Image-Turbo (Diffusion T2I) ROCm 7.2
|
|
nightly-1-gpu-zimage-turbo-rocm720:
|
|
if: (github.repository == 'sgl-project/sglang' || github.event_name == 'pull_request') && (!(inputs.job_filter || inputs.job_select) || (inputs.job_filter || inputs.job_select) == 'all' || contains(format(',{0},', inputs.job_filter || inputs.job_select), ',nightly-1-gpu-zimage-turbo-rocm720,'))
|
|
runs-on: linux-mi325-1gpu-sglang
|
|
steps:
|
|
- name: Checkout code
|
|
uses: actions/checkout@v4
|
|
with:
|
|
ref: ${{ inputs.ref || github.sha }}
|
|
|
|
- name: Ensure VRAM is clear
|
|
run: bash scripts/ci/amd/ensure_vram_clear.sh rocm
|
|
|
|
- name: Setup docker (ROCm 7.2)
|
|
run: |
|
|
touch github_summary.md
|
|
bash scripts/ci/amd/amd_ci_start_container.sh --rocm-version rocm720
|
|
env:
|
|
GITHUB_WORKSPACE: ${{ github.workspace }}
|
|
|
|
- name: Install dependencies
|
|
run: bash scripts/ci/amd/amd_ci_install_dependency.sh
|
|
|
|
- name: Z-Image-Turbo Diffusion Test ROCm 7.2 (1-GPU)
|
|
timeout-minutes: 45
|
|
run: |
|
|
bash scripts/ci/amd/amd_ci_exec.sh -w /sglang-checkout \
|
|
-e GITHUB_STEP_SUMMARY="/sglang-checkout/github_summary.md" \
|
|
-e SGLANG_DIFFUSION_ARTIFACT_DIR="/sglang-checkout/diffusion-artifacts" \
|
|
pytest test/registered/amd/test_zimage_turbo.py -v -s ${{ inputs.continue_on_error && '|| true' || '' }} || TEST_EXIT_CODE=$?
|
|
echo "$(<github_summary.md )" >> $GITHUB_STEP_SUMMARY || true
|
|
exit ${TEST_EXIT_CODE:-0}
|
|
|
|
- name: Upload generated images
|
|
if: always()
|
|
uses: actions/upload-artifact@v4
|
|
with:
|
|
name: zimage-turbo-outputs-rocm720
|
|
path: diffusion-artifacts/
|
|
if-no-files-found: ignore
|
|
retention-days: 30
|
|
|
|
# ============================================== MI35x ROCm 7.2 Tests ==============================================
|
|
# MI35x 1-GPU ROCm 7.2 tests
|
|
nightly-test-1-gpu-mi35x-rocm720:
|
|
if: (github.repository == 'sgl-project/sglang' || github.event_name == 'pull_request') && (!(inputs.job_filter || inputs.job_select) || (inputs.job_filter || inputs.job_select) == 'all' || contains(format(',{0},', inputs.job_filter || inputs.job_select), ',nightly-test-1-gpu-mi35x-rocm720,'))
|
|
runs-on: linux-mi35x-gpu-1
|
|
steps:
|
|
- name: Checkout code
|
|
uses: actions/checkout@v4
|
|
with:
|
|
ref: ${{ inputs.ref || github.sha }}
|
|
|
|
- name: Ensure VRAM is clear
|
|
run: bash scripts/ci/amd/ensure_vram_clear.sh rocm
|
|
|
|
- name: Setup docker (ROCm 7.2)
|
|
run: |
|
|
touch github_summary.md
|
|
bash scripts/ci/amd/amd_ci_start_container.sh --rocm-version rocm720
|
|
env:
|
|
GITHUB_WORKSPACE: ${{ github.workspace }}
|
|
|
|
- name: Install dependencies
|
|
run: |
|
|
bash scripts/ci/amd/amd_ci_install_dependency.sh
|
|
- name: Nightly Test MI35x ROCm 7.2 (1-GPU)
|
|
timeout-minutes: 90
|
|
run: |
|
|
bash scripts/ci/amd/amd_ci_exec.sh -w /sglang-checkout/test \
|
|
-e GITHUB_STEP_SUMMARY="/sglang-checkout/github_summary.md" \
|
|
python3 run_suite.py --hw amd --suite nightly-amd-1-gpu-mi35x --nightly --timeout-per-file 900 ${{ (github.event_name == 'schedule' || inputs.continue_on_error) && '--continue-on-error' || '' }} || TEST_EXIT_CODE=$?
|
|
echo "$(<github_summary.md )" >> $GITHUB_STEP_SUMMARY || true
|
|
exit ${TEST_EXIT_CODE:-0}
|
|
|
|
# MI35x 8-GPU Accuracy Tests - GPT-OSS (ROCm 7.2)
|
|
nightly-accuracy-8-gpu-mi35x-rocm720:
|
|
if: (github.repository == 'sgl-project/sglang' || github.event_name == 'pull_request') && (!(inputs.job_filter || inputs.job_select) || (inputs.job_filter || inputs.job_select) == 'all' || contains(format(',{0},', inputs.job_filter || inputs.job_select), ',nightly-accuracy-8-gpu-mi35x-rocm720,'))
|
|
runs-on: linux-mi35x-gpu-8
|
|
steps:
|
|
- name: Checkout code
|
|
uses: actions/checkout@v4
|
|
with:
|
|
ref: ${{ inputs.ref || github.sha }}
|
|
|
|
- name: Ensure VRAM is clear
|
|
run: bash scripts/ci/amd/ensure_vram_clear.sh rocm
|
|
|
|
- name: Setup docker (ROCm 7.2)
|
|
run: |
|
|
touch github_summary.md
|
|
bash scripts/ci/amd/amd_ci_start_container.sh --rocm-version rocm720
|
|
env:
|
|
GITHUB_WORKSPACE: ${{ github.workspace }}
|
|
|
|
- name: Install dependencies
|
|
run: |
|
|
bash scripts/ci/amd/amd_ci_install_dependency.sh --skip-test-time-deps
|
|
# Install tabulate for run_suite.py (missing in MI35x container)
|
|
bash scripts/ci/amd/amd_ci_exec.sh pip install tabulate
|
|
|
|
- name: Accuracy Test MI35x ROCm 7.2 (8-GPU GPT-OSS)
|
|
timeout-minutes: 180
|
|
run: |
|
|
bash scripts/ci/amd/amd_ci_exec.sh -w /sglang-checkout/test \
|
|
-e GITHUB_STEP_SUMMARY="/sglang-checkout/github_summary.md" \
|
|
python3 run_suite.py --hw amd --suite nightly-amd-8-gpu-mi35x --nightly --timeout-per-file 7200 ${{ (github.event_name == 'schedule' || inputs.continue_on_error) && '--continue-on-error' || '' }} || TEST_EXIT_CODE=$?
|
|
echo "$(<github_summary.md )" >> $GITHUB_STEP_SUMMARY || true
|
|
exit ${TEST_EXIT_CODE:-0}
|
|
|
|
# MI35x 8-GPU Grok1-INT4 (Accuracy + Performance) ROCm 7.2
|
|
nightly-8-gpu-mi35x-grok1-int4-rocm720:
|
|
if: (github.repository == 'sgl-project/sglang' || github.event_name == 'pull_request') && (!(inputs.job_filter || inputs.job_select) || (inputs.job_filter || inputs.job_select) == 'all' || contains(format(',{0},', inputs.job_filter || inputs.job_select), ',nightly-8-gpu-mi35x-grok1-int4-rocm720,'))
|
|
runs-on: linux-mi35x-gpu-8
|
|
steps:
|
|
- name: Checkout code
|
|
uses: actions/checkout@v4
|
|
with:
|
|
ref: ${{ inputs.ref || github.sha }}
|
|
|
|
- name: Ensure VRAM is clear
|
|
run: bash scripts/ci/amd/ensure_vram_clear.sh rocm
|
|
|
|
- name: Setup docker (ROCm 7.2)
|
|
run: |
|
|
touch github_summary.md
|
|
bash scripts/ci/amd/amd_ci_start_container.sh --rocm-version rocm720
|
|
env:
|
|
GITHUB_WORKSPACE: ${{ github.workspace }}
|
|
|
|
- name: Install dependencies
|
|
run: |
|
|
bash scripts/ci/amd/amd_ci_install_dependency.sh --skip-test-time-deps
|
|
# Install tabulate for run_suite.py (missing in MI35x container)
|
|
bash scripts/ci/amd/amd_ci_exec.sh pip install tabulate
|
|
|
|
- name: Accuracy Test MI35x ROCm 7.2 (8-GPU Grok1-INT4)
|
|
timeout-minutes: 60
|
|
run: |
|
|
> github_summary.md # Clear summary file
|
|
bash scripts/ci/amd/amd_ci_exec.sh -w /sglang-checkout/test \
|
|
-e RCCL_MSCCL_ENABLE=0 \
|
|
-e GITHUB_STEP_SUMMARY="/sglang-checkout/github_summary.md" \
|
|
python3 run_suite.py --hw amd --suite nightly-amd-accuracy-8-gpu-mi35x-grok1-int4 --nightly --timeout-per-file 3600 ${{ (github.event_name == 'schedule' || inputs.continue_on_error) && '--continue-on-error' || '' }} || TEST_EXIT_CODE=$?
|
|
echo "$(<github_summary.md )" >> $GITHUB_STEP_SUMMARY || true
|
|
exit ${TEST_EXIT_CODE:-0}
|
|
|
|
- name: Performance Test MI35x ROCm 7.2 (8-GPU Grok1-INT4)
|
|
timeout-minutes: 60
|
|
continue-on-error: true
|
|
run: |
|
|
> github_summary.md # Clear summary file
|
|
bash scripts/ci/amd/amd_ci_exec.sh -w /sglang-checkout/test \
|
|
-e RCCL_MSCCL_ENABLE=0 \
|
|
-e GITHUB_STEP_SUMMARY="/sglang-checkout/github_summary.md" \
|
|
python3 run_suite.py --hw amd --suite nightly-perf-8-gpu-mi35x-grok1-int4 --nightly --timeout-per-file 3600 ${{ (github.event_name == 'schedule' || inputs.continue_on_error) && '--continue-on-error' || '' }} || TEST_EXIT_CODE=$?
|
|
echo "$(<github_summary.md )" >> $GITHUB_STEP_SUMMARY || true
|
|
exit ${TEST_EXIT_CODE:-0}
|
|
|
|
# MI35x 8-GPU Grok2 (Accuracy + Performance) ROCm 7.2
|
|
nightly-8-gpu-mi35x-grok2-rocm720:
|
|
if: (github.repository == 'sgl-project/sglang' || github.event_name == 'pull_request') && (!(inputs.job_filter || inputs.job_select) || (inputs.job_filter || inputs.job_select) == 'all' || contains(format(',{0},', inputs.job_filter || inputs.job_select), ',nightly-8-gpu-mi35x-grok2-rocm720,'))
|
|
runs-on: linux-mi35x-gpu-8
|
|
steps:
|
|
- name: Checkout code
|
|
uses: actions/checkout@v4
|
|
with:
|
|
ref: ${{ inputs.ref || github.sha }}
|
|
|
|
- name: Ensure VRAM is clear
|
|
run: bash scripts/ci/amd/ensure_vram_clear.sh rocm
|
|
|
|
- name: Setup docker (ROCm 7.2)
|
|
run: |
|
|
touch github_summary.md
|
|
bash scripts/ci/amd/amd_ci_start_container.sh --rocm-version rocm720
|
|
env:
|
|
GITHUB_WORKSPACE: ${{ github.workspace }}
|
|
|
|
- name: Install dependencies
|
|
run: |
|
|
bash scripts/ci/amd/amd_ci_install_dependency.sh --skip-test-time-deps
|
|
# Install tabulate for run_suite.py (missing in MI35x container)
|
|
bash scripts/ci/amd/amd_ci_exec.sh pip install tabulate
|
|
|
|
- name: Accuracy Test MI35x ROCm 7.2 (8-GPU Grok2)
|
|
timeout-minutes: 60
|
|
run: |
|
|
> github_summary.md # Clear summary file
|
|
bash scripts/ci/amd/amd_ci_exec.sh -w /sglang-checkout/test \
|
|
-e RCCL_MSCCL_ENABLE=0 \
|
|
-e GITHUB_STEP_SUMMARY="/sglang-checkout/github_summary.md" \
|
|
python3 run_suite.py --hw amd --suite nightly-amd-accuracy-8-gpu-mi35x-grok2 --nightly --timeout-per-file 3600 ${{ (github.event_name == 'schedule' || inputs.continue_on_error) && '--continue-on-error' || '' }} || TEST_EXIT_CODE=$?
|
|
echo "$(<github_summary.md )" >> $GITHUB_STEP_SUMMARY || true
|
|
exit ${TEST_EXIT_CODE:-0}
|
|
|
|
- name: Performance Test MI35x ROCm 7.2 (8-GPU Grok2)
|
|
timeout-minutes: 60
|
|
continue-on-error: true
|
|
run: |
|
|
> github_summary.md # Clear summary file
|
|
bash scripts/ci/amd/amd_ci_exec.sh -w /sglang-checkout/test \
|
|
-e RCCL_MSCCL_ENABLE=0 \
|
|
-e GITHUB_STEP_SUMMARY="/sglang-checkout/github_summary.md" \
|
|
python3 run_suite.py --hw amd --suite nightly-perf-8-gpu-mi35x-grok2 --nightly --timeout-per-file 3600 ${{ (github.event_name == 'schedule' || inputs.continue_on_error) && '--continue-on-error' || '' }} || TEST_EXIT_CODE=$?
|
|
echo "$(<github_summary.md )" >> $GITHUB_STEP_SUMMARY || true
|
|
exit ${TEST_EXIT_CODE:-0}
|
|
|
|
# MI35x 8-GPU DeepSeek-R1-MXFP4 (Accuracy + Performance) ROCm 7.2
|
|
nightly-8-gpu-mi35x-deepseek-r1-mxfp4-rocm720:
|
|
if: (github.repository == 'sgl-project/sglang' || github.event_name == 'pull_request') && (!(inputs.job_filter || inputs.job_select) || (inputs.job_filter || inputs.job_select) == 'all' || contains(format(',{0},', inputs.job_filter || inputs.job_select), ',nightly-8-gpu-mi35x-deepseek-r1-mxfp4-rocm720,'))
|
|
runs-on: linux-mi35x-gpu-8
|
|
steps:
|
|
- name: Checkout code
|
|
uses: actions/checkout@v4
|
|
with:
|
|
ref: ${{ inputs.ref || github.sha }}
|
|
|
|
- name: Ensure VRAM is clear
|
|
run: bash scripts/ci/amd/ensure_vram_clear.sh rocm
|
|
|
|
- name: Setup docker (ROCm 7.2)
|
|
run: |
|
|
touch github_summary.md
|
|
bash scripts/ci/amd/amd_ci_start_container.sh --rocm-version rocm720
|
|
env:
|
|
GITHUB_WORKSPACE: ${{ github.workspace }}
|
|
|
|
- name: Install dependencies
|
|
run: |
|
|
bash scripts/ci/amd/amd_ci_install_dependency.sh --skip-test-time-deps
|
|
# Install tabulate for run_suite.py (missing in MI35x container)
|
|
bash scripts/ci/amd/amd_ci_exec.sh pip install tabulate
|
|
|
|
- name: Accuracy Test MI35x ROCm 7.2 (8-GPU DeepSeek-R1-MXFP4)
|
|
timeout-minutes: 180
|
|
run: |
|
|
> github_summary.md # Clear summary file
|
|
bash scripts/ci/amd/amd_ci_exec.sh -w /sglang-checkout/test \
|
|
-e GITHUB_STEP_SUMMARY="/sglang-checkout/github_summary.md" \
|
|
python3 run_suite.py --hw amd --suite nightly-amd-8-gpu-mi35x-deepseek-r1-mxfp4 --nightly --timeout-per-file 7200 ${{ (github.event_name == 'schedule' || inputs.continue_on_error) && '--continue-on-error' || '' }} || TEST_EXIT_CODE=$?
|
|
echo "$(<github_summary.md )" >> $GITHUB_STEP_SUMMARY || true
|
|
exit ${TEST_EXIT_CODE:-0}
|
|
|
|
- name: Performance Test MI35x ROCm 7.2 (8-GPU DeepSeek-R1-MXFP4)
|
|
timeout-minutes: 300
|
|
continue-on-error: true
|
|
run: |
|
|
> github_summary.md # Clear summary file
|
|
bash scripts/ci/amd/amd_ci_exec.sh -w /sglang-checkout/test \
|
|
-e GITHUB_STEP_SUMMARY="/sglang-checkout/github_summary.md" \
|
|
python3 registered/amd/perf/mi35x/test_deepseek_r1_mxfp4_perf_mi35x.py || TEST_EXIT_CODE=$?
|
|
echo "$(<github_summary.md )" >> $GITHUB_STEP_SUMMARY || true
|
|
exit ${TEST_EXIT_CODE:-0}
|
|
|
|
# MI35x 8-GPU DeepSeek-R1-MXFP4 KV FP8 (Accuracy + Performance) ROCm 7.2
|
|
nightly-8-gpu-mi35x-deepseek-r1-mxfp4-kv-fp8-rocm720:
|
|
if: (github.repository == 'sgl-project/sglang' || github.event_name == 'pull_request') && (!(inputs.job_filter || inputs.job_select) || (inputs.job_filter || inputs.job_select) == 'all' || contains(format(',{0},', inputs.job_filter || inputs.job_select), ',nightly-8-gpu-mi35x-deepseek-r1-mxfp4-kv-fp8-rocm720,'))
|
|
runs-on: linux-mi35x-gpu-8
|
|
steps:
|
|
- name: Checkout code
|
|
uses: actions/checkout@v4
|
|
with:
|
|
ref: ${{ inputs.ref || github.sha }}
|
|
|
|
- name: Ensure VRAM is clear
|
|
run: bash scripts/ci/amd/ensure_vram_clear.sh rocm
|
|
|
|
- name: Setup docker (ROCm 7.2)
|
|
run: |
|
|
touch github_summary.md
|
|
bash scripts/ci/amd/amd_ci_start_container.sh --rocm-version rocm720
|
|
env:
|
|
GITHUB_WORKSPACE: ${{ github.workspace }}
|
|
|
|
- name: Install dependencies
|
|
run: |
|
|
bash scripts/ci/amd/amd_ci_install_dependency.sh --skip-test-time-deps
|
|
# Install tabulate for run_suite.py (missing in MI35x container)
|
|
bash scripts/ci/amd/amd_ci_exec.sh pip install tabulate
|
|
|
|
- name: Accuracy Test MI35x ROCm 7.2 (8-GPU DeepSeek-R1-MXFP4 KV FP8)
|
|
timeout-minutes: 180
|
|
run: |
|
|
> github_summary.md # Clear summary file
|
|
bash scripts/ci/amd/amd_ci_exec.sh -w /sglang-checkout/test \
|
|
-e GITHUB_STEP_SUMMARY="/sglang-checkout/github_summary.md" \
|
|
python3 run_suite.py --hw amd --suite nightly-amd-8-gpu-mi35x-deepseek-r1-mxfp4-kv-fp8 --nightly --timeout-per-file 7200 ${{ (github.event_name == 'schedule' || inputs.continue_on_error) && '--continue-on-error' || '' }} || TEST_EXIT_CODE=$?
|
|
echo "$(<github_summary.md )" >> $GITHUB_STEP_SUMMARY || true
|
|
exit ${TEST_EXIT_CODE:-0}
|
|
|
|
- name: Performance Test MI35x ROCm 7.2 (8-GPU DeepSeek-R1-MXFP4 KV FP8)
|
|
timeout-minutes: 300
|
|
continue-on-error: true
|
|
run: |
|
|
> github_summary.md # Clear summary file
|
|
bash scripts/ci/amd/amd_ci_exec.sh -w /sglang-checkout/test \
|
|
-e GITHUB_STEP_SUMMARY="/sglang-checkout/github_summary.md" \
|
|
python3 registered/amd/perf/mi35x/test_deepseek_r1_mxfp4_kv_fp8_perf_mi35x.py || TEST_EXIT_CODE=$?
|
|
echo "$(<github_summary.md )" >> $GITHUB_STEP_SUMMARY || true
|
|
exit ${TEST_EXIT_CODE:-0}
|
|
|
|
# MI35x 8-GPU DeepSeek-R1-MXFP4 AllReduce Fusion (Accuracy + Performance) ROCm 7.2
|
|
nightly-8-gpu-mi35x-deepseek-r1-mxfp4-ar-fusion-rocm720:
|
|
if: (github.repository == 'sgl-project/sglang' || github.event_name == 'pull_request') && (!(inputs.job_filter || inputs.job_select) || (inputs.job_filter || inputs.job_select) == 'all' || contains(format(',{0},', inputs.job_filter || inputs.job_select), ',nightly-8-gpu-mi35x-deepseek-r1-mxfp4-ar-fusion-rocm720,'))
|
|
runs-on: linux-mi35x-gpu-8
|
|
steps:
|
|
- name: Checkout code
|
|
uses: actions/checkout@v4
|
|
with:
|
|
ref: ${{ inputs.ref || github.sha }}
|
|
|
|
- name: Ensure VRAM is clear
|
|
run: bash scripts/ci/amd/ensure_vram_clear.sh rocm
|
|
|
|
- name: Setup docker (ROCm 7.2)
|
|
run: |
|
|
touch github_summary.md
|
|
bash scripts/ci/amd/amd_ci_start_container.sh --rocm-version rocm720
|
|
env:
|
|
GITHUB_WORKSPACE: ${{ github.workspace }}
|
|
|
|
- name: Install dependencies
|
|
run: |
|
|
bash scripts/ci/amd/amd_ci_install_dependency.sh --skip-test-time-deps
|
|
# Install tabulate for run_suite.py (missing in MI35x container)
|
|
bash scripts/ci/amd/amd_ci_exec.sh pip install tabulate
|
|
|
|
- name: Accuracy Test MI35x ROCm 7.2 (8-GPU DeepSeek-R1-MXFP4 AllReduce Fusion)
|
|
timeout-minutes: 180
|
|
run: |
|
|
> github_summary.md # Clear summary file
|
|
bash scripts/ci/amd/amd_ci_exec.sh -w /sglang-checkout/test \
|
|
-e GITHUB_STEP_SUMMARY="/sglang-checkout/github_summary.md" \
|
|
python3 run_suite.py --hw amd --suite nightly-amd-8-gpu-mi35x-deepseek-r1-mxfp4-ar-fusion --nightly --timeout-per-file 7200 ${{ (github.event_name == 'schedule' || inputs.continue_on_error) && '--continue-on-error' || '' }} || TEST_EXIT_CODE=$?
|
|
echo "$(<github_summary.md )" >> $GITHUB_STEP_SUMMARY || true
|
|
exit ${TEST_EXIT_CODE:-0}
|
|
|
|
- name: Performance Test MI35x ROCm 7.2 (8-GPU DeepSeek-R1-MXFP4 AllReduce Fusion)
|
|
timeout-minutes: 300
|
|
continue-on-error: true
|
|
run: |
|
|
> github_summary.md # Clear summary file
|
|
bash scripts/ci/amd/amd_ci_exec.sh -w /sglang-checkout/test \
|
|
-e GITHUB_STEP_SUMMARY="/sglang-checkout/github_summary.md" \
|
|
python3 registered/amd/perf/mi35x/test_deepseek_r1_mxfp4_ar_fusion_perf_mi35x.py || TEST_EXIT_CODE=$?
|
|
echo "$(<github_summary.md )" >> $GITHUB_STEP_SUMMARY || true
|
|
exit ${TEST_EXIT_CODE:-0}
|
|
|
|
# MI35x 8-GPU DeepSeek-V3.2 Accuracy Test (ROCm 7.2)
|
|
nightly-accuracy-8-gpu-mi35x-deepseek-v32-rocm720:
|
|
if: (github.repository == 'sgl-project/sglang' || github.event_name == 'pull_request') && (!(inputs.job_filter || inputs.job_select) || (inputs.job_filter || inputs.job_select) == 'all' || contains(format(',{0},', inputs.job_filter || inputs.job_select), ',nightly-accuracy-8-gpu-mi35x-deepseek-v32-rocm720,'))
|
|
runs-on: linux-mi35x-gpu-8
|
|
steps:
|
|
- name: Checkout code
|
|
uses: actions/checkout@v4
|
|
with:
|
|
ref: ${{ inputs.ref || github.sha }}
|
|
|
|
- name: Ensure VRAM is clear
|
|
run: bash scripts/ci/amd/ensure_vram_clear.sh rocm
|
|
|
|
- name: Setup docker (ROCm 7.2)
|
|
run: |
|
|
touch github_summary.md
|
|
bash scripts/ci/amd/amd_ci_start_container.sh --rocm-version rocm720
|
|
env:
|
|
GITHUB_WORKSPACE: ${{ github.workspace }}
|
|
|
|
- name: Install dependencies
|
|
run: |
|
|
bash scripts/ci/amd/amd_ci_install_dependency.sh --skip-test-time-deps
|
|
# Install tabulate for run_suite.py (missing in MI35x container)
|
|
bash scripts/ci/amd/amd_ci_exec.sh pip install tabulate
|
|
|
|
- name: Accuracy Test MI35x ROCm 7.2 (8-GPU DeepSeek-V3.2)
|
|
timeout-minutes: 120
|
|
run: |
|
|
> github_summary.md # Clear summary file
|
|
bash scripts/ci/amd/amd_ci_exec.sh -w /sglang-checkout/test \
|
|
-e GITHUB_STEP_SUMMARY="/sglang-checkout/github_summary.md" \
|
|
python3 run_suite.py --hw amd --suite nightly-amd-8-gpu-mi35x-deepseek-v32 --nightly --timeout-per-file 3600 ${{ (github.event_name == 'schedule' || inputs.continue_on_error) && '--continue-on-error' || '' }} || TEST_EXIT_CODE=$?
|
|
echo "$(<github_summary.md )" >> $GITHUB_STEP_SUMMARY || true
|
|
exit ${TEST_EXIT_CODE:-0}
|
|
|
|
# MI35x 8-GPU DeepSeek-V3.2 TP+MTP Accuracy Test (ROCm 7.2)
|
|
nightly-accuracy-8-gpu-mi35x-deepseek-v32-mtp-rocm720:
|
|
if: (github.repository == 'sgl-project/sglang' || github.event_name == 'pull_request') && (!(inputs.job_filter || inputs.job_select) || (inputs.job_filter || inputs.job_select) == 'all' || contains(format(',{0},', inputs.job_filter || inputs.job_select), ',nightly-accuracy-8-gpu-mi35x-deepseek-v32-mtp-rocm720,'))
|
|
runs-on: linux-mi35x-gpu-8
|
|
steps:
|
|
- name: Checkout code
|
|
uses: actions/checkout@v4
|
|
with:
|
|
ref: ${{ inputs.ref || github.sha }}
|
|
|
|
- name: Ensure VRAM is clear
|
|
run: bash scripts/ci/amd/ensure_vram_clear.sh rocm
|
|
|
|
- name: Setup docker (ROCm 7.2)
|
|
run: |
|
|
touch github_summary.md
|
|
bash scripts/ci/amd/amd_ci_start_container.sh --rocm-version rocm720
|
|
env:
|
|
GITHUB_WORKSPACE: ${{ github.workspace }}
|
|
|
|
- name: Install dependencies
|
|
run: |
|
|
bash scripts/ci/amd/amd_ci_install_dependency.sh --skip-test-time-deps
|
|
# Install tabulate for run_suite.py (missing in MI35x container)
|
|
bash scripts/ci/amd/amd_ci_exec.sh pip install tabulate
|
|
|
|
- name: Accuracy Test MI35x ROCm 7.2 (8-GPU DeepSeek-V3.2 TP+MTP)
|
|
timeout-minutes: 120
|
|
run: |
|
|
> github_summary.md # Clear summary file
|
|
bash scripts/ci/amd/amd_ci_exec.sh -w /sglang-checkout/test \
|
|
-e GITHUB_STEP_SUMMARY="/sglang-checkout/github_summary.md" \
|
|
python3 run_suite.py --hw amd --suite nightly-amd-accuracy-8-gpu-mi35x-deepseek-v32-mtp --nightly --timeout-per-file 3600 ${{ (github.event_name == 'schedule' || inputs.continue_on_error) && '--continue-on-error' || '' }} || TEST_EXIT_CODE=$?
|
|
echo "$(<github_summary.md )" >> $GITHUB_STEP_SUMMARY || true
|
|
exit ${TEST_EXIT_CODE:-0}
|
|
|
|
# MI35x 8-GPU DeepSeek-V3.2 Performance Test (Basic) ROCm 7.2
|
|
nightly-perf-8-gpu-mi35x-deepseek-v32-basic-rocm720:
|
|
if: (github.repository == 'sgl-project/sglang' || github.event_name == 'pull_request') && (!(inputs.job_filter || inputs.job_select) || (inputs.job_filter || inputs.job_select) == 'all' || contains(format(',{0},', inputs.job_filter || inputs.job_select), ',nightly-perf-8-gpu-mi35x-deepseek-v32-basic-rocm720,'))
|
|
runs-on: linux-mi35x-gpu-8
|
|
steps:
|
|
- name: Checkout code
|
|
uses: actions/checkout@v4
|
|
with:
|
|
ref: ${{ inputs.ref || github.sha }}
|
|
|
|
- name: Ensure VRAM is clear
|
|
run: bash scripts/ci/amd/ensure_vram_clear.sh rocm
|
|
|
|
- name: Setup docker (ROCm 7.2)
|
|
run: |
|
|
touch github_summary.md
|
|
bash scripts/ci/amd/amd_ci_start_container.sh --rocm-version rocm720
|
|
env:
|
|
GITHUB_WORKSPACE: ${{ github.workspace }}
|
|
|
|
- name: Install dependencies
|
|
run: |
|
|
bash scripts/ci/amd/amd_ci_install_dependency.sh --skip-test-time-deps
|
|
# Install tabulate for run_suite.py (missing in MI35x container)
|
|
bash scripts/ci/amd/amd_ci_exec.sh pip install tabulate
|
|
|
|
- name: Performance Test MI35x ROCm 7.2 (8-GPU DeepSeek-V3.2 Basic)
|
|
timeout-minutes: 150
|
|
run: |
|
|
> github_summary.md # Clear summary file
|
|
bash scripts/ci/amd/amd_ci_exec.sh -w /sglang-checkout/test \
|
|
-e GITHUB_STEP_SUMMARY="/sglang-checkout/github_summary.md" \
|
|
python3 run_suite.py --hw amd --suite nightly-perf-8-gpu-mi35x-deepseek-v32-basic --nightly --timeout-per-file 5400 ${{ (github.event_name == 'schedule' || inputs.continue_on_error) && '--continue-on-error' || '' }} || TEST_EXIT_CODE=$?
|
|
echo "$(<github_summary.md )" >> $GITHUB_STEP_SUMMARY || true
|
|
exit ${TEST_EXIT_CODE:-0}
|
|
|
|
# MI35x 8-GPU Kimi-K2.6 (Accuracy) ROCm 7.2
|
|
nightly-8-gpu-mi35x-kimi-k26-rocm720:
|
|
if: (github.repository == 'sgl-project/sglang' || github.event_name == 'pull_request') && (!(inputs.job_filter || inputs.job_select) || (inputs.job_filter || inputs.job_select) == 'all' || contains(format(',{0},', inputs.job_filter || inputs.job_select), ',nightly-8-gpu-mi35x-kimi-k26-rocm720,'))
|
|
runs-on: linux-mi35x-gpu-8
|
|
steps:
|
|
- name: Checkout code
|
|
uses: actions/checkout@v4
|
|
with:
|
|
ref: ${{ inputs.ref || github.sha }}
|
|
|
|
- name: Ensure VRAM is clear
|
|
run: bash scripts/ci/amd/ensure_vram_clear.sh rocm
|
|
|
|
- name: Setup docker (ROCm 7.2)
|
|
run: |
|
|
touch github_summary.md
|
|
bash scripts/ci/amd/amd_ci_start_container.sh --rocm-version rocm720
|
|
env:
|
|
GITHUB_WORKSPACE: ${{ github.workspace }}
|
|
|
|
- name: Install dependencies
|
|
run: |
|
|
bash scripts/ci/amd/amd_ci_install_dependency.sh --skip-test-time-deps
|
|
# Install tabulate for run_suite.py (missing in MI35x container)
|
|
bash scripts/ci/amd/amd_ci_exec.sh pip install tabulate
|
|
|
|
- name: Accuracy Test MI35x ROCm 7.2 (8-GPU Kimi-K2.6)
|
|
timeout-minutes: 180
|
|
run: |
|
|
> github_summary.md # Clear summary file
|
|
bash scripts/ci/amd/amd_ci_exec.sh -w /sglang-checkout/test \
|
|
-e GITHUB_STEP_SUMMARY="/sglang-checkout/github_summary.md" \
|
|
python3 run_suite.py --hw amd --suite nightly-amd-accuracy-8-gpu-mi35x-kimi-k26 --nightly --timeout-per-file 7200 ${{ (github.event_name == 'schedule' || inputs.continue_on_error) && '--continue-on-error' || '' }} || TEST_EXIT_CODE=$?
|
|
echo "$(<github_summary.md )" >> $GITHUB_STEP_SUMMARY || true
|
|
exit ${TEST_EXIT_CODE:-0}
|
|
|
|
# MI35x 8-GPU Qwen3-235B-MXFP4 (Accuracy + Performance) ROCm 7.2
|
|
nightly-8-gpu-mi35x-qwen3-235b-mxfp4-rocm720:
|
|
if: (github.repository == 'sgl-project/sglang' || github.event_name == 'pull_request') && (!(inputs.job_filter || inputs.job_select) || (inputs.job_filter || inputs.job_select) == 'all' || contains(format(',{0},', inputs.job_filter || inputs.job_select), ',nightly-8-gpu-mi35x-qwen3-235b-mxfp4-rocm720,'))
|
|
runs-on: linux-mi35x-gpu-8
|
|
steps:
|
|
- name: Checkout code
|
|
uses: actions/checkout@v4
|
|
with:
|
|
ref: ${{ inputs.ref || github.sha }}
|
|
|
|
- name: Ensure VRAM is clear
|
|
run: bash scripts/ci/amd/ensure_vram_clear.sh rocm
|
|
|
|
- name: Setup docker (ROCm 7.2)
|
|
run: |
|
|
touch github_summary.md
|
|
bash scripts/ci/amd/amd_ci_start_container.sh --rocm-version rocm720
|
|
env:
|
|
GITHUB_WORKSPACE: ${{ github.workspace }}
|
|
|
|
- name: Install dependencies
|
|
run: |
|
|
bash scripts/ci/amd/amd_ci_install_dependency.sh --skip-test-time-deps
|
|
# Install tabulate for run_suite.py (missing in MI35x container)
|
|
bash scripts/ci/amd/amd_ci_exec.sh pip install tabulate
|
|
|
|
- name: Accuracy Test + Performance Test MI35x ROCm 7.2 (8-GPU Qwen3-235B-MXFP4)
|
|
timeout-minutes: 120
|
|
run: |
|
|
> github_summary.md # Clear summary file
|
|
bash scripts/ci/amd/amd_ci_exec.sh -w /sglang-checkout/test \
|
|
-e GITHUB_STEP_SUMMARY="/sglang-checkout/github_summary.md" \
|
|
python3 run_suite.py --hw amd --suite nightly-8-gpu-mi35x-qwen3-235b-mxfp4 --nightly --timeout-per-file 3600 ${{ (github.event_name == 'schedule' || inputs.continue_on_error) && '--continue-on-error' || '' }} || TEST_EXIT_CODE=$?
|
|
echo "$(<github_summary.md )" >> $GITHUB_STEP_SUMMARY || true
|
|
exit ${TEST_EXIT_CODE:-0}
|
|
|
|
# MI35x 8-GPU Qwen 3.5 (Accuracy + Performance combined) ROCm 7.2
|
|
nightly-8-gpu-mi35x-qwen35-rocm720:
|
|
if: (github.repository == 'sgl-project/sglang' || github.event_name == 'pull_request') && (!(inputs.job_filter || inputs.job_select) || (inputs.job_filter || inputs.job_select) == 'all' || contains(format(',{0},', inputs.job_filter || inputs.job_select), ',nightly-8-gpu-mi35x-qwen35-rocm720,'))
|
|
runs-on: linux-mi35x-gpu-8
|
|
steps:
|
|
- name: Checkout code
|
|
uses: actions/checkout@v4
|
|
with:
|
|
ref: ${{ inputs.ref || github.sha }}
|
|
|
|
- name: Ensure VRAM is clear
|
|
run: bash scripts/ci/amd/ensure_vram_clear.sh rocm
|
|
|
|
- name: Setup docker (ROCm 7.2)
|
|
run: |
|
|
touch github_summary.md
|
|
bash scripts/ci/amd/amd_ci_start_container.sh --rocm-version rocm720
|
|
env:
|
|
GITHUB_WORKSPACE: ${{ github.workspace }}
|
|
|
|
- name: Install dependencies
|
|
run: |
|
|
bash scripts/ci/amd/amd_ci_install_dependency.sh --skip-aiter-build --skip-test-time-deps
|
|
bash scripts/ci/amd/amd_ci_exec.sh pip install tabulate
|
|
bash scripts/ci/amd/amd_ci_exec.sh pip install mistral-common "lm-eval[api]"
|
|
|
|
- name: Accuracy Test MI35x ROCm 7.2 (8-GPU Qwen 3.5)
|
|
timeout-minutes: 120
|
|
run: |
|
|
> github_summary.md # Clear summary file
|
|
bash scripts/ci/amd/amd_ci_exec.sh -w /sglang-checkout/test \
|
|
-e GITHUB_STEP_SUMMARY="/sglang-checkout/github_summary.md" \
|
|
python3 run_suite.py --hw amd --suite nightly-amd-accuracy-8-gpu-mi35x-qwen35 --nightly --timeout-per-file 3600 ${{ (github.event_name == 'schedule' || inputs.continue_on_error) && '--continue-on-error' || '' }} || TEST_EXIT_CODE=$?
|
|
echo "$(<github_summary.md )" >> $GITHUB_STEP_SUMMARY || true
|
|
exit ${TEST_EXIT_CODE:-0}
|
|
|
|
- name: Performance Test MI35x ROCm 7.2 (8-GPU Qwen 3.5 FP8)
|
|
timeout-minutes: 120
|
|
continue-on-error: true
|
|
run: |
|
|
> github_summary.md # Clear summary file
|
|
bash scripts/ci/amd/amd_ci_exec.sh -w /sglang-checkout/test \
|
|
-e SGLANG_USE_AITER=1 \
|
|
-e GITHUB_STEP_SUMMARY="/sglang-checkout/github_summary.md" \
|
|
python3 run_suite.py --hw amd --suite nightly-perf-8-gpu-mi35x-qwen35-fp8 --nightly --timeout-per-file 5400 ${{ (github.event_name == 'schedule' || inputs.continue_on_error) && '--continue-on-error' || '' }} || TEST_EXIT_CODE=$?
|
|
echo "$(<github_summary.md )" >> $GITHUB_STEP_SUMMARY || true
|
|
exit ${TEST_EXIT_CODE:-0}
|
|
|
|
# MI35x 8-GPU GLM-5.1 (Accuracy + Performance combined) ROCm 7.2
|
|
nightly-8-gpu-mi35x-glm51-rocm720:
|
|
if: (github.repository == 'sgl-project/sglang' || github.event_name == 'pull_request') && (!(inputs.job_filter || inputs.job_select) || (inputs.job_filter || inputs.job_select) == 'all' || contains(format(',{0},', inputs.job_filter || inputs.job_select), ',nightly-8-gpu-mi35x-glm51-rocm720,'))
|
|
runs-on: linux-mi35x-gpu-8
|
|
steps:
|
|
- name: Checkout code
|
|
uses: actions/checkout@v4
|
|
with:
|
|
ref: ${{ inputs.ref || github.sha }}
|
|
|
|
- name: Ensure VRAM is clear
|
|
run: bash scripts/ci/amd/ensure_vram_clear.sh rocm
|
|
|
|
- name: Setup docker (ROCm 7.2)
|
|
run: |
|
|
touch github_summary.md
|
|
bash scripts/ci/amd/amd_ci_start_container.sh --rocm-version rocm720
|
|
env:
|
|
GITHUB_WORKSPACE: ${{ github.workspace }}
|
|
|
|
- name: Install dependencies
|
|
run: |
|
|
bash scripts/ci/amd/amd_ci_install_dependency.sh --skip-test-time-deps
|
|
bash scripts/ci/amd/amd_ci_exec.sh pip install tabulate
|
|
bash scripts/ci/amd/amd_ci_exec.sh pip install git+https://github.com/huggingface/transformers.git@96f807a33b75
|
|
|
|
- name: Accuracy Test MI35x ROCm 7.2 (8-GPU GLM-5.1 NSA)
|
|
timeout-minutes: 180
|
|
run: |
|
|
> github_summary.md # Clear summary file
|
|
bash scripts/ci/amd/amd_ci_exec.sh -w /sglang-checkout/test \
|
|
-e GITHUB_STEP_SUMMARY="/sglang-checkout/github_summary.md" \
|
|
python3 run_suite.py --hw amd --suite nightly-amd-8-gpu-mi35x-glm51 --nightly --timeout-per-file 7200 ${{ (github.event_name == 'schedule' || inputs.continue_on_error) && '--continue-on-error' || '' }} || TEST_EXIT_CODE=$?
|
|
echo "$(<github_summary.md )" >> $GITHUB_STEP_SUMMARY || true
|
|
exit ${TEST_EXIT_CODE:-0}
|
|
|
|
- name: Performance Test MI35x ROCm 7.2 (8-GPU GLM-5.1)
|
|
timeout-minutes: 120
|
|
continue-on-error: true
|
|
run: |
|
|
> github_summary.md # Clear summary file
|
|
bash scripts/ci/amd/amd_ci_exec.sh -w /sglang-checkout/test \
|
|
-e GITHUB_STEP_SUMMARY="/sglang-checkout/github_summary.md" \
|
|
python3 run_suite.py --hw amd --suite nightly-perf-8-gpu-mi35x-glm51 --nightly --timeout-per-file 5400 ${{ (github.event_name == 'schedule' || inputs.continue_on_error) && '--continue-on-error' || '' }} || TEST_EXIT_CODE=$?
|
|
echo "$(<github_summary.md )" >> $GITHUB_STEP_SUMMARY || true
|
|
exit ${TEST_EXIT_CODE:-0}
|
|
|
|
# MI35x 8-GPU GLM-5-MXFP4 (Accuracy + Performance combined) ROCm 7.2
|
|
nightly-8-gpu-mi35x-glm5-mxfp4-rocm720:
|
|
if: (github.repository == 'sgl-project/sglang' || github.event_name == 'pull_request') && (!(inputs.job_filter || inputs.job_select) || (inputs.job_filter || inputs.job_select) == 'all' || contains(format(',{0},', inputs.job_filter || inputs.job_select), ',nightly-8-gpu-mi35x-glm5-mxfp4-rocm720,'))
|
|
runs-on: linux-mi35x-gpu-8
|
|
steps:
|
|
- name: Checkout code
|
|
uses: actions/checkout@v4
|
|
with:
|
|
ref: ${{ inputs.ref || github.sha }}
|
|
|
|
- name: Ensure VRAM is clear
|
|
run: bash scripts/ci/amd/ensure_vram_clear.sh rocm
|
|
|
|
- name: Setup docker (ROCm 7.2)
|
|
run: |
|
|
touch github_summary.md
|
|
bash scripts/ci/amd/amd_ci_start_container.sh --rocm-version rocm720
|
|
env:
|
|
GITHUB_WORKSPACE: ${{ github.workspace }}
|
|
|
|
- name: Install dependencies
|
|
run: |
|
|
bash scripts/ci/amd/amd_ci_install_dependency.sh --skip-test-time-deps
|
|
bash scripts/ci/amd/amd_ci_exec.sh pip install tabulate
|
|
bash scripts/ci/amd/amd_ci_exec.sh pip install git+https://github.com/huggingface/transformers.git@96f807a33b75
|
|
|
|
- name: Accuracy Test MI35x ROCm 7.2 (8-GPU GLM-5-MXFP4)
|
|
timeout-minutes: 180
|
|
run: |
|
|
> github_summary.md
|
|
bash scripts/ci/amd/amd_ci_exec.sh -w /sglang-checkout/test \
|
|
-e SGLANG_USE_AITER=1 \
|
|
-e GITHUB_STEP_SUMMARY="/sglang-checkout/github_summary.md" \
|
|
python3 run_suite.py --hw amd --suite nightly-amd-8-gpu-mi35x-glm5-mxfp4 --nightly --timeout-per-file 7200 ${{ (github.event_name == 'schedule' || inputs.continue_on_error) && '--continue-on-error' || '' }} || TEST_EXIT_CODE=$?
|
|
echo "$(<github_summary.md )" >> $GITHUB_STEP_SUMMARY || true
|
|
exit ${TEST_EXIT_CODE:-0}
|
|
|
|
- name: Performance Test MI35x ROCm 7.2 (8-GPU GLM-5-MXFP4)
|
|
timeout-minutes: 300
|
|
continue-on-error: true
|
|
run: |
|
|
> github_summary.md
|
|
bash scripts/ci/amd/amd_ci_exec.sh -w /sglang-checkout/test \
|
|
-e SGLANG_USE_AITER=1 \
|
|
-e GITHUB_STEP_SUMMARY="/sglang-checkout/github_summary.md" \
|
|
python3 registered/amd/perf/mi35x/test_glm5_mxfp4_perf_mi35x.py || TEST_EXIT_CODE=$?
|
|
echo "$(<github_summary.md )" >> $GITHUB_STEP_SUMMARY || true
|
|
exit ${TEST_EXIT_CODE:-0}
|
|
|
|
# MI35x 8-GPU DeepSeek-V3.2 Performance Test (MTP) ROCm 7.2
|
|
nightly-perf-8-gpu-mi35x-deepseek-v32-mtp-rocm720:
|
|
if: (github.repository == 'sgl-project/sglang' || github.event_name == 'pull_request') && (!(inputs.job_filter || inputs.job_select) || (inputs.job_filter || inputs.job_select) == 'all' || contains(format(',{0},', inputs.job_filter || inputs.job_select), ',nightly-perf-8-gpu-mi35x-deepseek-v32-mtp-rocm720,'))
|
|
runs-on: linux-mi35x-gpu-8
|
|
steps:
|
|
- name: Checkout code
|
|
uses: actions/checkout@v4
|
|
with:
|
|
ref: ${{ inputs.ref || github.sha }}
|
|
|
|
- name: Ensure VRAM is clear
|
|
run: bash scripts/ci/amd/ensure_vram_clear.sh rocm
|
|
|
|
- name: Setup docker (ROCm 7.2)
|
|
run: |
|
|
touch github_summary.md
|
|
bash scripts/ci/amd/amd_ci_start_container.sh --rocm-version rocm720
|
|
env:
|
|
GITHUB_WORKSPACE: ${{ github.workspace }}
|
|
|
|
- name: Install dependencies
|
|
run: |
|
|
bash scripts/ci/amd/amd_ci_install_dependency.sh --skip-test-time-deps
|
|
# Install tabulate for run_suite.py (missing in MI35x container)
|
|
bash scripts/ci/amd/amd_ci_exec.sh pip install tabulate
|
|
|
|
- name: Performance Test MI35x ROCm 7.2 (8-GPU DeepSeek-V3.2 MTP)
|
|
timeout-minutes: 180
|
|
run: |
|
|
> github_summary.md # Clear summary file
|
|
bash scripts/ci/amd/amd_ci_exec.sh -w /sglang-checkout/test \
|
|
-e GITHUB_STEP_SUMMARY="/sglang-checkout/github_summary.md" \
|
|
python3 run_suite.py --hw amd --suite nightly-perf-8-gpu-mi35x-deepseek-v32-mtp --nightly --timeout-per-file 7200 ${{ (github.event_name == 'schedule' || inputs.continue_on_error) && '--continue-on-error' || '' }} || TEST_EXIT_CODE=$?
|
|
echo "$(<github_summary.md )" >> $GITHUB_STEP_SUMMARY || true
|
|
exit ${TEST_EXIT_CODE:-0}
|
|
|
|
# MI35x 8-GPU DeepSeek-V4-Flash FP8 + FP4 (Accuracy + Performance combined) ROCm 7.2
|
|
# NOTE on runtime sourcing: the DSv4 docker image (tag suffix `-DSv4`) bakes
|
|
# in sglang built from a specific commit of the amd/deepseek_v4 branch (the
|
|
# 7-char sha in the image tag is that commit). To keep the runtime as exactly
|
|
# that image-frozen sglang/aiter, we pass `--skip-sglang-build` and
|
|
# `--skip-aiter-build` so install_dependency.sh does NOT `pip install -e
|
|
# /sglang-checkout/python` (which would override the image's sglang with
|
|
# whatever this checkout happens to be) and does NOT rebuild aiter from this
|
|
# checkout's docker/rocm.Dockerfile. The /sglang-checkout mount is still used
|
|
# for shell scripts and for run_suite.py discovering test files; it does not
|
|
# poison Python imports because the image's site-packages .pth points at
|
|
# /sgl-workspace/sglang/python (a different path).
|
|
nightly-8-gpu-mi35x-deepseek-v4-flash-rocm720:
|
|
if: (github.repository == 'sgl-project/sglang' || github.event_name == 'pull_request') && (!(inputs.job_filter || inputs.job_select) || (inputs.job_filter || inputs.job_select) == 'all' || contains(format(',{0},', inputs.job_filter || inputs.job_select), ',nightly-8-gpu-mi35x-deepseek-v4-flash-rocm720,'))
|
|
runs-on: linux-mi35x-gpu-8
|
|
steps:
|
|
- name: Checkout code
|
|
uses: actions/checkout@v4
|
|
with:
|
|
ref: ${{ inputs.ref || github.sha }}
|
|
|
|
- name: Resolve DSv4 image tag
|
|
id: dsv4_image
|
|
run: |
|
|
# Pick the latest Docker Hub tag matching rocm720-mi35x-<sha7>-<YYYYMMDD>-DSv4.
|
|
# Docker Hub returns results sorted by last_updated DESC by default, so the
|
|
# first regex match is the most recent daily build.
|
|
AUTH_HEADER=()
|
|
if [[ -n "${DOCKERHUB_AMD_USERNAME:-}" && -n "${DOCKERHUB_AMD_TOKEN:-}" ]]; then
|
|
TOKEN=$(curl -s -H "Content-Type: application/json" \
|
|
-X POST -d "{\"username\":\"${DOCKERHUB_AMD_USERNAME}\",\"password\":\"${DOCKERHUB_AMD_TOKEN}\"}" \
|
|
https://hub.docker.com/v2/users/login/ | python3 -c "import json,sys; print(json.load(sys.stdin).get('token',''))")
|
|
if [[ -n "$TOKEN" ]]; then
|
|
AUTH_HEADER=(-H "Authorization: JWT $TOKEN")
|
|
fi
|
|
fi
|
|
TAG=$(curl -s "${AUTH_HEADER[@]}" \
|
|
"https://hub.docker.com/v2/repositories/rocm/sgl-dev/tags?page_size=100&name=DSv4" \
|
|
| grep -oE '"name":"rocm720-mi35x-[a-f0-9]{7}-[0-9]{8}-DSv4"' \
|
|
| head -n 1 | cut -d'"' -f4)
|
|
if [ -z "$TAG" ]; then
|
|
echo "::error::No DSv4 image found matching rocm720-mi35x-<sha7>-<YYYYMMDD>-DSv4 on Docker Hub"
|
|
exit 1
|
|
fi
|
|
echo "image=rocm/sgl-dev:$TAG" >> "$GITHUB_OUTPUT"
|
|
echo "Resolved DSv4 image: rocm/sgl-dev:$TAG"
|
|
|
|
- name: Ensure VRAM is clear
|
|
run: bash scripts/ci/amd/ensure_vram_clear.sh rocm
|
|
|
|
- name: Setup docker (ROCm 7.2 DSv4)
|
|
run: |
|
|
touch github_summary.md
|
|
bash scripts/ci/amd/amd_ci_start_container.sh --custom-image ${{ steps.dsv4_image.outputs.image }}
|
|
env:
|
|
GITHUB_WORKSPACE: ${{ github.workspace }}
|
|
|
|
- name: Install dependencies (preserve DSv4 sglang/aiter from image)
|
|
run: |
|
|
# --skip-sglang-build: keep the image's pre-installed DSv4 sglang
|
|
# (default would `pip install -e /sglang-checkout/python` and clobber it with main's source).
|
|
# --skip-aiter-build: keep the image's DSv4-tuned aiter
|
|
# (default reads /sglang-checkout/docker/rocm.Dockerfile from main and rebuilds aiter to that commit).
|
|
# --skip-test-time-deps: GSM8K + bench_one_batch_server don't need lmms-eval / human-eval.
|
|
bash scripts/ci/amd/amd_ci_install_dependency.sh \
|
|
--skip-sglang-build --skip-aiter-build --skip-test-time-deps
|
|
# tabulate is the only thing run_suite.py imports that may not be in the DSv4 image.
|
|
bash scripts/ci/amd/amd_ci_exec.sh pip install tabulate
|
|
|
|
- name: Accuracy + Performance Test MI35x ROCm 7.2 (8-GPU DeepSeek-V4-Flash FP8 + FP4)
|
|
timeout-minutes: 300
|
|
run: |
|
|
> github_summary.md # Clear summary file
|
|
bash scripts/ci/amd/amd_ci_exec.sh -w /sglang-checkout/test \
|
|
-e GITHUB_STEP_SUMMARY="/sglang-checkout/github_summary.md" \
|
|
python3 run_suite.py --hw amd --suite nightly-amd-8-gpu-mi35x-deepseek-v4-flash --nightly --timeout-per-file 7200 ${{ (github.event_name == 'schedule' || inputs.continue_on_error) && '--continue-on-error' || '' }} || TEST_EXIT_CODE=$?
|
|
echo "$(<github_summary.md )" >> $GITHUB_STEP_SUMMARY || true
|
|
exit ${TEST_EXIT_CODE:-0}
|
|
|
|
# MI35x 8-GPU DeepSeek-V4-Pro FP8 + FP4 (Accuracy + Performance combined) ROCm 7.2
|
|
# Pro is 1.6T (vs Flash 285B); load + warmup is much longer, so timeout-per-file
|
|
# and the job timeout are both larger than the Flash job.
|
|
# Same image / branch / install strategy as the Flash job above — see the comment
|
|
# block on `nightly-8-gpu-mi35x-deepseek-v4-flash-rocm720` for the rationale.
|
|
nightly-8-gpu-mi35x-deepseek-v4-pro-rocm720:
|
|
if: (github.repository == 'sgl-project/sglang' || github.event_name == 'pull_request') && (!(inputs.job_filter || inputs.job_select) || (inputs.job_filter || inputs.job_select) == 'all' || contains(format(',{0},', inputs.job_filter || inputs.job_select), ',nightly-8-gpu-mi35x-deepseek-v4-pro-rocm720,'))
|
|
runs-on: linux-mi35x-gpu-8
|
|
steps:
|
|
- name: Checkout code
|
|
uses: actions/checkout@v4
|
|
with:
|
|
ref: ${{ inputs.ref || github.sha }}
|
|
|
|
- name: Resolve DSv4 image tag
|
|
id: dsv4_image
|
|
run: |
|
|
AUTH_HEADER=()
|
|
if [[ -n "${DOCKERHUB_AMD_USERNAME:-}" && -n "${DOCKERHUB_AMD_TOKEN:-}" ]]; then
|
|
TOKEN=$(curl -s -H "Content-Type: application/json" \
|
|
-X POST -d "{\"username\":\"${DOCKERHUB_AMD_USERNAME}\",\"password\":\"${DOCKERHUB_AMD_TOKEN}\"}" \
|
|
https://hub.docker.com/v2/users/login/ | python3 -c "import json,sys; print(json.load(sys.stdin).get('token',''))")
|
|
if [[ -n "$TOKEN" ]]; then
|
|
AUTH_HEADER=(-H "Authorization: JWT $TOKEN")
|
|
fi
|
|
fi
|
|
TAG=$(curl -s "${AUTH_HEADER[@]}" \
|
|
"https://hub.docker.com/v2/repositories/rocm/sgl-dev/tags?page_size=100&name=DSv4" \
|
|
| grep -oE '"name":"rocm720-mi35x-[a-f0-9]{7}-[0-9]{8}-DSv4"' \
|
|
| head -n 1 | cut -d'"' -f4)
|
|
if [ -z "$TAG" ]; then
|
|
echo "::error::No DSv4 image found matching rocm720-mi35x-<sha7>-<YYYYMMDD>-DSv4 on Docker Hub"
|
|
exit 1
|
|
fi
|
|
echo "image=rocm/sgl-dev:$TAG" >> "$GITHUB_OUTPUT"
|
|
echo "Resolved DSv4 image: rocm/sgl-dev:$TAG"
|
|
|
|
- name: Ensure VRAM is clear
|
|
run: bash scripts/ci/amd/ensure_vram_clear.sh rocm
|
|
|
|
- name: Setup docker (ROCm 7.2 DSv4)
|
|
run: |
|
|
touch github_summary.md
|
|
bash scripts/ci/amd/amd_ci_start_container.sh --custom-image ${{ steps.dsv4_image.outputs.image }}
|
|
env:
|
|
GITHUB_WORKSPACE: ${{ github.workspace }}
|
|
|
|
- name: Install dependencies (preserve DSv4 sglang/aiter from image)
|
|
run: |
|
|
bash scripts/ci/amd/amd_ci_install_dependency.sh \
|
|
--skip-sglang-build --skip-aiter-build --skip-test-time-deps
|
|
bash scripts/ci/amd/amd_ci_exec.sh pip install tabulate
|
|
|
|
- name: Accuracy + Performance Test MI35x ROCm 7.2 (8-GPU DeepSeek-V4-Pro FP8 + FP4)
|
|
timeout-minutes: 480
|
|
run: |
|
|
> github_summary.md # Clear summary file
|
|
bash scripts/ci/amd/amd_ci_exec.sh -w /sglang-checkout/test \
|
|
-e GITHUB_STEP_SUMMARY="/sglang-checkout/github_summary.md" \
|
|
python3 run_suite.py --hw amd --suite nightly-amd-8-gpu-mi35x-deepseek-v4-pro --nightly --timeout-per-file 14400 ${{ (github.event_name == 'schedule' || inputs.continue_on_error) && '--continue-on-error' || '' }} || TEST_EXIT_CODE=$?
|
|
echo "$(<github_summary.md )" >> $GITHUB_STEP_SUMMARY || true
|
|
exit ${TEST_EXIT_CODE:-0}
|
|
|
|
check-all-jobs:
|
|
if: always() && (github.repository == 'sgl-project/sglang' || github.event_name == 'pull_request' || github.event_name == 'workflow_dispatch')
|
|
needs:
|
|
# MI30x ROCm 7.2 Unit Tests
|
|
- nightly-test-1-gpu-unit-rocm720
|
|
# MI30x ROCm 7.2 Accuracy Tests
|
|
- nightly-accuracy-2-gpu-rocm720
|
|
- nightly-accuracy-2-gpu-vlm-rocm720
|
|
# MI30x ROCm 7.2 Performance Tests
|
|
- nightly-perf-2-gpu-text-rocm720
|
|
- nightly-perf-2-gpu-vlm-rocm720
|
|
# MI30x ROCm 7.2 4-GPU Tests
|
|
- nightly-4-gpu-rocm720
|
|
- nightly-accuracy-8-gpu-rocm720
|
|
# MI30x ROCm 7.2 Combined Accuracy + Performance Tests
|
|
- nightly-8-gpu-grok1-int4-rocm720
|
|
- nightly-8-gpu-grok2-rocm720
|
|
- nightly-8-gpu-deepseek-v31-rocm720
|
|
- nightly-8-gpu-deepseek-v32-rocm720
|
|
- nightly-8-gpu-deepseek-v32-mtp-rocm720
|
|
- nightly-8-gpu-deepseek-v3-kv-fp8-rocm720
|
|
- nightly-8-gpu-kimi-k26-rocm720
|
|
- nightly-8-gpu-qwen3-235b-rocm720
|
|
- nightly-8-gpu-qwen35-rocm720
|
|
- nightly-8-gpu-glm51-rocm720
|
|
- nightly-8-gpu-minimax-m27-rocm720
|
|
# MI30x ROCm 7.2 Diffusion Tests
|
|
- nightly-1-gpu-zimage-turbo-rocm720
|
|
# MI35x ROCm 7.2 jobs
|
|
- nightly-test-1-gpu-mi35x-rocm720
|
|
- nightly-accuracy-8-gpu-mi35x-rocm720
|
|
- nightly-8-gpu-mi35x-grok1-int4-rocm720
|
|
- nightly-8-gpu-mi35x-grok2-rocm720
|
|
- nightly-8-gpu-mi35x-deepseek-r1-mxfp4-rocm720
|
|
- nightly-8-gpu-mi35x-deepseek-r1-mxfp4-kv-fp8-rocm720
|
|
- nightly-8-gpu-mi35x-deepseek-r1-mxfp4-ar-fusion-rocm720
|
|
- nightly-accuracy-8-gpu-mi35x-deepseek-v32-rocm720
|
|
- nightly-accuracy-8-gpu-mi35x-deepseek-v32-mtp-rocm720
|
|
- nightly-perf-8-gpu-mi35x-deepseek-v32-basic-rocm720
|
|
- nightly-perf-8-gpu-mi35x-deepseek-v32-mtp-rocm720
|
|
- nightly-8-gpu-mi35x-deepseek-v4-flash-rocm720
|
|
- nightly-8-gpu-mi35x-deepseek-v4-pro-rocm720
|
|
- nightly-8-gpu-mi35x-kimi-k26-rocm720
|
|
- nightly-8-gpu-mi35x-qwen3-235b-mxfp4-rocm720
|
|
- nightly-8-gpu-mi35x-qwen35-rocm720
|
|
- nightly-8-gpu-mi35x-glm51-rocm720
|
|
- nightly-8-gpu-mi35x-glm5-mxfp4-rocm720
|
|
runs-on: ubuntu-latest
|
|
steps:
|
|
- name: Check if any job failed
|
|
run: |
|
|
if [[ "${{ contains(needs.*.result, 'failure') }}" == "true" ]]; then
|
|
echo "One or more ROCm 7.2 nightly test jobs failed"
|
|
exit 1
|
|
fi
|
|
if [[ "${{ contains(needs.*.result, 'cancelled') }}" == "true" ]]; then
|
|
echo "One or more ROCm 7.2 nightly test jobs were cancelled"
|
|
exit 1
|
|
fi
|
|
echo "All ROCm 7.2 nightly test jobs passed"
|