[AMD] Add DeepSeek-V3.2 and VLMs model in nightly tests (#17179)

Co-authored-by: michaelzhang-ai <michaelzhang-ai@users.noreply.github.com> Co-authored-by: YC Tseng <yctseng@amd.com> Co-authored-by: Bingxu Chen <bingxche@amd.com>
2026-01-19 20:31:56 -08:00
parent 6988a0f570
commit a3addd6203
18 changed files with 1139 additions and 98 deletions
--- a/.github/workflows/nightly-test-amd.yml
+++ b/.github/workflows/nightly-test-amd.yml
@@ -22,6 +22,8 @@ on:
          # MI30x Accuracy Tests (GSM8K / MMMU)
          - 'nightly-accuracy-2-gpu'
          - 'nightly-accuracy-2-gpu-vlm'
+          - 'nightly-perf-2-gpu-text'
+          - 'nightly-perf-2-gpu-vlm'
          - 'nightly-accuracy-8-gpu'
          - 'nightly-accuracy-8-gpu-deepseek-r1'
          # MI30x Accuracy + Performance Tests (combined)
@@ -31,10 +33,12 @@ on:
          # MI35x jobs
          - 'nightly-test-1-gpu-mi35x'
          - 'nightly-accuracy-8-gpu-mi35x'
-          - 'nightly-accuracy-8-gpu-mi35x-deepseek-r1'
          - 'nightly-8-gpu-mi35x-grok1-int4'
          - 'nightly-8-gpu-mi35x-grok2'
          - 'nightly-8-gpu-mi35x-deepseek-r1-mxfp4'
+          - 'nightly-accuracy-8-gpu-mi35x-deepseek-v32'
+          - 'nightly-perf-8-gpu-mi35x-deepseek-v32-basic'
+          - 'nightly-perf-8-gpu-mi35x-deepseek-v32-mtp'
  workflow_call:
    inputs:
      ref:
@@ -106,6 +110,7 @@ jobs:

      - name: Nightly Test (2-GPU)
        run: |
+          > github_summary.md  # Clear summary file
          bash scripts/ci/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 || TEST_EXIT_CODE=$?
@@ -135,12 +140,75 @@ jobs:
      - name: Nightly Accuracy Test (2-GPU VLM MMMU)
        timeout-minutes: 180
        run: |
+          > github_summary.md  # Clear summary file
          bash scripts/ci/amd_ci_exec.sh -w /sglang-checkout/test \
            -e GITHUB_STEP_SUMMARY="/sglang-checkout/github_summary.md" \
            python3 run_suite.py --hw amd --suite nightly-amd-accuracy-2-gpu-vlm --nightly --timeout-per-file 7200 || TEST_EXIT_CODE=$?
          echo "$(<github_summary.md )" >> $GITHUB_STEP_SUMMARY || true
          exit ${TEST_EXIT_CODE:-0}

+  # 2-GPU Text Models Performance Tests
+  nightly-perf-2-gpu-text:
+    if: (github.repository == 'sgl-project/sglang' || github.event_name == 'pull_request') && (inputs.job_filter == '' || inputs.job_filter == 'all' || inputs.job_filter == 'nightly-perf-2-gpu-text')
+    runs-on: linux-mi325-gpu-2
+    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_ci_start_container.sh
+        env:
+          GITHUB_WORKSPACE: ${{ github.workspace }}
+
+      - name: Install dependencies
+        run: bash scripts/ci/amd_ci_install_dependency.sh
+
+      - name: Performance Test (2-GPU Text Models)
+        timeout-minutes: 120
+        run: |
+          > github_summary.md  # Clear summary file
+          bash scripts/ci/amd_ci_exec.sh -w /sglang-checkout/test \
+            -e SGLANG_USE_AITER=1 \
+            -e GITHUB_STEP_SUMMARY="/sglang-checkout/github_summary.md" \
+            python3 run_suite.py --hw amd --suite nightly-amd-perf-text-2-gpu --nightly --timeout-per-file 3600 || TEST_EXIT_CODE=$?
+          echo "$(<github_summary.md )" >> $GITHUB_STEP_SUMMARY || true
+          exit ${TEST_EXIT_CODE:-0}
+
+  # 2-GPU VLM Performance Tests
+  nightly-perf-2-gpu-vlm:
+    if: (github.repository == 'sgl-project/sglang' || github.event_name == 'pull_request') && (inputs.job_filter == '' || inputs.job_filter == 'all' || inputs.job_filter == 'nightly-perf-2-gpu-vlm')
+    runs-on: linux-mi325-gpu-2
+    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_ci_start_container.sh
+        env:
+          GITHUB_WORKSPACE: ${{ github.workspace }}
+
+      - name: Install dependencies
+        run: bash scripts/ci/amd_ci_install_dependency.sh
+
+      - name: Performance Test (2-GPU VLM Models)
+        timeout-minutes: 180
+        run: |
+          > github_summary.md  # Clear summary file
+          bash scripts/ci/amd_ci_exec.sh -w /sglang-checkout/test \
+            -e SGLANG_USE_AITER=1 \
+            -e GITHUB_STEP_SUMMARY="/sglang-checkout/github_summary.md" \
+            python3 run_suite.py --hw amd --suite nightly-amd-perf-vlm-2-gpu --nightly --timeout-per-file 7200 || TEST_EXIT_CODE=$?
+          echo "$(<github_summary.md )" >> $GITHUB_STEP_SUMMARY || true
+          exit ${TEST_EXIT_CODE:-0}
+
  # 8-GPU Accuracy Tests - GPT-OSS, Grok1-FP8 (accuracy only)
  nightly-accuracy-8-gpu:
    if: (github.repository == 'sgl-project/sglang' || github.event_name == 'pull_request') && (inputs.job_filter == '' || inputs.job_filter == 'all' || inputs.job_filter == 'nightly-accuracy-8-gpu')
@@ -404,38 +472,6 @@ jobs:
          echo "$(<github_summary.md )" >> $GITHUB_STEP_SUMMARY || true
          exit ${TEST_EXIT_CODE:-0}

-  # MI35x 8-GPU DeepSeek-R1-0528 Accuracy Test (separate job due to long loading time)
-  nightly-accuracy-8-gpu-mi35x-deepseek-r1:
-    if: (github.repository == 'sgl-project/sglang' || github.event_name == 'pull_request') && (inputs.job_filter == '' || inputs.job_filter == 'all' || inputs.job_filter == 'nightly-accuracy-8-gpu-mi35x-deepseek-r1')
-    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_ci_start_container.sh
-        env:
-          GITHUB_WORKSPACE: ${{ github.workspace }}
-
-      - name: Install dependencies
-        run: |
-          bash scripts/ci/amd_ci_install_dependency.sh
-          # Install tabulate for run_suite.py (missing in MI35x container)
-          bash scripts/ci/amd_ci_exec.sh pip install tabulate
-
-      - name: Accuracy Test MI35x (8-GPU DeepSeek-R1-0528)
-        timeout-minutes: 240
-        run: |
-          bash scripts/ci/amd_ci_exec.sh -w /sglang-checkout/test \
-            -e GITHUB_STEP_SUMMARY="/sglang-checkout/github_summary.md" \
-            python3 run_suite.py --hw amd --suite nightly-amd-accuracy-8-gpu-mi35x-deepseek-r1 --nightly --timeout-per-file 7200 || TEST_EXIT_CODE=$?
-          echo "$(<github_summary.md )" >> $GITHUB_STEP_SUMMARY || true
-          exit ${TEST_EXIT_CODE:-0}
-
  # MI35x 8-GPU Grok1-INT4 (Accuracy + Performance combined)
  nightly-8-gpu-mi35x-grok1-int4:
    if: (github.repository == 'sgl-project/sglang' || github.event_name == 'pull_request') && (inputs.job_filter == '' || inputs.job_filter == 'all' || inputs.job_filter == 'nightly-8-gpu-mi35x-grok1-int4')
@@ -572,6 +608,105 @@ jobs:
          echo "$(<github_summary.md )" >> $GITHUB_STEP_SUMMARY || true
          exit ${TEST_EXIT_CODE:-0}

+  # MI35x 8-GPU DeepSeek-V3.2 Accuracy Test
+  nightly-accuracy-8-gpu-mi35x-deepseek-v32:
+    if: (github.repository == 'sgl-project/sglang' || github.event_name == 'pull_request') && (inputs.job_filter == '' || inputs.job_filter == 'all' || inputs.job_filter == 'nightly-accuracy-8-gpu-mi35x-deepseek-v32')
+    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_ci_start_container.sh
+        env:
+          GITHUB_WORKSPACE: ${{ github.workspace }}
+
+      - name: Install dependencies
+        run: |
+          bash scripts/ci/amd_ci_install_dependency.sh
+          # Install tabulate for run_suite.py (missing in MI35x container)
+          bash scripts/ci/amd_ci_exec.sh pip install tabulate
+
+      - name: Accuracy Test MI35x (8-GPU DeepSeek-V3.2)
+        timeout-minutes: 120
+        run: |
+          > github_summary.md  # Clear summary file
+          bash scripts/ci/amd_ci_exec.sh -w /sglang-checkout/test \
+            -e GITHUB_STEP_SUMMARY="/sglang-checkout/github_summary.md" \
+            python3 run_suite.py --hw amd --suite nightly-amd-8-gpu-mi35x-deepseek-v32 --nightly --timeout-per-file 3600 || TEST_EXIT_CODE=$?
+          echo "$(<github_summary.md )" >> $GITHUB_STEP_SUMMARY || true
+          exit ${TEST_EXIT_CODE:-0}
+
+  # MI35x 8-GPU DeepSeek-V3.2 Performance Test (Basic)
+  nightly-perf-8-gpu-mi35x-deepseek-v32-basic:
+    if: (github.repository == 'sgl-project/sglang' || github.event_name == 'pull_request') && (inputs.job_filter == '' || inputs.job_filter == 'all' || inputs.job_filter == 'nightly-perf-8-gpu-mi35x-deepseek-v32-basic')
+    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_ci_start_container.sh
+        env:
+          GITHUB_WORKSPACE: ${{ github.workspace }}
+
+      - name: Install dependencies
+        run: |
+          bash scripts/ci/amd_ci_install_dependency.sh
+          # Install tabulate for run_suite.py (missing in MI35x container)
+          bash scripts/ci/amd_ci_exec.sh pip install tabulate
+
+      - name: Performance Test MI35x (8-GPU DeepSeek-V3.2 Basic)
+        timeout-minutes: 150
+        run: |
+          > github_summary.md  # Clear summary file
+          bash scripts/ci/amd_ci_exec.sh -w /sglang-checkout/test \
+            -e GITHUB_STEP_SUMMARY="/sglang-checkout/github_summary.md" \
+            python3 run_suite.py --hw amd --suite nightly-perf-8-gpu-mi35x-deepseek-v32-basic --nightly --timeout-per-file 5400 || TEST_EXIT_CODE=$?
+          echo "$(<github_summary.md )" >> $GITHUB_STEP_SUMMARY || true
+          exit ${TEST_EXIT_CODE:-0}
+
+  # MI35x 8-GPU DeepSeek-V3.2 Performance Test (MTP)
+  nightly-perf-8-gpu-mi35x-deepseek-v32-mtp:
+    if: (github.repository == 'sgl-project/sglang' || github.event_name == 'pull_request') && (inputs.job_filter == '' || inputs.job_filter == 'all' || inputs.job_filter == 'nightly-perf-8-gpu-mi35x-deepseek-v32-mtp')
+    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_ci_start_container.sh
+        env:
+          GITHUB_WORKSPACE: ${{ github.workspace }}
+
+      - name: Install dependencies
+        run: |
+          bash scripts/ci/amd_ci_install_dependency.sh
+          # Install tabulate for run_suite.py (missing in MI35x container)
+          bash scripts/ci/amd_ci_exec.sh pip install tabulate
+
+      - name: Performance Test MI35x (8-GPU DeepSeek-V3.2 MTP)
+        timeout-minutes: 150
+        run: |
+          > github_summary.md  # Clear summary file
+          bash scripts/ci/amd_ci_exec.sh -w /sglang-checkout/test \
+            -e GITHUB_STEP_SUMMARY="/sglang-checkout/github_summary.md" \
+            python3 run_suite.py --hw amd --suite nightly-perf-8-gpu-mi35x-deepseek-v32-mtp --nightly --timeout-per-file 5400 || TEST_EXIT_CODE=$?
+          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:
@@ -580,6 +715,9 @@ jobs:
      # MI30x Accuracy Tests
      - nightly-accuracy-2-gpu
      - nightly-accuracy-2-gpu-vlm
+      # MI30x Performance Tests
+      - nightly-perf-2-gpu-text
+      - nightly-perf-2-gpu-vlm
      - nightly-accuracy-8-gpu
      - nightly-accuracy-8-gpu-deepseek-r1
      # MI30x Combined Accuracy + Performance Tests
@@ -589,10 +727,12 @@ jobs:
      # MI35x jobs
      - nightly-test-1-gpu-mi35x
      - nightly-accuracy-8-gpu-mi35x
-      - nightly-accuracy-8-gpu-mi35x-deepseek-r1
      - nightly-8-gpu-mi35x-grok1-int4
      - nightly-8-gpu-mi35x-grok2
      - nightly-8-gpu-mi35x-deepseek-r1-mxfp4
+      - nightly-accuracy-8-gpu-mi35x-deepseek-v32
+      - nightly-perf-8-gpu-mi35x-deepseek-v32-basic
+      - nightly-perf-8-gpu-mi35x-deepseek-v32-mtp
    runs-on: ubuntu-latest
    steps:
      - name: Check if any job failed