[ci] change test whl into python 312 and use test images

.github/workflows/api_eval.yml

@@ -43,7 +43,7 @@ env:
   HOST_LOCALTIME: /usr/share/zoneinfo/Asia/Shanghai
   ACTIONS_ALLOW_USE_UNSECURE_NODE_VERSION: true
   REPORT_DIR: /nvme/qa_test_models/evaluation_report/allure_report/${{ inputs.repo_ref }}_${{ github.run_id }}
-  COV_PARAM: --cov /opt/py3/lib/python3.10/site-packages/lmdeploy
+  COV_PARAM: --cov /opt/py3/lib/python3.12/site-packages/lmdeploy
   TEST_CODE_PATH: /nvme/qa_test_models/test_pkg/lmdeploy/${{ inputs.repo_ref }}_${{ github.run_id }}
   OFFLINE_CODE_PATH: /nvme/qa_test_models/offline_pkg/lmdeploy
   COMPASS_DATA_CACHE: /nvme/qa_test_models/compass_data_cache
@@ -58,7 +58,7 @@ jobs:
     if: ${{github.event_name == 'schedule' || (!cancelled() && !inputs.offline_mode)}}
     strategy:
       matrix:
-        pyver: [py310]
+        pyver: [py312]
     runs-on: ubuntu-latest
     env:
       PYTHON_VERSION: ${{ matrix.pyver }}
@@ -132,7 +132,7 @@ jobs:
         if: ${{github.event_name == 'schedule' || !inputs.offline_mode}}
         uses: actions/download-artifact@v4
         with:
-          name: my-artifact-${{ github.run_id }}-py310
+          name: my-artifact-${{ github.run_id }}-py312
       - name: Copy Artifacts
         if: ${{github.event_name == 'schedule' || !inputs.offline_mode}}
         run: rm ${{env.TEST_CODE_PATH}}/lmdeploy-*.whl -f && cp lmdeploy-*.whl ${{env.TEST_CODE_PATH}}

.github/workflows/benchmark.yml

@@ -28,6 +28,16 @@ on:
         description: 'Whether start a offline mode, if true, you should prepare code and whl package by yourself'
         type: boolean
         default: false
+      docker_tag:
+        required: true
+        description: 'Docker tag'
+        type: string
+        default: 'latest-cu12.8'
+      result_tag:
+        required: true
+        description: 'result_tag if is not none, benchmark results will be uploaded to feishu'
+        type: string
+        default: "default"
 
 env:
   HOST_PIP_CACHE_DIR: /nvme/github-actions/pip-cache
@@ -45,7 +55,7 @@ jobs:
     if: ${{github.event_name == 'schedule' || (!cancelled() && !inputs.offline_mode)}}
     strategy:
       matrix:
-        pyver: [py310]
+        pyver: [py312]
     runs-on: ubuntu-latest
     env:
       PYTHON_VERSION: ${{ matrix.pyver }}
@@ -93,7 +103,7 @@ jobs:
     runs-on: [self-hosted, linux-a100]
     timeout-minutes: 50
     container:
-      image: openmmlab/lmdeploy:latest-cu12.8
+      image: openmmlab/lmdeploy:${{ inputs.docker_tag || 'latest-cu12.8' }}
       options: "--gpus=all --ipc=host --user root -e PIP_CACHE_DIR=/root/.cache/pip -e NVIDIA_DISABLE_REQUIRE=1 --pull never"
       volumes:
         - /nvme/qa_test_models:/nvme/qa_test_models
@@ -117,7 +127,7 @@ jobs:
         if: ${{github.event_name == 'schedule' || !inputs.offline_mode}}
         uses: actions/download-artifact@v4
         with:
-          name: my-artifact-${{ github.run_id }}-py310
+          name: my-artifact-${{ github.run_id }}-py312
       - name: Copy Artifacts
         if: ${{github.event_name == 'schedule' || !inputs.offline_mode}}
         run: rm ${{env.TEST_CODE_PATH}}/lmdeploy-*.whl -f && cp lmdeploy-*.whl ${{env.TEST_CODE_PATH}}
@@ -153,7 +163,7 @@ jobs:
       TEST_ENV: ${{ matrix.transformers }}
     timeout-minutes: 480
     container:
-      image: openmmlab/lmdeploy:latest-cu12.8
+      image: openmmlab/lmdeploy:${{ inputs.docker_tag || 'latest-cu12.8' }}
       options: "--gpus=all --ipc=host --user root -e PIP_CACHE_DIR=/root/.cache/pip -e NVIDIA_DISABLE_REQUIRE=1 --pull never"
       volumes:
         - /nvme/github-actions/pip-cache:/root/.cache/pip
@@ -197,11 +207,25 @@ jobs:
         if: contains(fromJson(github.event.inputs.backend), 'pytorch') && !contains(fromJson(github.event.inputs.backend), 'turbomind')
         run: |
             pytest autotest/benchmark/test_${{matrix.benchmark_type}}_performance.py -n ${{matrix.n}} -m '${{matrix.gpu_num}} and not pr_test and not function and pytorch' --alluredir=${{env.ALLURE_REPORT_DIR}}
+      - name: Generate result
+        if: always()
+        run: |
+            cd /nvme/qa_test_models/feishu_upload
+            python3 test_benchmark.py --root ${{env.REPORT_DIR}} --output ${{env.REPORT_DIR}}/${{inputs.result_tag}}.txt --hardware A100 --infer-version ${{inputs.result_tag}}
+      - name: Async result
+        if: always() && inputs.result_tag != 'default'
+        env:
+          FEISHU_APP_ID: ${{secrets.FEISHU_APP_ID}}
+          FEISHU_APP_SECRET: ${{secrets.FEISHU_APP_SECRET}}
+          FEISHU_TABLE_TOKEN: ${{secrets.FEISHU_TABLE_TOKEN}}
+          FEISHU_TABLE_ID: ${{secrets.BENCHMARK_FEISHU_TABLE_ID}}
+        run: |
+            cd /nvme/qa_test_models/feishu_upload
+            python3 main.py --skip-duplicates ${{env.REPORT_DIR}}/${{inputs.result_tag}}.txt --config config-benchmark.py
       - name: Clear workfile
         if: always()
         run: |
           echo "status=done" >> ${{env.REPORT_DIR}}/status.txt
-          chmod -R 777 $REPORT_DIR
           export workdir=$(pwd)
           cd ..
           rm -rf $workdir

.github/workflows/daily_ete_test.yml

@@ -38,6 +38,11 @@ on:
         description: 'regression functions'
         type: string
         default: "['quant', 'tools','restful','pipeline','benchmark','evaluation']"
+      docker_tag:
+        required: true
+        description: 'Docker tag'
+        type: string
+        default: 'nightly-test-cu12.8'
   schedule:
     - cron:  '00 14 * * 0-4'
 
@@ -48,7 +53,7 @@ env:
   ACTIONS_ALLOW_USE_UNSECURE_NODE_VERSION: true
   ROOT_DIR: /nvme/qa_test_models
   REPORT_DIR: /nvme/qa_test_models/test-reports/${{ inputs.repo_ref || 'main' }}_${{ github.run_id }}
-  COV_PARAM: --cov /opt/py3/lib/python3.10/site-packages/lmdeploy
+  COV_PARAM: --cov /opt/py3/lib/python3.12/site-packages/lmdeploy
   TEST_CODE_PATH: /nvme/qa_test_models/test_pkg/lmdeploy/${{ inputs.repo_ref || 'main' }}_${{ github.run_id }}
   OFFLINE_CODE_PATH: /nvme/qa_test_models/offline_pkg/lmdeploy
   OFFLINE_REQUIREMENTS: /nvme/qa_test_models/offline_pkg/requirements.txt
@@ -60,7 +65,7 @@ jobs:
     if: ${{!cancelled() && (github.event_name == 'schedule' || !inputs.offline_mode)}}
     strategy:
       matrix:
-        pyver: [py310]
+        pyver: [py312]
     runs-on: ubuntu-latest
     env:
       PYTHON_VERSION: ${{ matrix.pyver }}
@@ -109,7 +114,7 @@ jobs:
     runs-on: [self-hosted, linux-a100]
     timeout-minutes: 50
     container:
-      image: openmmlab/lmdeploy:latest-cu12.8
+      image: openmmlab/lmdeploy:${{ inputs.docker_tag || 'nightly-test-cu12.8' }}
       options: "--gpus=all --ipc=host --user root -e PIP_CACHE_DIR=/root/.cache/pip -e NVIDIA_DISABLE_REQUIRE=1 --pull never"
       volumes:
         - /nvme/qa_test_models:/nvme/qa_test_models
@@ -131,7 +136,7 @@ jobs:
         if: ${{github.event_name == 'schedule' || !inputs.offline_mode}}
         uses: actions/download-artifact@v4
         with:
-          name: my-artifact-${{ github.run_id }}-py310
+          name: my-artifact-${{ github.run_id }}-py312
       - name: Copy Artifacts
         if: ${{github.event_name == 'schedule' || !inputs.offline_mode}}
         run: rm ${{env.TEST_CODE_PATH}}/lmdeploy-*.whl -f && cp lmdeploy-*.whl ${{env.TEST_CODE_PATH}}
@@ -140,9 +145,13 @@ jobs:
         run: rm ${{env.TEST_CODE_PATH}}/lmdeploy-*.whl -f && cp ${{env.OFFLINE_CODE_PATH}}/lmdeploy-*.whl ${{env.TEST_CODE_PATH}}
       - name: Mark as start
         run: |
-          chmod -R 777 ${{env.TEST_CODE_PATH}}
           mkdir ${{env.REPORT_DIR}} -p
           echo "starttime=$(date +%s)" > ${{env.REPORT_DIR}}/status.txt
+      - name: Clear workfile
+        if: always()
+        run: |
+          chmod -R 777 ${{env.TEST_CODE_PATH}}
+          chmod -R 777 ${{env.REPORT_DIR}}
 
   test_quantization:
     needs: download_pkgs
@@ -158,7 +167,7 @@ jobs:
       MODELSCOPE_MODULES_CACHE: /nvme/qa_test_models/modelscope_modules
       TEST_ENV: ${{ matrix.transformers }}
     container:
-      image: openmmlab/lmdeploy:latest-cu12.8
+      image: openmmlab/lmdeploy:${{ inputs.docker_tag || 'nightly-test-cu12.8' }}
       options: "--gpus=all --ipc=host --user root -e PIP_CACHE_DIR=/root/.cache/pip -e NVIDIA_DISABLE_REQUIRE=1 --pull never"
       volumes:
         - /nvme/github-actions/pip-cache:/root/.cache/pip
@@ -177,7 +186,7 @@ jobs:
           echo "starttime=$(date +%s)" > ${{env.REPORT_DIR}}/status.txt
       - name: Install lmdeploy - dependency
         run: |
-          python3 -m pip install auto_gptq matplotlib attrdict
+          python3 -m pip install matplotlib attrdict
           python3 -m pip install -r requirements/lite.txt
       - name: Install lmdeploy
         run: |
@@ -210,7 +219,6 @@ jobs:
         if: always()
         run: |
           echo "status=done" >> ${{env.REPORT_DIR}}/status.txt
-          chmod -R 777 ${{env.ROOT_DIR}}
           export workdir=$(pwd)
           cd ..
           rm -rf $workdir
@@ -246,7 +254,7 @@ jobs:
       MODELSCOPE_MODULES_CACHE: /nvme/qa_test_models/modelscope_modules
       TEST_ENV: ${{ matrix.transformers }}
     container:
-      image: openmmlab/lmdeploy:latest-cu12.8
+      image: openmmlab/lmdeploy:${{ inputs.docker_tag || 'nightly-test-cu12.8' }}
       options: "--gpus=all --ipc=host --user root -e PIP_CACHE_DIR=/root/.cache/pip -e NVIDIA_DISABLE_REQUIRE=1 --pull never"
       volumes:
         - /nvme/github-actions/pip-cache:/root/.cache/pip
@@ -330,7 +338,6 @@ jobs:
         if: always()
         run: |
           echo "status=done" >> ${{env.REPORT_DIR}}/status.txt
-          chmod -R 777 ${{env.ROOT_DIR}}
           export workdir=$(pwd)
           cd ..
           rm -rf $workdir
@@ -347,6 +354,36 @@ jobs:
         backend: ${{ fromJSON(inputs.backend || '["turbomind", "pytorch"]')}}
         model_path: ['Qwen/Qwen3-8B-Base', 'Qwen/Qwen3-30B-A3B', 'Qwen/Qwen3-32B', 'OpenGVLab/InternVL3_5-30B-A3B', 'OpenGVLab/InternVL3-38B', 'Qwen/Qwen3-VL-8B-Instruct', 'Qwen/Qwen3-VL-30B-A3B-Instruct']
         include:
+          - tp: 2
+            model: Qwen3.5-35B-A3B
+            model_path: Qwen/Qwen3.5-35B-A3B
+            case_info: ['chat_completions_v1', 'generate']
+            generate_type: all
+            extra: '--logprobs-mode raw_logprobs --enable-return-routed-experts'
+            backend: pytorch
+          - tp: 2
+            model: Qwen3.5-35B-A3B
+            model_path: Qwen/Qwen3.5-35B-A3B
+            case_info: ['chat_completions_v1', 'generate']
+            generate_type: logprob
+            extra: '--logprobs-mode raw_logprobs'
+            backend: turbomind
+          - tp: 2
+            model: Qwen3.5-27B
+            model_path: Qwen/Qwen3.5-27B
+            case_info: ['chat_completions_v1', 'generate']
+            generate_type: logprob
+            extra: '--logprobs-mode raw_logprobs'
+          - tp: 2
+            model: Qwen3.5-35B-A3B-Base
+            model_path: Qwen/Qwen3.5-35B-A3B-Base
+            case_info: ['completions_v1']
+            generate_type: base
+          - tp: 1
+            model: Qwen3.5-2B-Base
+            model_path: Qwen/Qwen3.5-2B-Base
+            case_info: ['completions_v1']
+            generate_type: base
           - tp: 2
             model: Qwen3-8B-Base
             model_path: Qwen/Qwen3-8B-Base
@@ -422,7 +459,7 @@ jobs:
             extra: '--logprobs-mode raw_logprobs'
     timeout-minutes: 60
     container:
-      image: openmmlab/lmdeploy:latest-cu12.8
+      image: openmmlab/lmdeploy:${{ inputs.docker_tag || 'nightly-test-cu12.8' }}
       options: "--gpus=all --ipc=host --user root -e PIP_CACHE_DIR=/root/.cache/pip -e NVIDIA_DISABLE_REQUIRE=1 --pull never"
       volumes:
         - /nvme/github-actions/pip-cache:/root/.cache/pip
@@ -527,7 +564,6 @@ jobs:
         if: always()
         run: |
           echo "status=done" >> ${{env.REPORT_DIR}}/status.txt
-          chmod -R 777 ${{env.ROOT_DIR}}
           export workdir=$(pwd)
           cd ..
           rm -rf $workdir
@@ -540,7 +576,7 @@ jobs:
     needs: test_quantization
     timeout-minutes: 240
     container:
-      image: openmmlab/lmdeploy:latest-cu12.8
+      image: openmmlab/lmdeploy:${{ inputs.docker_tag || 'nightly-test-cu12.8' }}
       options: "--gpus=all --ipc=host --user root -e PIP_CACHE_DIR=/root/.cache/pip -e NVIDIA_DISABLE_REQUIRE=1 --pull never"
       volumes:
         - /nvme/github-actions/pip-cache:/root/.cache/pip
@@ -590,7 +626,6 @@ jobs:
         if: always()
         run: |
           echo "status=done" >> ${{env.REPORT_DIR}}/status.txt
-          chmod -R 777 ${{env.ROOT_DIR}}
           export workdir=$(pwd)
           cd ..
           rm -rf $workdir
@@ -604,7 +639,7 @@ jobs:
     needs: test_quantization
     timeout-minutes: 120
     container:
-      image: openmmlab/lmdeploy:latest-cu12.8
+      image: openmmlab/lmdeploy:${{ inputs.docker_tag || 'nightly-test-cu12.8' }}
       options: "--gpus=all --ipc=host --user root -e PIP_CACHE_DIR=/root/.cache/pip -e NVIDIA_DISABLE_REQUIRE=1 --pull never"
       volumes:
         - /nvme/github-actions/pip-cache:/root/.cache/pip
@@ -646,7 +681,6 @@ jobs:
         if: always()
         run: |
           echo "status=done" >> ${{env.REPORT_DIR}}/status.txt
-          chmod -R 777 ${{env.ROOT_DIR}}
           export workdir=$(pwd)
           cd ..
           rm -rf $workdir
@@ -671,7 +705,7 @@ jobs:
             generate_type: base
     timeout-minutes: 60
     container:
-      image: openmmlab/lmdeploy:latest-cu12.8
+      image: openmmlab/lmdeploy:${{ inputs.docker_tag || 'nightly-test-cu12.8' }}
       options: "--gpus=all --ipc=host --user root -e PIP_CACHE_DIR=/root/.cache/pip -e NVIDIA_DISABLE_REQUIRE=1 --pull never"
       volumes:
         - /nvme/github-actions/pip-cache:/root/.cache/pip
@@ -777,7 +811,6 @@ jobs:
         if: always()
         run: |
           echo "status=done" >> ${{env.REPORT_DIR}}/status.txt
-          chmod -R 777 ${{env.ROOT_DIR}}
           export workdir=$(pwd)
           cd ..
           rm -rf $workdir
@@ -790,7 +823,7 @@ jobs:
     needs: test_quantization
     timeout-minutes: 240
     container:
-      image: openmmlab/lmdeploy:latest-cu12.8
+      image: openmmlab/lmdeploy:${{ inputs.docker_tag || 'nightly-test-cu12.8' }}
       options: "--gpus=all --ipc=host --user root -e PIP_CACHE_DIR=/root/.cache/pip -e NVIDIA_DISABLE_REQUIRE=1 --pull never"
       volumes:
         - /nvme/github-actions/pip-cache:/root/.cache/pip
@@ -841,7 +874,6 @@ jobs:
         if: always()
         run: |
           echo "status=done" >> ${{env.REPORT_DIR}}/status.txt
-          chmod -R 777 ${{env.ROOT_DIR}}
           export workdir=$(pwd)
           cd ..
           rm -rf $workdir
@@ -854,7 +886,7 @@ jobs:
     needs: [test_tools, test_restful, test_pipeline, test_benchmark]
     timeout-minutes: 5
     container:
-      image: openmmlab/lmdeploy:latest-cu12.8
+      image: openmmlab/lmdeploy:${{ inputs.docker_tag || 'nightly-test-cu12.8' }}
       options: "--gpus=all --ipc=host --user root -e PIP_CACHE_DIR=/root/.cache/pip -e NVIDIA_DISABLE_REQUIRE=1 --pull never"
       volumes:
         - /nvme/github-actions/pip-cache:/root/.cache/pip
@@ -866,7 +898,6 @@ jobs:
         run: cp -r ${{env.TEST_CODE_PATH}}/. .
       - name: Install lmdeploy
         run: |
-          echo "status=done" >> ${{env.REPORT_DIR}}/status.txt
           python3 -m pip uninstall lmdeploy -y && python3 -m pip install lmdeploy-*.whl --no-deps
           python3 -m pip install -r requirements/test.txt
       - name: Get coverage report
@@ -879,7 +910,7 @@ jobs:
       - name: Clear workfile
         if: always()
         run: |
-          chmod -R 777 ${{env.ROOT_DIR}}
+          chmod -R 777 ${{env.REPORT_DIR}}
           export workdir=$(pwd)
           cd ..
           rm -rf $workdir