minke.yu/sglang
1
0
Fork 0
Code Issues Pull Requests Actions 12 Packages Projects Releases Wiki Activity

[AMD] Test DeepSeek V4 FlashMLA backend variants nightly (#28290)

Browse Source
This commit is contained in:
Bingxu Chen
2026-06-16 01:03:26 -07:00
committed by GitHub
parent c5b9106c1a
commit 0fc2bc4a8b
5 changed files with 48 additions and 18 deletions
Download Patch File Download Diff File Expand all files Collapse all files
.github/workflows/nightly-test-amd-rocm720.yml
+32 -6
View File
@@ -1250,15 +1250,28 @@ jobs:
bash scripts/ci/amd/amd_ci_install_dependency.sh --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-Flash FP8 + FP4)
- name: Accuracy + Performance Test MI35x ROCm 7.2 (8-GPU DeepSeek-V4-Flash FP8 + FP4, unified_kv_triton)
timeout-minutes: 300
run: |
> github_summary.md # Clear summary file
echo "## SGLANG_HACK_FLASHMLA_BACKEND=unified_kv_triton" >> github_summary.md
bash scripts/ci/amd/amd_ci_exec.sh -w /sglang-checkout/test \
-e SGLANG_HACK_FLASHMLA_BACKEND=unified_kv_triton \
-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 ${{ (github.event_name == 'schedule' || inputs.continue_on_error) && '--continue-on-error' || '' }} || TEST_EXIT_CODE=$?
python3 run_suite.py --hw amd --suite nightly-amd-8-gpu-mi35x-deepseek-v4-flash --nightly --timeout-per-file 7200 ${{ (github.event_name == 'schedule' || inputs.continue_on_error) && '--continue-on-error' || '' }}
echo "$(<github_summary.md )" >> $GITHUB_STEP_SUMMARY || true
- name: Accuracy + Performance Test MI35x ROCm 7.2 (8-GPU DeepSeek-V4-Flash FP8 + FP4, triton)
if: ${{ !cancelled() }}
timeout-minutes: 300
run: |
> github_summary.md # Clear summary file
echo "## SGLANG_HACK_FLASHMLA_BACKEND=triton" >> github_summary.md
bash scripts/ci/amd/amd_ci_exec.sh -w /sglang-checkout/test \
-e SGLANG_HACK_FLASHMLA_BACKEND=triton \
-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 ${{ (github.event_name == 'schedule' || inputs.continue_on_error) && '--continue-on-error' || '' }}
echo "$(<github_summary.md )" >> $GITHUB_STEP_SUMMARY || true
exit ${TEST_EXIT_CODE:-0}
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,'))
@@ -1285,15 +1298,28 @@ jobs:
bash scripts/ci/amd/amd_ci_install_dependency.sh --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)
- name: Accuracy + Performance Test MI35x ROCm 7.2 (8-GPU DeepSeek-V4-Pro FP8 + FP4, unified_kv_triton)
timeout-minutes: 480
run: |
> github_summary.md # Clear summary file
echo "## SGLANG_HACK_FLASHMLA_BACKEND=unified_kv_triton" >> github_summary.md
bash scripts/ci/amd/amd_ci_exec.sh -w /sglang-checkout/test \
-e SGLANG_HACK_FLASHMLA_BACKEND=unified_kv_triton \
-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 ${{ (github.event_name == 'schedule' || inputs.continue_on_error) && '--continue-on-error' || '' }} || TEST_EXIT_CODE=$?
python3 run_suite.py --hw amd --suite nightly-amd-8-gpu-mi35x-deepseek-v4-pro --nightly --timeout-per-file 14400 ${{ (github.event_name == 'schedule' || inputs.continue_on_error) && '--continue-on-error' || '' }}
echo "$(<github_summary.md )" >> $GITHUB_STEP_SUMMARY || true
- name: Accuracy + Performance Test MI35x ROCm 7.2 (8-GPU DeepSeek-V4-Pro FP8 + FP4, triton)
if: ${{ !cancelled() }}
timeout-minutes: 480
run: |
> github_summary.md # Clear summary file
echo "## SGLANG_HACK_FLASHMLA_BACKEND=triton" >> github_summary.md
bash scripts/ci/amd/amd_ci_exec.sh -w /sglang-checkout/test \
-e SGLANG_HACK_FLASHMLA_BACKEND=triton \
-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 ${{ (github.event_name == 'schedule' || inputs.continue_on_error) && '--continue-on-error' || '' }}
echo "$(<github_summary.md )" >> $GITHUB_STEP_SUMMARY || true
exit ${TEST_EXIT_CODE:-0}
# ==============================================================================
# 8-GPU Kimi-K2.6 (MI30x + MI35x)