[CI] CI refactor (#4928)

1. rename workflow to better name
2. fix lint error
3. remove accuracy report doc and test

- vLLM version: v0.12.0
- vLLM main:
ad32e3e19c

Signed-off-by: wangxiyuan <wangxiyuan1007@gmail.com>

.github/workflows/schedule_test_vllm_main.yaml

@@ -17,9 +17,9 @@
 name: 'ascend test / vllm main'
 
 on:
-  # Run full e2e tests per 2h
+  # Run full e2e tests per 4h
   schedule:
-    - cron: '0 */2 * * *'
+    - cron: '0 */4 * * *'
   workflow_dispatch:
 
 # Bash shells do not use ~/.profile or ~/.bashrc so these shells need to be explicitly

.github/workflows/vllm_ascend_test_report.yaml

@@ -1,172 +0,0 @@
-#
-# Copyright (c) 2025 Huawei Technologies Co., Ltd. All Rights Reserved.
-#
-# Licensed under the Apache License, Version 2.0 (the "License");
-# you may not use this file except in compliance with the License.
-# You may obtain a copy of the License at
-#
-#     http://www.apache.org/licenses/LICENSE-2.0
-#
-# Unless required by applicable law or agreed to in writing, software
-# distributed under the License is distributed on an "AS IS" BASIS,
-# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
-# See the License for the specific language governing permissions and
-# limitations under the License.
-# This file is a part of the vllm-ascend project.
-#
-
-# This test will be triggered:
-# 1. schedule
-# 2. pull_request change the related files
-# 3. workflow_dispatch with models input
-
-name: ascend test / accuracy report
-
-on:
-  pull_request:
-    branches:
-      - 'main'
-      - '*-dev'
-    paths:
-      - '.github/workflows/vllm_ascend_test_report.yaml'
-      - 'tests/e2e/models/test_lm_eval_correctness.py'
-  workflow_dispatch:
-    inputs:
-      vllm-ascend-version:
-        description: 'vllm-ascend:'
-        required: true
-        type: choice
-        # Current supported vLLM versions
-        options:
-          - latest
-          - main
-        default: main
-
-# Bash shells do not use ~/.profile or ~/.bashrc so these shells need to be explicitly
-# declared as "shell: bash -el {0}" on steps that need to be properly activated.
-# It's used to activate ascend-toolkit environment variables.
-defaults:
-  run:
-    shell: bash -el {0}
-
-# only cancel in-progress runs of the same workflow
-concurrency:
-  group: ${{ github.workflow }}-${{ github.ref }}
-  cancel-in-progress: true
-
-jobs:
-  run:
-    strategy:
-      fail-fast: false
-      matrix:
-        include:
-          - runner: linux-aarch64-a2-1
-            model_list:
-              - Qwen3-8B
-              - Qwen2.5-VL-7B-Instruct
-              - Qwen2-Audio-7B-Instruct
-          - runner: linux-aarch64-a2-2
-            model_list:
-              - Qwen3-30B-A3B
-              - Qwen3-VL-30B-A3B-Instruct
-              - DeepSeek-V2-Lite
-    uses: ./.github/workflows/_e2e_nightly_single_node_models.yaml
-    with:
-      vllm: v0.12.0
-      runner: ${{ matrix.runner }}
-      image: swr.cn-southwest-2.myhuaweicloud.com/base_image/ascend-ci/cann:8.3.rc2-910b-ubuntu22.04-py3.11
-      model_list: ${{ toJson(matrix.model_list) }}
-      upload: ${{ github.event_name == 'workflow_dispatch' && github.event.inputs.vllm-ascend-version == 'latest' }}
-
-  create_pr:
-    runs-on: ubuntu-latest
-    needs: run
-    if: ${{ github.event_name == 'workflow_dispatch' && github.event.inputs.vllm-ascend-version == 'latest' }}
-    env:
-      UPSTREAM_REPO: vllm-project/vllm-ascend
-    steps:
-      - name: Checkout repository
-        uses: actions/checkout@v6.0.1
-        with:
-          repository: vllm-ascend-ci/vllm-ascend
-          token: ${{ secrets.PAT_TOKEN }}
-          ref: main
-      
-      - name: Add upstream remote
-        run: |
-          git remote add upstream https://github.com/${{ env.UPSTREAM_REPO }}.git
-          git fetch upstream
-          git remote -v
-
-      - name: Set Git user info dynamically
-        run: |
-          git config user.name "${{ github.actor }}"
-          git config user.email "${{ github.actor }}@users.noreply.github.com"
-
-      - name: Create or switch to branch
-        run: |
-          TIMESTAMP=$(date +%Y%m%d%H%M%S)
-          BRANCH_NAME="auto-pr/accuracy-report-${TIMESTAMP}"
-          echo "BRANCH_NAME=${BRANCH_NAME}" >> $GITHUB_ENV
-          git checkout -B "${BRANCH_NAME}" upstream/main
-
-      - name: Download only current run reports
-        uses: actions/download-artifact@v6
-        with:
-          path: ./docs/source/developer_guide/evaluation/accuracy_report
-          pattern: report-*
-          github-token: ${{ secrets.GITHUB_TOKEN }}
-          run-id: ${{ github.run_id }}
-
-      - name: Delete old report
-        run: |
-          find ./docs/source/developer_guide/evaluation/accuracy_report -maxdepth 1 -type f -name '*.md' ! -name 'index.md' -delete
-          find ./docs/source/developer_guide/evaluation/accuracy_report -mindepth 2 -type f -name '*.md' -exec mv -f {} ./docs/source/developer_guide/evaluation/accuracy_report \;
-          find ./docs/source/developer_guide/evaluation/accuracy_report -mindepth 1 -type d -empty -delete
-      
-      - name: Update accuracy_report/index.md
-        run: |
-          REPORT_DIR="./docs/source/developer_guide/evaluation/accuracy_report"
-          INDEX_MD="$REPORT_DIR/index.md"
-          {
-            echo "# Accuracy Report"
-            echo ""
-            echo ":::{toctree}"
-            echo ":caption: Accuracy Report"
-            echo ":maxdepth: 1"
-            
-            for report in "$REPORT_DIR"/*.md; do
-              filename="$(basename "$report" .md)"
-              if [ "$filename" != "index" ]; then
-                echo "$filename"
-              fi
-            done
-            echo ":::"
-          } > "$INDEX_MD"
-
-      - name: push accuracy report
-        env:
-          GITHUB_TOKEN: ${{ secrets.PAT_TOKEN }}
-        run: |
-          git add ./docs/source/developer_guide/evaluation/accuracy_report/*.md
-          git commit -s -m "[Doc] Update accuracy reports for ${{ env.BRANCH_NAME }}"
-          git push -f origin "${{ env.BRANCH_NAME }}"
-
-      - name: Create PR in upstream via API
-        uses: actions/github-script@v8
-        with:
-          github-token: ${{ secrets.PAT_TOKEN }}
-          script: |
-            const pr = await github.rest.pulls.create({
-              owner: 'vllm-project',
-              repo: 'vllm-ascend',
-              head: `vllm-ascend-ci:${{ env.BRANCH_NAME }}`,
-              base: 'main',
-              title: `[Doc] Update accuracy reports for ${{ env.BRANCH_NAME }}`,
-              body: `The accuracy results running on NPU Altlas A2 have changed, updating reports for: All models
-            
-              - [Workflow run][1]
-              
-              [1]: ${{ github.server_url }}/${{ github.repository }}/actions/runs/${{ github.run_id }}`
-            });
-            core.info(`Created PR #${pr.data.number}`);

docs/source/developer_guide/evaluation/accuracy_report/DeepSeek-V2-Lite.md

@@ -1,20 +0,0 @@
-# deepseek-ai/DeepSeek-V2-Lite
-
-- **vLLM Version**: vLLM: 0.10.1.1 ([1da94e6](https://github.com/vllm-project/vllm/commit/1da94e6)), **vLLM Ascend Version**: v0.10.1rc1 ([7e16b4a](https://github.com/vllm-project/vllm-ascend/commit/7e16b4a))  
-- **Software Environment**: **CANN**: 8.2.RC1, **PyTorch**: 2.7.1, **torch-npu**: 2.7.1.dev20250724  
-- **Hardware Environment**: Atlas A2 Series  
-- **Parallel mode**: TP2
-- **Execution mode**: ACLGraph
-
-**Command**:  
-
-```bash
-export MODEL_ARGS='pretrained=deepseek-ai/DeepSeek-V2-Lite,tensor_parallel_size=2,dtype=auto,trust_remote_code=True,max_model_len=4096,enforce_eager=True'
-lm_eval --model vllm --model_args $MODEL_ARGS --tasks gsm8k \
-    --batch_size auto
-```
-
-| Task                  | Metric      | Value     | Stderr |
-|-----------------------|-------------|----------:|-------:|
-| gsm8k | exact_match,strict-match | ✅0.3813 | ± 0.0134 |
-| gsm8k | exact_match,flexible-extract | ✅0.3836 | ± 0.0134 |

docs/source/developer_guide/evaluation/accuracy_report/Qwen2.5-VL-7B-Instruct.md

@@ -1,19 +0,0 @@
-# Qwen/Qwen2.5-VL-7B-Instruct
-
-- **vLLM Version**: vLLM: 0.10.1.1 ([1da94e6](https://github.com/vllm-project/vllm/commit/1da94e6)), **vLLM Ascend Version**: v0.10.1rc1 ([7e16b4a](https://github.com/vllm-project/vllm-ascend/commit/7e16b4a))  
-- **Software Environment**: **CANN**: 8.2.RC1, **PyTorch**: 2.7.1, **torch-npu**: 2.7.1.dev20250724  
-- **Hardware Environment**: Atlas A2 Series  
-- **Parallel mode**: TP1
-- **Execution mode**: ACLGraph
-
-**Command**:  
-
-```bash
-export MODEL_ARGS='pretrained=Qwen/Qwen2.5-VL-7B-Instruct,tensor_parallel_size=1,dtype=auto,trust_remote_code=False,max_model_len=8192'
-lm_eval --model vllm-vlm --model_args $MODEL_ARGS --tasks mmmu_val \
- --apply_chat_template True   --fewshot_as_multiturn True    --batch_size auto
-```
-
-| Task                  | Metric      | Value     | Stderr |
-|-----------------------|-------------|----------:|-------:|
-| mmmu_val | acc,none | ✅0.52 | ± 0.0162 |

docs/source/developer_guide/evaluation/accuracy_report/Qwen3-30B-A3B.md

@@ -1,21 +0,0 @@
-# Qwen/Qwen3-30B-A3B
-
-- **vLLM Version**: vLLM: 0.10.1.1 ([1da94e6](https://github.com/vllm-project/vllm/commit/1da94e6)), **vLLM Ascend Version**: v0.10.1rc1 ([7e16b4a](https://github.com/vllm-project/vllm-ascend/commit/7e16b4a))  
-- **Software Environment**: **CANN**: 8.2.RC1, **PyTorch**: 2.7.1, **torch-npu**: 2.7.1.dev20250724  
-- **Hardware Environment**: Atlas A2 Series  
-- **Parallel mode**: TP2 + EP
-- **Execution mode**: ACLGraph
-
-**Command**:  
-
-```bash
-export MODEL_ARGS='pretrained=Qwen/Qwen3-30B-A3B,tensor_parallel_size=2,dtype=auto,trust_remote_code=False,max_model_len=4096,gpu_memory_utilization=0.6,enable_expert_parallel=True'
-lm_eval --model vllm --model_args $MODEL_ARGS --tasks gsm8k,ceval-valid \
-   --num_fewshot 5   --batch_size auto
-```
-
-| Task                  | Metric      | Value     | Stderr |
-|-----------------------|-------------|----------:|-------:|
-| gsm8k | exact_match,strict-match | ✅0.8923 | ± 0.0085 |
-| gsm8k | exact_match,flexible-extract | ✅0.8506 | ± 0.0098 |
-| ceval-valid | acc,none | ✅0.8358 | ± 0.0099 |

docs/source/developer_guide/evaluation/accuracy_report/Qwen3-8B-Base.md

@@ -1,21 +0,0 @@
-# Qwen/Qwen3-8B-Base
-
-- **vLLM Version**: vLLM: 0.10.1.1 ([1da94e6](https://github.com/vllm-project/vllm/commit/1da94e6)), **vLLM Ascend Version**: v0.10.1rc1 ([7e16b4a](https://github.com/vllm-project/vllm-ascend/commit/7e16b4a))  
-- **Software Environment**: **CANN**: 8.2.RC1, **PyTorch**: 2.7.1, **torch-npu**: 2.7.1.dev20250724  
-- **Hardware Environment**: Atlas A2 Series  
-- **Parallel mode**: TP1
-- **Execution mode**: ACLGraph
-
-**Command**:  
-
-```bash
-export MODEL_ARGS='pretrained=Qwen/Qwen3-8B-Base,tensor_parallel_size=1,dtype=auto,trust_remote_code=False,max_model_len=4096'
-lm_eval --model vllm --model_args $MODEL_ARGS --tasks gsm8k,ceval-valid \
- --apply_chat_template True   --fewshot_as_multiturn True   --num_fewshot 5   --batch_size auto
-```
-
-| Task                  | Metric      | Value     | Stderr |
-|-----------------------|-------------|----------:|-------:|
-| gsm8k | exact_match,strict-match | ✅0.8271 | ± 0.0104 |
-| gsm8k | exact_match,flexible-extract | ✅0.8294 | ± 0.0104 |
-| ceval-valid | acc,none | ✅0.815 | ± 0.0103 |

docs/source/developer_guide/evaluation/accuracy_report/index.md

@@ -1,10 +0,0 @@
-# Accuracy Report
-
-:::{toctree}
-:caption: Accuracy Report
-:maxdepth: 1
-DeepSeek-V2-Lite
-Qwen2.5-VL-7B-Instruct
-Qwen3-30B-A3B
-Qwen3-8B-Base
-:::

docs/source/developer_guide/evaluation/index.md

@@ -7,5 +7,4 @@ using_evalscope
 using_lm_eval
 using_ais_bench
 using_opencompass
-accuracy_report/index
 :::

vllm_ascend/quantization/w8a8_dynamic.py

@@ -200,7 +200,6 @@ class AscendW8A8DynamicFusedMoEMethod:
         assert router_logits.shape[
             1] == global_num_experts - global_redundant_expert_num, "Number of global experts mismatch (excluding redundancy)"
 
-        topk_weights, topk_ids = None, None
         if self.multistream_overlap_gate:
             fc3_context = get_flash_common3_context()
             assert fc3_context is not None
@@ -219,7 +218,8 @@ class AscendW8A8DynamicFusedMoEMethod:
                 scoring_func=scoring_func,
                 e_score_correction_bias=e_score_correction_bias,
                 global_num_experts=global_num_experts)
-
+        assert topk_ids is not None
+        assert topk_weights is not None
         # this is a naive implementation for experts load balance so as
         # to avoid accumulating too much tokens on a single rank.
         # currently it is only activated when doing profile runs.