Skip to content

[AMD] Add GLM-5.1-FP8 nightly accuracy and performance benchmarks for MI30x and MI35x #22336

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
HaiShaw merged 4 commits into from
Apr 9, 2026
Merged

[AMD] Add GLM-5.1-FP8 nightly accuracy and performance benchmarks for MI30x and MI35x #22336

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
4 commits
Select commit Hold shift + click to select a range
5e0e308
[AMD] Add GLM-5.1-FP8 nightly accuracy and performance benchmarks for…
michaelzhang-ai Apr 8, 2026
1a664fa
[AMD][CI] Remove GLM-4.7-FP8 nightly job from ROCm 7.2 workflow
michaelzhang-ai Apr 9, 2026
eb5f0d6
[AMD][CI] Drop GLM-5 MI35x perf test — accuracy only
michaelzhang-ai Apr 9, 2026
55a15bf
[AMD][CI] Remove --ep-size from GLM-5.1 tests to match GLM-5 config
michaelzhang-ai Apr 9, 2026
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
88 changes: 67 additions & 21 deletions .github/workflows/nightly-test-amd-rocm720.yml
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -44,6 +44,7 @@ on:
- nightly-8-gpu-qwen3-235b-rocm720
- nightly-8-gpu-qwen35-rocm720
- nightly-8-gpu-glm5-rocm720
- nightly-8-gpu-glm51-rocm720
- nightly-8-gpu-minimax-m25-rocm720
- nightly-1-gpu-zimage-turbo-rocm720
- nightly-test-1-gpu-mi35x-rocm720
Expand All @@ -61,7 +62,7 @@ on:
- nightly-8-gpu-mi35x-qwen3-235b-mxfp4-rocm720
- nightly-8-gpu-mi35x-qwen35-rocm720
- nightly-8-gpu-mi35x-glm5-rocm720
- nightly-8-gpu-mi35x-glm47-fp8-rocm720
- nightly-8-gpu-mi35x-glm51-rocm720
- nightly-8-gpu-mi35x-minimax-m25-rocm720
job_filter:
description: 'Or type comma-separated job names (overrides dropdown if non-empty)'
Expand Down Expand Up @@ -709,6 +710,50 @@ jobs:
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.ref }}

- 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 ${{ inputs.continue_on_error && '--continue-on-error' || '' }} || TEST_EXIT_CODE=$?
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 ${{ inputs.continue_on_error && '--continue-on-error' || '' }} || TEST_EXIT_CODE=$?
echo "$(<github_summary.md )" >> $GITHUB_STEP_SUMMARY || true
exit ${TEST_EXIT_CODE:-0}

# 8-GPU MiniMax-M2.5 (Accuracy + Performance combined) ROCm 7.2
nightly-8-gpu-minimax-m25-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-m25-rocm720,'))
Expand Down Expand Up @@ -1288,7 +1333,7 @@ jobs:
echo "$(<github_summary.md )" >> $GITHUB_STEP_SUMMARY || true
exit ${TEST_EXIT_CODE:-0}

# MI35x 8-GPU GLM-5 (Accuracy + Performance combined) ROCm 7.2
# MI35x 8-GPU GLM-5 (Accuracy only) ROCm 7.2
nightly-8-gpu-mi35x-glm5-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-rocm720,'))
runs-on: linux-mi35x-gpu-8
Expand Down Expand Up @@ -1322,20 +1367,9 @@ jobs:
echo "$(<github_summary.md )" >> $GITHUB_STEP_SUMMARY || true
exit ${TEST_EXIT_CODE:-0}

- name: Performance Test MI35x ROCm 7.2 (8-GPU GLM-5)
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 GITHUB_STEP_SUMMARY="/sglang-checkout/github_summary.md" \
python3 run_suite.py --hw amd --suite nightly-perf-8-gpu-mi35x-glm5 --nightly --timeout-per-file 5400 ${{ inputs.continue_on_error && '--continue-on-error' || '' }} || TEST_EXIT_CODE=$?
echo "$(<github_summary.md )" >> $GITHUB_STEP_SUMMARY || true
exit ${TEST_EXIT_CODE:-0}

# MI35x 8-GPU GLM-4.7-FP8 (Accuracy) ROCm 7.2
nightly-8-gpu-mi35x-glm47-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-glm47-fp8-rocm720,'))
# 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
Expand All @@ -1352,17 +1386,28 @@ jobs:

- name: Install dependencies
run: |
bash scripts/ci/amd/amd_ci_install_dependency.sh
# Install tabulate for run_suite.py (missing in MI35x container)
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 ${{ inputs.continue_on_error && '--continue-on-error' || '' }} || TEST_EXIT_CODE=$?
echo "$(<github_summary.md )" >> $GITHUB_STEP_SUMMARY || true
exit ${TEST_EXIT_CODE:-0}

- name: Accuracy Test MI35x ROCm 7.2 (8-GPU GLM-4.7-FP8)
- 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-amd-8-gpu-mi35x-glm47-fp8 --nightly --timeout-per-file 3600 ${{ inputs.continue_on_error && '--continue-on-error' || '' }} || TEST_EXIT_CODE=$?
python3 run_suite.py --hw amd --suite nightly-perf-8-gpu-mi35x-glm51 --nightly --timeout-per-file 5400 ${{ inputs.continue_on_error && '--continue-on-error' || '' }} || TEST_EXIT_CODE=$?
echo "$(<github_summary.md )" >> $GITHUB_STEP_SUMMARY || true
exit ${TEST_EXIT_CODE:-0}

Expand Down Expand Up @@ -1469,6 +1514,7 @@ jobs:
- nightly-8-gpu-qwen3-235b-rocm720
- nightly-8-gpu-qwen35-rocm720
- nightly-8-gpu-glm5-rocm720
- nightly-8-gpu-glm51-rocm720
- nightly-8-gpu-minimax-m25-rocm720
# MI30x ROCm 7.2 Diffusion Tests
- nightly-1-gpu-zimage-turbo-rocm720
Expand All @@ -1488,7 +1534,7 @@ jobs:
- nightly-8-gpu-mi35x-qwen3-235b-mxfp4-rocm720
- nightly-8-gpu-mi35x-qwen35-rocm720
- nightly-8-gpu-mi35x-glm5-rocm720
- nightly-8-gpu-mi35x-glm47-fp8-rocm720
- nightly-8-gpu-mi35x-glm51-rocm720
- nightly-8-gpu-mi35x-minimax-m25-rocm720
runs-on: ubuntu-latest
steps:
Expand Down
91 changes: 87 additions & 4 deletions .github/workflows/nightly-test-amd.yml
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -44,6 +44,8 @@ on:
- nightly-8-gpu-qwen3-235b
- nightly-8-gpu-qwen35
- nightly-8-gpu-glm5
- nightly-8-gpu-glm51
- nightly-8-gpu-glm51-mxfp4
- nightly-8-gpu-minimax-m25
- nightly-1-gpu-zimage-turbo
- nightly-test-1-gpu-mi35x
Expand All @@ -61,6 +63,8 @@ on:
- nightly-8-gpu-mi35x-qwen3-235b-mxfp4
- nightly-8-gpu-mi35x-qwen35
- nightly-8-gpu-mi35x-glm5
- nightly-8-gpu-mi35x-glm51
- nightly-8-gpu-mi35x-glm51-mxfp4
- nightly-8-gpu-mi35x-minimax-m25
job_filter:
description: 'Or type comma-separated job names (overrides dropdown if non-empty)'
Expand Down Expand Up @@ -712,6 +716,50 @@ jobs:
echo "$(<github_summary.md )" >> $GITHUB_STEP_SUMMARY || true
exit ${TEST_EXIT_CODE:-0}

# 8-GPU GLM-5.1 (Accuracy + Performance combined)
nightly-8-gpu-glm51:
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,'))
runs-on: linux-mi325-8gpu-sglang
steps:
- name: Checkout code
uses: actions/checkout@v4
with:
ref: ${{ inputs.ref || github.ref }}

- name: Setup docker
run: |
touch github_summary.md
bash scripts/ci/amd/amd_ci_start_container.sh
env:
GITHUB_WORKSPACE: ${{ github.workspace }}

- name: Install dependencies
run: |
bash scripts/ci/amd/amd_ci_install_dependency.sh
bash scripts/ci/amd/amd_ci_exec.sh pip install git+https://github.com/huggingface/transformers.git@96f807a33b75

- name: Accuracy Test (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 ${{ inputs.continue_on_error && '--continue-on-error' || '' }} || TEST_EXIT_CODE=$?
echo "$(<github_summary.md )" >> $GITHUB_STEP_SUMMARY || true
exit ${TEST_EXIT_CODE:-0}

- name: Performance Test (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 ${{ inputs.continue_on_error && '--continue-on-error' || '' }} || TEST_EXIT_CODE=$?
echo "$(<github_summary.md )" >> $GITHUB_STEP_SUMMARY || true
exit ${TEST_EXIT_CODE:-0}

# 8-GPU MiniMax-M2.5 (Accuracy + Performance combined)
nightly-8-gpu-minimax-m25:
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-m25,'))
Expand Down Expand Up @@ -1294,7 +1342,7 @@ jobs:
echo "$(<github_summary.md )" >> $GITHUB_STEP_SUMMARY || true
exit ${TEST_EXIT_CODE:-0}

# MI35x 8-GPU GLM-5 (Accuracy + Performance combined)
# MI35x 8-GPU GLM-5 (Accuracy only)
nightly-8-gpu-mi35x-glm5:
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,'))
runs-on: linux-mi35x-gpu-8
Expand Down Expand Up @@ -1328,14 +1376,47 @@ jobs:
echo "$(<github_summary.md )" >> $GITHUB_STEP_SUMMARY || true
exit ${TEST_EXIT_CODE:-0}

- name: Performance Test MI35x (8-GPU GLM-5)
# MI35x 8-GPU GLM-5.1 (Accuracy + Performance combined)
nightly-8-gpu-mi35x-glm51:
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,'))
runs-on: linux-mi35x-gpu-8
steps:
- name: Checkout code
uses: actions/checkout@v4
with:
ref: ${{ inputs.ref || github.ref }}

- name: Setup docker
run: |
touch github_summary.md
bash scripts/ci/amd/amd_ci_start_container.sh
env:
GITHUB_WORKSPACE: ${{ github.workspace }}

- name: Install dependencies
run: |
bash scripts/ci/amd/amd_ci_install_dependency.sh
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 (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 ${{ inputs.continue_on_error && '--continue-on-error' || '' }} || TEST_EXIT_CODE=$?
echo "$(<github_summary.md )" >> $GITHUB_STEP_SUMMARY || true
exit ${TEST_EXIT_CODE:-0}

- name: Performance Test MI35x (8-GPU GLM-5.1)
timeout-minutes: 120
continue-on-error: true # Perf test failure doesn't fail the job if accuracy passed
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-glm5 --nightly --timeout-per-file 5400 ${{ inputs.continue_on_error && '--continue-on-error' || '' }} || TEST_EXIT_CODE=$?
python3 run_suite.py --hw amd --suite nightly-perf-8-gpu-mi35x-glm51 --nightly --timeout-per-file 5400 ${{ inputs.continue_on_error && '--continue-on-error' || '' }} || TEST_EXIT_CODE=$?
echo "$(<github_summary.md )" >> $GITHUB_STEP_SUMMARY || true
exit ${TEST_EXIT_CODE:-0}

Expand Down Expand Up @@ -1442,6 +1523,7 @@ jobs:
- nightly-8-gpu-qwen3-235b
- nightly-8-gpu-qwen35
- nightly-8-gpu-glm5
- nightly-8-gpu-glm51
- nightly-8-gpu-minimax-m25
# MI30x Diffusion Tests
- nightly-1-gpu-zimage-turbo
Expand All @@ -1459,6 +1541,7 @@ jobs:
- nightly-8-gpu-mi35x-qwen3-235b-mxfp4
- nightly-8-gpu-mi35x-qwen35
- nightly-8-gpu-mi35x-glm5
- nightly-8-gpu-mi35x-glm51
- nightly-8-gpu-mi35x-minimax-m25
# MI35x perf jobs excluded from check - perf failures don't block CI
# - nightly-perf-8-gpu-mi35x-deepseek-v32-basic
Expand Down
Loading
Loading