Merge tag 'v0.10.2rc1' into v0.10.2rc1-ori

.github/scale-config.yml

+# scale-config.yml:
+#   Powers what instance types are available for GHA auto-scaled
+#   runners. Runners listed here will be available as self hosted
+#   runners, configuration is directly pulled from the main branch.
+# runner_types:
+#   runner_label:
+#     instance_type: m4.large
+#     os: linux
+#     # min_available defaults to the global cfg in the ALI Terraform
+#     min_available: undefined
+#     # when max_available value is not defined, no max runners is enforced
+#     max_available: undefined
+#     disk_size: 50
+#     is_ephemeral: true
+
+runner_types:
+  linux.2xlarge:
+    disk_size: 150
+    instance_type: c5.2xlarge
+    is_ephemeral: true
+    os: linux

.github/workflows/issue_autolabel.yml

+name: Label issues based on keywords
+on:
+  issues:
+    types: [opened, edited, reopened]
+permissions:
+  issues: write          # needed so the workflow can add labels
+  contents: read
+concurrency:
+  group: issue-labeler-${{ github.event.issue.number }}
+  cancel-in-progress: true
+jobs:
+  add-labels:
+    runs-on: ubuntu-latest
+    steps:
+      - name: Label issues based on keywords
+        uses: actions/github-script@60a0d83039c74a4aee543508d2ffcb1c3799cdea  # v7.0.1
+        with:
+          script: |
+            // Configuration: Add new labels and keywords here
+            const labelConfig = {
+              rocm: {
+                // Keyword search - matches whole words only (with word boundaries)
+                keywords: [
+                  {
+                    term: "composable kernel",
+                    searchIn: "both"
+                  },
+                  {
+                    term: "rccl",
+                    searchIn: "body"  // only search in body
+                  },
+                  {
+                    term: "migraphx",
+                    searchIn: "title"  // only search in title
+                  },
+                  {
+                    term: "hipgraph",
+                    searchIn: "both"
+                  },
+                  {
+                    term: "ROCm System Management Interface",
+                    searchIn: "body"
+                  },
+                ],
+                
+                // Substring search - matches anywhere in text (partial matches)
+                substrings: [
+                  {
+                    term: "VLLM_ROCM_",
+                    searchIn: "both"
+                  },
+                  {
+                    term: "aiter",
+                    searchIn: "title"
+                  },
+                  {
+                    term: "rocm",
+                    searchIn: "title"
+                  },
+                  {
+                    term: "amd",
+                    searchIn: "title"
+                  },
+                  {
+                    term: "hip-",
+                    searchIn: "both"
+                  },
+                  {
+                    term: "gfx",
+                    searchIn: "both"
+                  },
+                  {
+                    term: "cdna",
+                    searchIn: "both"
+                  },
+                  {
+                    term: "rdna",
+                    searchIn: "both"
+                  },
+                  {
+                    term: "torch_hip",
+                    searchIn: "body"  // only in body
+                  },
+                  {
+                    term: "_hip",
+                    searchIn: "both"
+                  },
+                  {
+                    term: "hip_",
+                    searchIn: "both"
+                  },
+                  
+                  // ROCm tools and libraries
+                  {
+                    term: "hipify",
+                    searchIn: "both"
+                  },
+                ],
+                
+                // Regex patterns - for complex pattern matching
+                regexPatterns: [
+                  {
+                    pattern: "\\bmi\\d{3}[a-z]*\\b",
+                    description: "AMD GPU names (mi + 3 digits + optional letters)",
+                    flags: "gi",
+                    searchIn: "both"  // "title", "body", or "both"
+                  }
+                ],
+              },
+            };
+            
+            // Helper function to create regex based on search type
+            function createSearchRegex(term, type) {
+              // Escape special regex characters in the term
+              const escapedTerm = term.replace(/[.*+?^${}()|[\]\\]/g, '\\$&');
+              
+              switch (type) {
+                case 'keyword':
+                  // Word boundary search - matches whole words only
+                  return new RegExp(`\\b${escapedTerm}\\b`, "gi");
+                case 'substring':
+                  // Substring search - matches anywhere in the text
+                  return new RegExp(escapedTerm, "gi");
+                default:
+                  throw new Error(`Unknown search type: ${type}`);
+              }
+            }
+            
+            // Helper function to find matching terms in text with line information
+            function findMatchingTermsWithLines(text, searchTerms = [], searchType = 'keyword', searchLocation = '') {
+              const matches = [];
+              const lines = text.split('\n');
+              
+              for (const termConfig of searchTerms) {
+                let regex;
+                let term, searchIn, pattern, description, flags;
+                
+                // Handle different input formats (string or object)
+                if (typeof termConfig === 'string') {
+                  term = termConfig;
+                  searchIn = 'both'; // default
+                } else {
+                  term = termConfig.term;
+                  searchIn = termConfig.searchIn || 'both';
+                  pattern = termConfig.pattern;
+                  description = termConfig.description;
+                  flags = termConfig.flags;
+                }
+                
+                // Skip if this term shouldn't be searched in the current location
+                if (searchIn !== 'both' && searchIn !== searchLocation) {
+                  continue;
+                }
+                
+                // Create appropriate regex
+                if (searchType === 'regex') {
+                  regex = new RegExp(pattern, flags || "gi");
+                } else {
+                  regex = createSearchRegex(term, searchType);
+                }
+                
+                const termMatches = [];
+                
+                // Check each line for matches
+                lines.forEach((line, lineIndex) => {
+                  const lineMatches = line.match(regex);
+                  if (lineMatches) {
+                    lineMatches.forEach(match => {
+                      termMatches.push({
+                        match: match,
+                        lineNumber: lineIndex + 1,
+                        lineContent: line.trim(),
+                        searchType: searchType,
+                        searchLocation: searchLocation,
+                        originalTerm: term || pattern,
+                        description: description,
+                        // Show context around the match in the line
+                        context: line.length > 100 ? 
+                          line.substring(Math.max(0, line.toLowerCase().indexOf(match.toLowerCase()) - 30), 
+                                       line.toLowerCase().indexOf(match.toLowerCase()) + match.length + 30) + '...' 
+                          : line.trim()
+                      });
+                    });
+                  }
+                });
+                
+                if (termMatches.length > 0) {
+                  matches.push({
+                    term: term || (description || pattern),
+                    searchType: searchType,
+                    searchLocation: searchLocation,
+                    searchIn: searchIn,
+                    pattern: pattern,
+                    matches: termMatches,
+                    count: termMatches.length
+                  });
+                }
+              }
+              
+              return matches;
+            }
+            
+            // Helper function to check if label should be added
+            async function processLabel(labelName, config) {
+              const body = context.payload.issue.body || "";
+              const title = context.payload.issue.title || "";
+              
+              core.notice(`Processing label: ${labelName}`);
+              core.notice(`Issue Title: "${title}"`);
+              core.notice(`Issue Body length: ${body.length} characters`);
+              
+              let shouldAddLabel = false;
+              let allMatches = [];
+              let reason = '';
+              
+              const keywords = config.keywords || [];
+              const substrings = config.substrings || [];
+              const regexPatterns = config.regexPatterns || [];
+              
+              core.notice(`Searching with ${keywords.length} keywords, ${substrings.length} substrings, and ${regexPatterns.length} regex patterns`);
+              
+              // Search in title
+              if (title.trim()) {
+                core.notice(`Searching in title: "${title}"`);
+                
+                const titleKeywordMatches = findMatchingTermsWithLines(title, keywords, 'keyword', 'title');
+                const titleSubstringMatches = findMatchingTermsWithLines(title, substrings, 'substring', 'title');
+                const titleRegexMatches = findMatchingTermsWithLines(title, regexPatterns, 'regex', 'title');
+                
+                allMatches.push(...titleKeywordMatches, ...titleSubstringMatches, ...titleRegexMatches);
+              }
+              
+              // Search in body
+              if (body.trim()) {
+                core.notice(`Searching in body (${body.length} characters)`);
+                
+                const bodyKeywordMatches = findMatchingTermsWithLines(body, keywords, 'keyword', 'body');
+                const bodySubstringMatches = findMatchingTermsWithLines(body, substrings, 'substring', 'body');
+                const bodyRegexMatches = findMatchingTermsWithLines(body, regexPatterns, 'regex', 'body');
+                
+                allMatches.push(...bodyKeywordMatches, ...bodySubstringMatches, ...bodyRegexMatches);
+              }
+              
+              if (allMatches.length > 0) {
+                core.notice(`Found ${allMatches.length} matching term(s):`);
+                
+                for (const termMatch of allMatches) {
+                  const locationText = termMatch.searchLocation === 'title' ? 'title' : 'body';
+                  const searchInText = termMatch.searchIn === 'both' ? 'both' : termMatch.searchIn;
+                  
+                  if (termMatch.searchType === 'regex') {
+                    core.notice(`  📍 Regex: "${termMatch.term}" (pattern: ${termMatch.pattern}) found ${termMatch.count} time(s) in ${locationText} (configured to search in: ${searchInText}):`);
+                  } else {
+                    core.notice(`  📍 Term: "${termMatch.term}" (${termMatch.searchType} search) found ${termMatch.count} time(s) in ${locationText} (configured to search in: ${searchInText}):`);
+                  }
+                  
+                  // Show details for each match
+                  termMatch.matches.forEach((match, index) => {
+                    core.notice(`    ${index + 1}. Line ${match.lineNumber} in ${match.searchLocation}: "${match.match}" [${match.searchType}]`);
+                    if (match.description) {
+                      core.notice(`       Description: ${match.description}`);
+                    }
+                    core.notice(`       Context: ${match.context}`);
+                    if (match.lineContent !== match.context) {
+                      core.notice(`       Full line: ${match.lineContent}`);
+                    }
+                  });
+                }
+                
+                shouldAddLabel = true;
+                const totalMatches = allMatches.reduce((sum, t) => sum + t.count, 0);
+                const titleMatches = allMatches.filter(t => t.searchLocation === 'title').reduce((sum, t) => sum + t.count, 0);
+                const bodyMatches = allMatches.filter(t => t.searchLocation === 'body').reduce((sum, t) => sum + t.count, 0);
+                const keywordMatches = allMatches.filter(t => t.searchType === 'keyword').reduce((sum, t) => sum + t.count, 0);
+                const substringMatches = allMatches.filter(t => t.searchType === 'substring').reduce((sum, t) => sum + t.count, 0);
+                const regexMatches = allMatches.filter(t => t.searchType === 'regex').reduce((sum, t) => sum + t.count, 0);
+                
+                reason = `Found ${totalMatches} total matches (${titleMatches} in title, ${bodyMatches} in body) - ${keywordMatches} keyword matches, ${substringMatches} substring matches, ${regexMatches} regex matches`;
+              }
+              
+              core.notice(`Final decision: ${shouldAddLabel ? 'ADD LABEL' : 'DO NOT ADD LABEL'}`);
+              core.notice(`Reason: ${reason || 'No matching terms found'}`);
+              
+              if (shouldAddLabel) {
+                const existingLabels = context.payload.issue.labels.map(l => l.name);
+                if (!existingLabels.includes(labelName)) {
+                  await github.rest.issues.addLabels({
+                    owner: context.repo.owner,
+                    repo: context.repo.repo,
+                    issue_number: context.issue.number,
+                    labels: [labelName],
+                  });
+                  core.notice(`Label "${labelName}" added. ${reason}`);
+                  return true;
+                }
+                core.notice(`Label "${labelName}" already present.`);
+                return false;
+              }
+              
+              core.notice(`No matching terms found for label "${labelName}".`);
+              return false;
+            }
+            
+            // Process all configured labels
+            const processLabels = Object.entries(labelConfig)
+              .map(([labelName, config]) => processLabel(labelName, config));
+            const labelsAdded = await Promise.all(processLabels);
+            const numLabelsAdded = labelsAdded.reduce((x, y) => x + y, 0);
+            core.notice(`Processing complete. ${numLabelsAdded} label(s) added.`);
\ No newline at end of file

.github/workflows/lint-and-deploy.yaml

-name: Lint and Deploy Charts
-
-on: pull_request
-
-concurrency:
-  group: ${{ github.workflow }}-${{ github.ref }}
-  cancel-in-progress: true
-
-permissions:
-  contents: read
-
-jobs:
-  lint-and-deploy:
-    runs-on: ubuntu-latest
-    steps:
-      - name: Checkout
-        uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683 # v4.2.2
-        with:
-          fetch-depth: 0
-
-      - name: Set up Helm
-        uses: azure/setup-helm@b9e51907a09c216f16ebe8536097933489208112 # v4.3.0
-        with:
-          version: v3.14.4
-
-       #Python is required because ct lint runs Yamale and yamllint which require Python.
-      - uses: actions/setup-python@42375524e23c412d93fb67b49958b491fce71c38 # v5.4.0
-        with:
-          python-version: '3.13'
-
-      - name: Set up chart-testing
-        uses: helm/chart-testing-action@0d28d3144d3a25ea2cc349d6e59901c4ff469b3b # v2.7.0
-        with:
-          version: v3.10.1
-
-      - name: Run chart-testing (lint)
-        run: ct lint --target-branch ${{ github.event.repository.default_branch }} --chart-dirs examples/online_serving/chart-helm --charts examples/online_serving/chart-helm
-
-      - name: Setup minio
-        run: |
-          docker network create vllm-net
-          docker run -d -p 9000:9000 --name minio --net vllm-net \
-                     -e "MINIO_ACCESS_KEY=minioadmin" \
-                     -e "MINIO_SECRET_KEY=minioadmin" \
-                     -v /tmp/data:/data \
-                     -v /tmp/config:/root/.minio \
-                     minio/minio server /data
-          export AWS_ACCESS_KEY_ID=minioadmin
-          export AWS_SECRET_ACCESS_KEY=minioadmin
-          export AWS_EC2_METADATA_DISABLED=true
-          mkdir opt-125m
-          cd opt-125m && curl -O -Ls "https://huggingface.co/facebook/opt-125m/resolve/main/{pytorch_model.bin,config.json,generation_config.json,merges.txt,special_tokens_map.json,tokenizer_config.json,vocab.json}" && cd ..
-          aws --endpoint-url http://127.0.0.1:9000/ s3 mb s3://testbucket
-          aws --endpoint-url http://127.0.0.1:9000/ s3 cp opt-125m/ s3://testbucket/opt-125m --recursive
-
-      - name: Create kind cluster
-        uses: helm/kind-action@a1b0e391336a6ee6713a0583f8c6240d70863de3 # v1.12.0
-
-      - name: Build the Docker image vllm cpu
-        run: docker buildx build -f docker/Dockerfile.cpu -t vllm-cpu-env .
-
-      - name: Configuration of docker images, network and namespace for the kind cluster
-        run: |
-          docker pull amazon/aws-cli:2.6.4
-          kind load docker-image  amazon/aws-cli:2.6.4 --name chart-testing
-          kind load docker-image vllm-cpu-env:latest --name chart-testing
-          docker network connect vllm-net "$(docker ps -aqf "name=chart-testing-control-plane")"
-          kubectl create ns ns-vllm
-
-      - name: Run chart-testing (install)
-        run: |
-          export AWS_ACCESS_KEY_ID=minioadmin
-          export AWS_SECRET_ACCESS_KEY=minioadmin
-          sleep 30 && kubectl -n ns-vllm logs -f "$(kubectl -n ns-vllm get pods | awk '/deployment/ {print $1;exit}')" &
-          helm install --wait --wait-for-jobs --timeout 5m0s --debug --create-namespace --namespace=ns-vllm test-vllm examples/online_serving/chart-helm -f examples/online_serving/chart-helm/values.yaml --set secrets.s3endpoint=http://minio:9000 --set secrets.s3bucketname=testbucket --set secrets.s3accesskeyid=$AWS_ACCESS_KEY_ID --set secrets.s3accesskey=$AWS_SECRET_ACCESS_KEY --set resources.requests.cpu=1 --set resources.requests.memory=4Gi --set resources.limits.cpu=2 --set resources.limits.memory=5Gi --set image.env[0].name=VLLM_CPU_KVCACHE_SPACE --set image.env[1].name=VLLM_LOGGING_LEVEL --set image.env[2].name=VLLM_CPU_CI_ENV --set-string image.env[0].value="1" --set-string image.env[1].value="DEBUG" --set-string image.env[2].value="1" --set-string extraInit.s3modelpath="opt-125m/" --set-string 'resources.limits.nvidia\.com/gpu=0' --set-string 'resources.requests.nvidia\.com/gpu=0' --set-string image.repository="vllm-cpu-env"
-
-      - name: curl test
-        run: |
-          kubectl -n ns-vllm port-forward service/test-vllm-service 8001:80 &
-          sleep 10
-          CODE="$(curl -v -f --location http://localhost:8001/v1/completions \
-                  --header "Content-Type: application/json" \
-                  --data '{
-                          "model": "opt-125m",
-                          "prompt": "San Francisco is a",
-                          "max_tokens": 7,
-                          "temperature": 0
-                  }'):$CODE"
-          echo "$CODE"

.github/workflows/publish.yml

-# This workflow will upload a Python Package to Release asset
-# For more information see: https://help.github.com/en/actions/language-and-framework-guides/using-python-with-github-actions
-
-name: Create Release
-
-on:
-  push:
-    tags:
-      - v*
-
-# Needed to create release and upload assets
-permissions:
-  contents: write
-
-jobs:
-  release:
-    # Retrieve tag and create release
-    name: Create Release
-    runs-on: ubuntu-latest
-    outputs:
-      upload_url: ${{ steps.create_release.outputs.upload_url }}
-    steps:
-      - name: Checkout
-        uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683 # v4.2.2
-
-      - name: Extract branch info
-        shell: bash
-        run: |
-          echo "release_tag=${GITHUB_REF#refs/*/}" >> "$GITHUB_ENV"
-
-      - name: Create Release
-        id: create_release
-        uses: actions/github-script@60a0d83039c74a4aee543508d2ffcb1c3799cdea # v7.0.1
-        env:
-          RELEASE_TAG: ${{ env.release_tag }}
-        with:
-          github-token: "${{ secrets.GITHUB_TOKEN }}"
-          script: |
-            const script = require('.github/workflows/scripts/create_release.js')
-            await script(github, context, core)
-
-  # NOTE(simon): No longer build wheel using GitHub Actions. See buildkite's release workflow. 
-  # wheel:
-  #   name: Build Wheel
-  #   runs-on: ${{ matrix.os }}
-  #   needs: release
-
-  #   strategy:
-  #     fail-fast: false
-  #     matrix:
-  #         os: ['ubuntu-20.04']
-  #         python-version: ['3.9', '3.10', '3.11', '3.12']
-  #         pytorch-version: ['2.4.0']  # Must be the most recent version that meets requirements/cuda.txt.
-  #         cuda-version: ['11.8', '12.1']
-
-  #   steps:
-  #     - name: Checkout
-  #       uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683 # v4.2.2
-
-  #     - name: Setup ccache
-  #       uses: hendrikmuhs/ccache-action@ed74d11c0b343532753ecead8a951bb09bb34bc9 # v1.2.14
-  #       with:
-  #         create-symlink: true
-  #         key: ${{ github.job }}-${{ matrix.python-version }}-${{ matrix.cuda-version }}
-
-  #     - name: Set up Linux Env
-  #       if: ${{ runner.os == 'Linux' }}
-  #       run: |
-  #         bash -x .github/workflows/scripts/env.sh
-
-  #     - name: Set up Python
-  #       uses: actions/setup-python@0b93645e9fea7318ecaed2b359559ac225c90a2b # v5.3.0
-  #       with:
-  #           python-version: ${{ matrix.python-version }}
-
-  #     - name: Install CUDA ${{ matrix.cuda-version }}
-  #       run: |
-  #         bash -x .github/workflows/scripts/cuda-install.sh ${{ matrix.cuda-version }} ${{ matrix.os }}
-
-  #     - name: Install PyTorch ${{ matrix.pytorch-version }} with CUDA ${{ matrix.cuda-version }}
-  #       run: |
-  #         bash -x .github/workflows/scripts/pytorch-install.sh ${{ matrix.python-version }} ${{ matrix.pytorch-version }} ${{ matrix.cuda-version }}
-
-  #     - name: Build wheel
-  #       shell: bash
-  #       env:
-  #         CMAKE_BUILD_TYPE: Release # do not compile with debug symbol to reduce wheel size
-  #       run: |
-  #         bash -x .github/workflows/scripts/build.sh ${{ matrix.python-version }} ${{ matrix.cuda-version }}
-  #         wheel_name=$(find dist -name "*whl" -print0 | xargs -0 -n 1 basename)
-  #         asset_name=${wheel_name//"linux"/"manylinux1"}
-  #         echo "wheel_name=${wheel_name}" >> "$GITHUB_ENV"
-  #         echo "asset_name=${asset_name}" >> "$GITHUB_ENV"
-
-  #     - name: Upload Release Asset
-  #       uses: actions/upload-release-asset@e8f9f06c4b078e705bd2ea027f0926603fc9b4d5 # v1.0.2
-  #       env:
-  #         GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}
-  #       with:
-  #         upload_url: ${{ needs.release.outputs.upload_url }}
-  #         asset_path: ./dist/${{ env.wheel_name }}
-  #         asset_name: ${{ env.asset_name }}
-  #         asset_content_type: application/*
-
-      # (Danielkinz): This last step will publish the .whl to pypi. Warning: untested
-      # - name: Publish package
-      #   uses: pypa/gh-action-pypi-publish@release/v1.8
-      #   with:
-      #     repository-url: https://test.pypi.org/legacy/
-      #     password: ${{ secrets.PYPI_API_TOKEN }}
-      #     skip-existing: true

.github/workflows/reminder_comment.yml

@@ -12,16 +12,43 @@ jobs:
         uses: actions/github-script@60a0d83039c74a4aee543508d2ffcb1c3799cdea # v7.0.1
         with:
           script: |
-            github.rest.issues.createComment({
+            try {
+              // Get the PR author
+              const prAuthor = context.payload.pull_request.user.login;
+              
+              // Check if this is the author's first PR in this repository
+              // Use GitHub's search API to find all PRs by this author
+              const { data: searchResults } = await github.rest.search.issuesAndPullRequests({
+                q: `repo:${context.repo.owner}/${context.repo.repo} type:pr author:${prAuthor}`,
+                per_page: 100  
+              });
+              
+              const authorPRCount = searchResults.total_count;
+              
+              console.log(`Found ${authorPRCount} PRs by ${prAuthor}`);
+              
+              // Only post comment if this is the first PR (only one PR by this author)
+              if (authorPRCount === 1) {
+                console.log(`Posting welcome comment for first-time contributor: ${prAuthor}`);
+                await github.rest.issues.createComment({
                 owner: context.repo.owner,
                 repo: context.repo.repo,
                 issue_number: context.issue.number,
                 body: '👋 Hi! Thank you for contributing to the vLLM project.\n\n' +
                   '💬 Join our developer Slack at https://slack.vllm.ai to discuss your PR in #pr-reviews, coordinate on features in #feat- channels, or join special interest groups in #sig- channels.\n\n' +
-                'Just a reminder: PRs would not trigger full CI run by default. Instead, it would only run `fastcheck` CI which starts running only a small and essential subset of CI tests to quickly catch errors. You can run other CI tests on top of those by going to your `fastcheck` build on Buildkite UI (linked in the PR checks section) and unblock them. If you do not have permission to unblock, ping `simon-mo` or `khluu` to add you in our Buildkite org.\n\n' +
+                  'Just a reminder: PRs would not trigger full CI run by default. Instead, it would only run `fastcheck` CI which starts running only a small and essential subset of CI tests to quickly catch errors. \n\n' +
+                  'You ask your reviewers to trigger select CI tests on top of `fastcheck` CI. \n\n' +
                   'Once the PR is approved and ready to go, your PR reviewer(s) can run CI to test the changes comprehensively before merging.\n\n' +
                   'To run CI, PR reviewers can either: Add `ready` label to the PR or enable auto-merge.\n\n' +
+                  'If you have any questions, please reach out to us on Slack at https://slack.vllm.ai.\n\n' +
                   '🚀'
-            })
+                });
+              } else {
+                console.log(`Skipping comment for ${prAuthor} - not their first PR (${authorPRCount} PRs found)`);
+              }
+            } catch (error) {
+              console.error('Error checking PR history or posting comment:', error);
+              // Don't fail the workflow, just log the error
+            }
         env:
           GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}

.pre-commit-config.yaml

@@ -21,7 +21,7 @@ repos:
   - id: ruff-format
     files: ^(.buildkite|benchmarks|examples)/.*
 - repo: https://github.com/crate-ci/typos
-  rev: v1.34.0
+  rev: v1.35.5
   hooks:
   - id: typos
 - repo: https://github.com/PyCQA/isort

CMakeLists.txt

@@ -30,7 +30,7 @@ install(CODE "set(CMAKE_INSTALL_LOCAL_ONLY TRUE)" ALL_COMPONENTS)
 # Supported python versions.  These versions will be searched in order, the
 # first match will be selected.  These should be kept in sync with setup.py.
 #
-set(PYTHON_SUPPORTED_VERSIONS "3.9" "3.10" "3.11" "3.12")
+set(PYTHON_SUPPORTED_VERSIONS "3.9" "3.10" "3.11" "3.12" "3.13")
 
 # Supported AMD GPU architectures.
 set(HIP_SUPPORTED_ARCHS "gfx906;gfx908;gfx90a;gfx942;gfx950;gfx1030;gfx1100;gfx1101;gfx1200;gfx1201;gfx928;gfx936")
@@ -45,8 +45,8 @@ set(HIP_SUPPORTED_ARCHS "gfx906;gfx908;gfx90a;gfx942;gfx950;gfx1030;gfx1100;gfx1
 # requirements.txt files and should be kept consistent.  The ROCm torch
 # versions are derived from docker/Dockerfile.rocm
 #
-set(TORCH_SUPPORTED_VERSION_CUDA "2.7.1")
-set(TORCH_SUPPORTED_VERSION_ROCM "2.7.0")
+set(TORCH_SUPPORTED_VERSION_CUDA "2.8.0")
+set(TORCH_SUPPORTED_VERSION_ROCM "2.8.0")
 
 #
 # Try to find python package with an executable that exactly matches
@@ -358,9 +358,7 @@ if(VLLM_GPU_LANG STREQUAL "CUDA")
     list(APPEND VLLM_EXT_SRC ${MARLIN_TEMPLATE_KERNEL_SRC})
 
     set(MARLIN_SRCS
-       "csrc/quantization/marlin/dense/marlin_cuda_kernel.cu"
        "csrc/quantization/marlin/sparse/marlin_24_cuda_kernel.cu"
-       "csrc/quantization/marlin/qqq/marlin_qqq_gemm_kernel.cu"
        "csrc/quantization/gptq_marlin/gptq_marlin.cu"
        "csrc/quantization/gptq_marlin/gptq_marlin_repack.cu"
        "csrc/quantization/gptq_marlin/awq_marlin_repack.cu")
@@ -544,6 +542,7 @@ if(VLLM_GPU_LANG STREQUAL "CUDA")
   if(${CMAKE_CUDA_COMPILER_VERSION} VERSION_GREATER_EQUAL 12.8 AND FP4_ARCHS)
     set(SRCS
       "csrc/quantization/fp4/nvfp4_quant_kernels.cu"
+      "csrc/quantization/fp4/activation_nvfp4_quant_fusion_kernels.cu"
       "csrc/quantization/fp4/nvfp4_scaled_mm_sm120_kernels.cu")
     set_gencode_flags_for_srcs(
       SRCS "${SRCS}"
@@ -562,6 +561,7 @@ if(VLLM_GPU_LANG STREQUAL "CUDA")
   if(${CMAKE_CUDA_COMPILER_VERSION} VERSION_GREATER_EQUAL 12.8 AND FP4_ARCHS)
     set(SRCS
       "csrc/quantization/fp4/nvfp4_quant_kernels.cu"
+      "csrc/quantization/fp4/activation_nvfp4_quant_fusion_kernels.cu"
       "csrc/quantization/fp4/nvfp4_experts_quant.cu"
       "csrc/quantization/fp4/nvfp4_scaled_mm_kernels.cu"
       "csrc/quantization/fp4/nvfp4_blockwise_moe_kernel.cu")
@@ -753,6 +753,33 @@ if(VLLM_GPU_LANG STREQUAL "CUDA")
                      "found in CUDA target architectures")
     endif()
   endif()
+
+  # Only build W4A8 kernels if we are building for something compatible with sm90a
+  cuda_archs_loose_intersection(W4A8_ARCHS "9.0a" "${CUDA_ARCHS}")
+  if(${CMAKE_CUDA_COMPILER_VERSION} VERSION_GREATER_EQUAL 12.0 AND W4A8_ARCHS)
+    set(SRCS
+       "csrc/quantization/cutlass_w4a8/w4a8_mm_entry.cu")
+
+    set_gencode_flags_for_srcs(
+      SRCS "${SRCS}"
+      CUDA_ARCHS "${W4A8_ARCHS}")
+
+    list(APPEND VLLM_EXT_SRC "${SRCS}")
+
+    message(STATUS "Building W4A8 kernels for archs: ${W4A8_ARCHS}")
+  else()
+    if (NOT ${CMAKE_CUDA_COMPILER_VERSION} VERSION_GREATER_EQUAL 12.0
+        AND W4A8_ARCHS)
+      message(STATUS "Not building W4A8 kernels as CUDA Compiler version is "
+                     "not >= 12.0, we recommend upgrading to CUDA 12.0 or "
+                     "later if you intend on running w4a16 quantized models on "
+                     "Hopper.")
+    else()
+      message(STATUS "Not building W4A8 kernels as no compatible archs "
+                     "found in CUDA target architectures")
+    endif()
+  endif()
+
 # if CUDA endif
 endif()
 
@@ -793,7 +820,9 @@ set(VLLM_MOE_EXT_SRC
   "csrc/moe/topk_softmax_kernels.cu")
 
 if(VLLM_GPU_LANG STREQUAL "CUDA")
-  list(APPEND VLLM_MOE_EXT_SRC "csrc/moe/moe_wna16.cu")
+  list(APPEND VLLM_MOE_EXT_SRC
+    "csrc/moe/moe_wna16.cu"
+    "csrc/moe/grouped_topk_kernels.cu")
 endif()
 
 if(VLLM_GPU_LANG STREQUAL "CUDA")

README.md

@@ -18,14 +18,16 @@ Easy, fast, and cheap LLM serving for everyone
 
 *Latest News* 🔥
 
+- [2025/08] We hosted [vLLM Shanghai Meetup](https://mp.weixin.qq.com/s/pDmAXHcN7Iqc8sUKgJgGtg) focusing on building, developing, and integrating with vLLM! Please find the meetup slides [here](https://drive.google.com/drive/folders/1OvLx39wnCGy_WKq8SiVKf7YcxxYI3WCH).
+- [2025/08] We hosted [vLLM Korea Meetup](https://luma.com/cgcgprmh) with Red Hat and Rebellions! We shared the latest advancements in vLLM along with project spotlights from the vLLM Korea community. Please find the meetup slides [here](https://drive.google.com/file/d/1bcrrAE1rxUgx0mjIeOWT6hNe2RefC5Hm/view).
 - [2025/08] We hosted [vLLM Beijing Meetup](https://mp.weixin.qq.com/s/dgkWg1WFpWGO2jCdTqQHxA) focusing on large-scale LLM deployment! Please find the meetup slides [here](https://drive.google.com/drive/folders/1Pid6NSFLU43DZRi0EaTcPgXsAzDvbBqF) and the recording [here](https://www.chaspark.com/#/live/1166916873711665152).
-- [2025/05] We hosted [NYC vLLM Meetup](https://lu.ma/c1rqyf1f)! Please find the meetup slides [here](https://docs.google.com/presentation/d/1_q_aW_ioMJWUImf1s1YM-ZhjXz8cUeL0IJvaquOYBeA/edit?usp=sharing).
 - [2025/05] vLLM is now a hosted project under PyTorch Foundation! Please find the announcement [here](https://pytorch.org/blog/pytorch-foundation-welcomes-vllm/).
 - [2025/01] We are excited to announce the alpha release of vLLM V1: A major architectural upgrade with 1.7x speedup! Clean code, optimized execution loop, zero-overhead prefix caching, enhanced multimodal support, and more. Please check out our blog post [here](https://blog.vllm.ai/2025/01/27/v1-alpha-release.html).
 
 <details>
 <summary>Previous News</summary>
 
+- [2025/05] We hosted [NYC vLLM Meetup](https://lu.ma/c1rqyf1f)! Please find the meetup slides [here](https://docs.google.com/presentation/d/1_q_aW_ioMJWUImf1s1YM-ZhjXz8cUeL0IJvaquOYBeA/edit?usp=sharing).
 - [2025/04] We hosted [Asia Developer Day](https://www.sginnovate.com/event/limited-availability-morning-evening-slots-remaining-inaugural-vllm-asia-developer-day)! Please find the meetup slides from the vLLM team [here](https://docs.google.com/presentation/d/19cp6Qu8u48ihB91A064XfaXruNYiBOUKrBxAmDOllOo/edit?usp=sharing).
 - [2025/03] We hosted [vLLM x Ollama Inference Night](https://lu.ma/vllm-ollama)! Please find the meetup slides from the vLLM team [here](https://docs.google.com/presentation/d/16T2PDD1YwRnZ4Tu8Q5r6n53c5Lr5c73UV9Vd2_eBo4U/edit?usp=sharing).
 - [2025/03] We hosted [the first vLLM China Meetup](https://mp.weixin.qq.com/s/n77GibL2corAtQHtVEAzfg)! Please find the meetup slides from vLLM team [here](https://docs.google.com/presentation/d/1REHvfQMKGnvz6p3Fd23HhSO4c8j5WPGZV0bKYLwnHyQ/edit?usp=sharing).

SECURITY.md

@@ -42,4 +42,9 @@ For certain security issues of CRITICAL, HIGH, or MODERATE severity level, we ma
 
 * If you wish to be added to the prenotification group, please send an email copying all the members of the [vulnerability management team](https://docs.vllm.ai/en/latest/contributing/vulnerability_management.html). Each vendor contact will be analyzed on a case-by-case basis.
 
+* Organizations and vendors who either ship or use vLLM, are eligible to join the prenotification group if they meet at least one of the following qualifications
+    * Substantial internal deployment leveraging the upstream vLLM project.
+    * Established internal security teams and comprehensive compliance measures.
+    * Active and consistent contributions to the upstream vLLM project.
+
 * We may withdraw organizations from receiving future prenotifications if they release fixes or any other information about issues before they are public. Group membership may also change based on policy refinements for who may be included.

benchmarks/README.md

@@ -32,6 +32,14 @@ become available.
         <div>Note that the images need to be downloaded separately. For example, to download COCO's 2017 Train images:</div>
         <code>wget http://images.cocodataset.org/zips/train2017.zip</code>
       </td>
+    </tr>
+        <tr>
+      <td><strong>ShareGPT4Video (Video)</strong></td>
+      <td style="text-align: center;">✅</td>
+      <td style="text-align: center;">✅</td>
+      <td>
+        <code>git clone https://huggingface.co/datasets/ShareGPT4Video/ShareGPT4Video</code>
+      </td>
     </tr>
     <tr>
       <td><strong>BurstGPT</strong></td>
@@ -51,6 +59,12 @@ become available.
       <td style="text-align: center;">✅</td>
       <td><code>synthetic</code></td>
     </tr>
+    <tr>
+      <td><strong>RandomMultiModal (Image/Video)</strong></td>
+      <td style="text-align: center;">🟡</td>
+      <td style="text-align: center;">🚧</td>
+      <td><code>synthetic</code> </td>
+    </tr>
     <tr>
       <td><strong>Prefix Repetition</strong></td>
       <td style="text-align: center;">✅</td>
@@ -194,6 +208,7 @@ vllm serve Qwen/Qwen2-VL-7B-Instruct
 ```bash
 vllm bench serve \
   --backend openai-chat \
+  --endpoint-type openai-chat \
   --model Qwen/Qwen2-VL-7B-Instruct \
   --endpoint /v1/chat/completions \
   --dataset-name hf \
@@ -230,6 +245,7 @@ vllm serve Qwen/Qwen2-VL-7B-Instruct
 ```bash
 vllm bench serve \
   --backend openai-chat \
+  --endpoint-type openai-chat \
   --model Qwen/Qwen2-VL-7B-Instruct \
   --endpoint /v1/chat/completions \
   --dataset-name hf \
@@ -244,6 +260,7 @@ vllm bench serve \
 ```bash
 vllm bench serve \
   --backend openai-chat \
+  --endpoint-type openai-chat \
   --model Qwen/Qwen2-VL-7B-Instruct \
   --endpoint /v1/chat/completions \
   --dataset-name hf \
@@ -684,4 +701,102 @@ python benchmarks/benchmark_serving.py \
   --endpoint /v1/chat/completion
 ```
 
+### Videos (ShareGPT4Video)
+
+Start vLLM:
+
+```bash
+python -m vllm.entrypoints.openai.api_server \
+  --model Qwen/Qwen2.5-VL-7B-Instruct \
+  --dtype bfloat16 \
+  --limit-mm-per-prompt '{"video": 1}' \
+  --allowed-local-media-path /path/to/sharegpt4video/videos
+```
+
+Send requests with videos:
+
+```bash
+python benchmarks/benchmark_serving.py \
+  --backend openai-chat \
+  --model Qwen/Qwen2.5-VL-7B-Instruct \
+  --dataset-name sharegpt \
+  --dataset-path /path/to/ShareGPT4Video/llava_v1_5_mix665k_with_video_chatgpt72k_share4video28k.json \
+  --num-prompts 100 \
+  --save-result \
+  --result-dir ~/vllm_benchmark_results \
+  --save-detailed \
+  --endpoint /v1/chat/completion
+```
+
+### Synthetic Random Images (random-mm)
+
+Generate synthetic image inputs alongside random text prompts to stress-test vision models without external datasets.
+
+Notes:
+
+- Works only with online benchmark via the OpenAI  backend (`--backend openai-chat`) and endpoint `/v1/chat/completions`.
+- Video sampling is not yet implemented.
+
+Start the server (example):
+
+```bash
+vllm serve Qwen/Qwen2.5-VL-3B-Instruct \
+  --dtype bfloat16 \
+  --max-model-len 16384 \
+  --limit-mm-per-prompt '{"image": 3, "video": 0}' \
+  --mm-processor-kwargs max_pixels=1003520
+```
+
+Benchmark. It is recommended to use the flag `--ignore-eos` to simulate real responses. You can set the size of the output via the arg `random-output-len`.
+
+Ex.1: Fixed number of items and a single image resolution, enforcing generation of approx 40 tokens:
+
+```bash
+vllm bench serve \
+  --backend openai-chat \
+  --model Qwen/Qwen2.5-VL-3B-Instruct \
+  --endpoint /v1/chat/completions \
+  --dataset-name random-mm \
+  --num-prompts 100 \
+  --max-concurrency 10 \
+  --random-prefix-len 25 \
+  --random-input-len 300 \
+  --random-output-len 40 \
+  --random-range-ratio 0.2 \
+  --random-mm-base-items-per-request 2 \
+  --random-mm-limit-mm-per-prompt '{"image": 3, "video": 0}' \
+  --random-mm-bucket-config '{(224, 224, 1): 1.0}' \
+  --request-rate inf \
+  --ignore-eos \
+  --seed 42
+```
+
+The number of items per request can be controlled by passing multiple image buckets:
+
+```bash
+  --random-mm-base-items-per-request 2 \
+  --random-mm-num-mm-items-range-ratio 0.5 \
+  --random-mm-limit-mm-per-prompt '{"image": 4, "video": 0}' \
+  --random-mm-bucket-config '{(256, 256, 1): 0.7, (720, 1280, 1): 0.3}' \
+```
+
+Flags specific to `random-mm`:
+
+- `--random-mm-base-items-per-request`: base number of multimodal items per request.
+- `--random-mm-num-mm-items-range-ratio`: vary item count uniformly in the closed integer range [floor(n·(1−r)), ceil(n·(1+r))]. Set r=0 to keep it fixed; r=1 allows 0 items.
+- `--random-mm-limit-mm-per-prompt`: per-modality hard caps, e.g. '{"image": 3, "video": 0}'.
+- `--random-mm-bucket-config`: dict mapping (H, W, T) → probability. Entries with probability 0 are removed; remaining probabilities are renormalized to sum to 1. Use T=1 for images. Set any T>1 for videos (video sampling not yet supported).
+
+Behavioral notes:
+
+- If the requested base item count cannot be satisfied under the provided per-prompt limits, the tool raises an error rather than silently clamping.
+
+How sampling works:
+
+- Determine per-request item count k by sampling uniformly from the integer range defined by `--random-mm-base-items-per-request` and `--random-mm-num-mm-items-range-ratio`, then clamp k to at most the sum of per-modality limits.
+- For each of the k items, sample a bucket (H, W, T) according to the normalized probabilities in `--random-mm-bucket-config`, while tracking how many items of each modality have been added.
+- If a modality (e.g., image) reaches its limit from `--random-mm-limit-mm-per-prompt`, all buckets of that modality are excluded and the remaining bucket probabilities are renormalized before continuing.
+This should be seen as an edge case, and if this behavior can be avoided by setting `--random-mm-limit-mm-per-prompt` to a large number. Note that this might result in errors due to engine config `--limit-mm-per-prompt`.
+- The resulting request contains synthetic image data in `multi_modal_data` (OpenAI Chat format). When `random-mm` is used with the OpenAI Chat backend, prompts remain text and MM content is attached via `multi_modal_data`.
+
 </details>

benchmarks/backend_request_func.py

@@ -34,6 +34,7 @@ class RequestFuncInput:
     multi_modal_content: Optional[dict | list[dict]] = None
     ignore_eos: bool = False
     language: Optional[str] = None
+    request_id: Optional[str] = None
 
 
 @dataclass
@@ -71,6 +72,9 @@ async def async_request_tgi(
             "inputs": request_func_input.prompt,
             "parameters": params,
         }
+        headers = None
+        if request_func_input.request_id:
+            headers = {"x-request-id": request_func_input.request_id}
         output = RequestFuncOutput()
         output.prompt_len = request_func_input.prompt_len
         if request_func_input.ignore_eos:
@@ -82,7 +86,9 @@ async def async_request_tgi(
         st = time.perf_counter()
         most_recent_timestamp = st
         try:
-            async with session.post(url=api_url, json=payload) as response:
+            async with session.post(
+                url=api_url, json=payload, headers=headers
+            ) as response:
                 if response.status == 200:
                     async for chunk_bytes in response.content:
                         chunk_bytes = chunk_bytes.strip()
@@ -145,6 +151,9 @@ async def async_request_trt_llm(
         }
         if request_func_input.ignore_eos:
             payload["min_length"] = request_func_input.output_len
+        headers = None
+        if request_func_input.request_id:
+            headers = {"x-request-id": request_func_input.request_id}
         output = RequestFuncOutput()
         output.prompt_len = request_func_input.prompt_len
 
@@ -152,7 +161,9 @@ async def async_request_trt_llm(
         st = time.perf_counter()
         most_recent_timestamp = st
         try:
-            async with session.post(url=api_url, json=payload) as response:
+            async with session.post(
+                url=api_url, json=payload, headers=headers
+            ) as response:
                 if response.status == 200:
                     async for chunk_bytes in response.content:
                         chunk_bytes = chunk_bytes.strip()
@@ -211,6 +222,8 @@ async def async_request_deepspeed_mii(
             "top_p": 1.0,
         }
         headers = {"Authorization": f"Bearer {os.environ.get('OPENAI_API_KEY')}"}
+        if request_func_input.request_id:
+            headers["x-request-id"] = request_func_input.request_id
 
         output = RequestFuncOutput()
         output.prompt_len = request_func_input.prompt_len
@@ -283,6 +296,8 @@ async def async_request_openai_completions(
         if request_func_input.extra_body:
             payload.update(request_func_input.extra_body)
         headers = {"Authorization": f"Bearer {os.environ.get('OPENAI_API_KEY')}"}
+        if request_func_input.request_id:
+            headers["x-request-id"] = request_func_input.request_id
 
         output = RequestFuncOutput()
         output.prompt_len = request_func_input.prompt_len
@@ -395,6 +410,8 @@ async def async_request_openai_chat_completions(
             "Content-Type": "application/json",
             "Authorization": f"Bearer {os.environ.get('OPENAI_API_KEY')}",
         }
+        if request_func_input.request_id:
+            headers["x-request-id"] = request_func_input.request_id
 
         output = RequestFuncOutput()
         output.prompt_len = request_func_input.prompt_len
@@ -491,6 +508,8 @@ async def async_request_openai_audio(
         headers = {
             "Authorization": f"Bearer {os.environ.get('OPENAI_API_KEY')}",
         }
+        if request_func_input.request_id:
+            headers["x-request-id"] = request_func_input.request_id
 
         # Send audio file
         def to_bytes(y, sr):

benchmarks/benchmark_dataset.py

@@ -19,6 +19,7 @@ import logging
 import random
 from abc import ABC, abstractmethod
 from collections.abc import Mapping
+from copy import deepcopy
 from dataclasses import dataclass
 from functools import cache
 from io import BytesIO
@@ -54,6 +55,7 @@ class SampleRequest:
     expected_output_len: int
     multi_modal_data: Optional[Union[MultiModalDataDict, dict, list[dict]]] = None
     lora_request: Optional[LoRARequest] = None
+    request_id: Optional[str] = None
 
 
 # -----------------------------------------------------------------------------
@@ -155,7 +157,10 @@ class BenchmarkDataset(ABC):
 
     @abstractmethod
     def sample(
-        self, tokenizer: PreTrainedTokenizerBase, num_requests: int
+        self,
+        tokenizer: PreTrainedTokenizerBase,
+        num_requests: int,
+        request_id_prefix: str = "",
     ) -> list[SampleRequest]:
         """
         Abstract method to generate sample requests from the dataset.
@@ -167,6 +172,7 @@ class BenchmarkDataset(ABC):
             tokenizer (PreTrainedTokenizerBase): The tokenizer to be used
              for processing the dataset's text.
             num_requests (int): The number of sample requests to generate.
+            request_id_prefix (str) The prefix of request_id.
 
         Returns:
             list[SampleRequest]: A list of sample requests generated from the
@@ -175,7 +181,10 @@ class BenchmarkDataset(ABC):
         raise NotImplementedError("sample must be implemented in subclasses.")
 
     def maybe_oversample_requests(
-        self, requests: list[SampleRequest], num_requests: int
+        self,
+        requests: list[SampleRequest],
+        num_requests: int,
+        request_id_prefix: str = "",
     ) -> None:
         """
         Oversamples the list of requests if its size is less than the desired
@@ -183,11 +192,18 @@ class BenchmarkDataset(ABC):
 
         Args:
             requests (List[SampleRequest]): The current list of sampled
-            requests.  num_requests (int): The target number of requests.
+            requests.
+            num_requests (int): The target number of requests.
+            request_id_prefix (str) The prefix of the request ids.
         """
         if len(requests) < num_requests:
             random.seed(self.random_seed)
-            additional = random.choices(requests, k=num_requests - len(requests))
+            additional = deepcopy(
+                random.choices(requests, k=num_requests - len(requests))
+            )
+            for i in range(len(additional)):
+                req = additional[i]
+                req.request_id = request_id_prefix + str(len(requests) + i)
             requests.extend(additional)
             logger.info("Oversampled requests to reach %d total samples.", num_requests)
 
@@ -277,6 +293,41 @@ def process_image(image: Any) -> Mapping[str, Any]:
     )
 
 
+def process_video(video: Any) -> Mapping[str, Any]:
+    """
+    Process a single video input and return a multimedia content dictionary.
+
+    Supports the following input types:
+
+    1. Dictionary with raw video bytes: - Expects a dict with a 'bytes' key
+       containing raw video data.
+
+    2. String input: - Treats the string as a URL or local file path.  -
+       Prepends "file://" if the string doesn't start with "http://" or
+       "file://".  - Returns a dictionary with the image URL.
+
+    Raises:
+        ValueError: If the input is not a supported type.
+    """
+    if isinstance(video, dict) and "bytes" in video:
+        video_bytes = video["bytes"]
+        video_base64 = base64.b64encode(video_bytes).decode("utf-8")
+        return {
+            "type": "video_url",
+            "video_url": {"url": f"data:video/mp4;base64,{video_base64}"},
+        }
+
+    if isinstance(video, str):
+        video_url = (
+            video if video.startswith(("http://", "file://")) else f"file://{video}"
+        )
+        return {"type": "video_url", "video_url": {"url": video_url}}
+
+    raise ValueError(
+        f"Invalid video input {video}. Must be a string of local path/remote url, or a dictionary with raw video bytes in the form of `{{'bytes': raw_video_bytes}}`."  # noqa: E501
+    )
+
+
 # -----------------------------------------------------------------------------
 # Random Dataset Implementation (Synthetic Data)
 # -----------------------------------------------------------------------------
@@ -303,6 +354,7 @@ class RandomDataset(BenchmarkDataset):
         range_ratio: float = DEFAULT_RANGE_RATIO,
         input_len: int = DEFAULT_INPUT_LEN,
         output_len: int = DEFAULT_OUTPUT_LEN,
+        request_id_prefix: str = "",
         **kwargs,
     ) -> list[SampleRequest]:
         # Enforce range_ratio < 1
@@ -363,8 +415,10 @@ class RandomDataset(BenchmarkDataset):
                     prompt=prompt,
                     prompt_len=total_input_len,
                     expected_output_len=int(output_lens[i]),
+                    request_id=request_id_prefix + str(i),
                 )
             )
+
         return requests
 
 
@@ -406,9 +460,11 @@ class ShareGPTDataset(BenchmarkDataset):
         max_loras: Optional[int] = None,
         output_len: Optional[int] = None,
         enable_multimodal_chat: bool = False,
+        request_id_prefix: str = "",
         **kwargs,
     ) -> list:
         samples: list = []
+        ind = 0
         for entry in self.data:
             if len(samples) >= num_requests:
                 break
@@ -430,9 +486,10 @@ class ShareGPTDataset(BenchmarkDataset):
                 skip_min_output_len_check=output_len is not None,
             ):
                 continue
-            # TODO: Also support ShareGPT4Video.
             if image_path := entry.get("image"):
                 mm_content = process_image(image_path)
+            elif video_path := entry.get("video"):
+                mm_content = process_video(video_path)
             else:
                 mm_content = None
             if enable_multimodal_chat:
@@ -444,9 +501,11 @@ class ShareGPTDataset(BenchmarkDataset):
                     expected_output_len=new_output_len,
                     lora_request=lora_request,
                     multi_modal_data=mm_content,
+                    request_id=request_id_prefix + str(ind),
                 )
             )
-        self.maybe_oversample_requests(samples, num_requests)
+            ind += 1
+        self.maybe_oversample_requests(samples, num_requests, request_id_prefix)
         return samples
 
 
@@ -512,10 +571,11 @@ class CustomDataset(BenchmarkDataset):
         output_len: Optional[int] = None,
         enable_multimodal_chat: bool = False,
         skip_chat_template: bool = False,
+        request_id_prefix: str = "",
         **kwargs,
     ) -> list:
         sampled_requests = []
-        for item in self.data:
+        for i, item in enumerate(self.data):
             if len(sampled_requests) >= num_requests:
                 break
             prompt = item["prompt"]
@@ -534,9 +594,12 @@ class CustomDataset(BenchmarkDataset):
                     prompt=prompt,
                     prompt_len=prompt_len,
                     expected_output_len=output_len,
+                    request_id=request_id_prefix + str(i),
+                )
             )
+        self.maybe_oversample_requests(
+            sampled_requests, num_requests, request_id_prefix
         )
-        self.maybe_oversample_requests(sampled_requests, num_requests)
 
         return sampled_requests
 
@@ -578,6 +641,7 @@ class SonnetDataset(BenchmarkDataset):
         input_len: int = DEFAULT_INPUT_LEN,
         output_len: int = DEFAULT_OUTPUT_LEN,
         return_prompt_formatted: bool = False,
+        request_id_prefix: str = "",
         **kwargs,
     ) -> list:
         # Calculate average token length for a poem line.
@@ -603,6 +667,7 @@ class SonnetDataset(BenchmarkDataset):
         prefix_lines = self.data[:num_prefix_lines]
 
         samples = []
+        ind = 0
         while len(samples) < num_requests:
             extra_lines = random.choices(
                 self.data, k=num_input_lines - num_prefix_lines
@@ -613,14 +678,17 @@ class SonnetDataset(BenchmarkDataset):
                 msg, add_generation_prompt=True, tokenize=False
             )
             prompt_len = len(tokenizer(prompt_formatted).input_ids)
+
             if prompt_len <= input_len:
                 samples.append(
                     SampleRequest(
                         prompt=prompt_formatted if return_prompt_formatted else prompt,
                         prompt_len=prompt_len,
                         expected_output_len=output_len,
+                        request_id=request_id_prefix + str(ind),
                     )
                 )
+                ind += 1
         return samples
 
 
@@ -672,6 +740,7 @@ class BurstGPTDataset(BenchmarkDataset):
         num_requests: int,
         max_loras: Optional[int] = None,
         lora_path: Optional[str] = None,
+        request_id_prefix: str = "",
         **kwargs,
     ) -> list[SampleRequest]:
         samples = []
@@ -693,6 +762,7 @@ class BurstGPTDataset(BenchmarkDataset):
                     prompt_len=input_len,
                     expected_output_len=output_len,
                     lora_request=lora_req,
+                    request_id=request_id_prefix + str(i),
                 )
             )
         return samples
@@ -752,12 +822,14 @@ class ConversationDataset(HuggingFaceDataset):
         num_requests: int,
         output_len: Optional[int] = None,
         enable_multimodal_chat: bool = False,
+        request_id_prefix: str = "",
         **kwargs,
     ) -> list:
         # Filter examples with at least 2 conversations
         filtered_data = self.data.filter(lambda x: len(x["conversations"]) >= 2)
         sampled_requests = []
         dynamic_output = output_len is None
+        ind = 0
 
         for item in filtered_data:
             if len(sampled_requests) >= num_requests:
@@ -785,9 +857,13 @@ class ConversationDataset(HuggingFaceDataset):
                     prompt_len=prompt_len,
                     expected_output_len=output_len,
                     multi_modal_data=mm_content,
+                    request_id=request_id_prefix + str(ind),
                 )
             )
-        self.maybe_oversample_requests(sampled_requests, num_requests)
+            ind += 1
+        self.maybe_oversample_requests(
+            sampled_requests, num_requests, request_id_prefix
+        )
         return sampled_requests
 
 
@@ -814,11 +890,12 @@ class VisionArenaDataset(HuggingFaceDataset):
         num_requests: int,
         output_len: Optional[int] = None,
         enable_multimodal_chat: bool = False,
+        request_id_prefix: str = "",
         **kwargs,
     ) -> list:
         output_len = output_len if output_len is not None else self.DEFAULT_OUTPUT_LEN
         sampled_requests = []
-        for item in self.data:
+        for i, item in enumerate(self.data):
             if len(sampled_requests) >= num_requests:
                 break
             parser_fn = self.SUPPORTED_DATASET_PATHS.get(self.dataset_path)
@@ -838,9 +915,12 @@ class VisionArenaDataset(HuggingFaceDataset):
                     prompt_len=prompt_len,
                     expected_output_len=output_len,
                     multi_modal_data=mm_content,
+                    request_id=request_id_prefix + str(i),
+                )
             )
+        self.maybe_oversample_requests(
+            sampled_requests, num_requests, request_id_prefix
         )
-        self.maybe_oversample_requests(sampled_requests, num_requests)
         return sampled_requests
 
 
@@ -870,15 +950,18 @@ class InstructCoderDataset(HuggingFaceDataset):
         num_requests: int,
         output_len: Optional[int] = None,
         enable_multimodal_chat: bool = False,
+        request_id_prefix: str = "",
         **kwargs,
     ) -> list:
         output_len = output_len if output_len is not None else self.DEFAULT_OUTPUT_LEN
         sampled_requests = []
-        for item in self.data:
+        for i, item in enumerate(self.data):
             if len(sampled_requests) >= num_requests:
                 break
-            prompt = f"{item['input']}\n\n{item['instruction']} Just output \
-            the code, do not include any explanation."
+            prompt = (
+                f"{item['input']}\n\n{item['instruction']} Just output "
+                "the code, do not include any explanation."
+            )
 
             # apply template
             prompt = tokenizer.apply_chat_template(
@@ -892,9 +975,12 @@ class InstructCoderDataset(HuggingFaceDataset):
                     prompt=prompt,
                     prompt_len=prompt_len,
                     expected_output_len=output_len,
+                    request_id=request_id_prefix + str(i),
                 )
             )
-        self.maybe_oversample_requests(sampled_requests, num_requests)
+        self.maybe_oversample_requests(
+            sampled_requests, num_requests, request_id_prefix
+        )
         return sampled_requests
 
 
@@ -924,12 +1010,13 @@ class MTBenchDataset(HuggingFaceDataset):
         num_requests: int,
         output_len: Optional[int] = None,
         enable_multimodal_chat: bool = False,
+        request_id_prefix: str = "",
         **kwargs,
     ) -> list:
         output_len = output_len if output_len is not None else self.DEFAULT_OUTPUT_LEN
         sampled_requests = []
 
-        for item in self.data:
+        for i, item in enumerate(self.data):
             if len(sampled_requests) >= num_requests:
                 break
             prompt = item["turns"][0]
@@ -947,9 +1034,12 @@ class MTBenchDataset(HuggingFaceDataset):
                     prompt=prompt,
                     prompt_len=prompt_len,
                     expected_output_len=output_len,
+                    request_id=request_id_prefix + str(i),
+                )
             )
+        self.maybe_oversample_requests(
+            sampled_requests, num_requests, request_id_prefix
         )
-        self.maybe_oversample_requests(sampled_requests, num_requests)
         return sampled_requests
 
 
@@ -974,10 +1064,12 @@ class AIMODataset(HuggingFaceDataset):
         tokenizer: PreTrainedTokenizerBase,
         num_requests: int,
         output_len: Optional[int] = None,
+        request_id_prefix: str = "",
         **kwargs,
     ) -> list:
         sampled_requests = []
         dynamic_output = output_len is None
+        ind = 0
 
         for item in self.data:
             if len(sampled_requests) >= num_requests:
@@ -1000,9 +1092,13 @@ class AIMODataset(HuggingFaceDataset):
                     prompt_len=prompt_len,
                     expected_output_len=output_len,
                     multi_modal_data=None,
+                    request_id=request_id_prefix + str(ind),
                 )
             )
-        self.maybe_oversample_requests(sampled_requests, num_requests)
+            ind += 1
+        self.maybe_oversample_requests(
+            sampled_requests, num_requests, request_id_prefix
+        )
         return sampled_requests
 
 
@@ -1072,12 +1168,18 @@ class NextEditPredictionDataset(HuggingFaceDataset):
         "zed-industries/zeta": _format_zeta_prompt,
     }
 
-    def sample(self, tokenizer: PreTrainedTokenizerBase, num_requests: int, **kwargs):
+    def sample(
+        self,
+        tokenizer: PreTrainedTokenizerBase,
+        num_requests: int,
+        request_id_prefix: str = "",
+        **kwargs,
+    ):
         formatting_prompt_func = self.MAPPING_PROMPT_FUNCS.get(self.dataset_path)
         if formatting_prompt_func is None:
             raise ValueError(f"Unsupported dataset path: {self.dataset_path}")
         samples = []
-        for sample in self.data:
+        for i, sample in enumerate(self.data):
             sample = formatting_prompt_func(sample)
             samples.append(
                 SampleRequest(
@@ -1086,11 +1188,12 @@ class NextEditPredictionDataset(HuggingFaceDataset):
                     expected_output_len=len(
                         tokenizer(sample["expected_output"]).input_ids
                     ),
+                    request_id=request_id_prefix + str(i),
                 )
             )
             if len(samples) >= num_requests:
                 break
-        self.maybe_oversample_requests(samples, num_requests)
+        self.maybe_oversample_requests(samples, num_requests, request_id_prefix)
         return samples
 
 
@@ -1139,6 +1242,7 @@ class ASRDataset(HuggingFaceDataset):
         tokenizer: PreTrainedTokenizerBase,
         num_requests: int,
         output_len: Optional[int] = None,
+        request_id_prefix: str = "",
         **kwargs,
     ) -> list:
         import librosa
@@ -1148,6 +1252,7 @@ class ASRDataset(HuggingFaceDataset):
         prompt_len = len(tokenizer(prompt).input_ids)
         sampled_requests = []
         skipped = 0
+        ind = 0
         for item in self.data:
             if len(sampled_requests) >= num_requests:
                 break
@@ -1166,8 +1271,10 @@ class ASRDataset(HuggingFaceDataset):
                     prompt_len=prompt_len,
                     expected_output_len=output_len,
                     multi_modal_data=mm_content,
+                    request_id=request_id_prefix + str(ind),
                 )
             )
+            ind += 1
         if skipped:
             logger.warning(
                 "%d samples discarded from dataset due to"
@@ -1175,5 +1282,7 @@ class ASRDataset(HuggingFaceDataset):
                 " what Whisper supports.",
                 skipped,
             )
-        self.maybe_oversample_requests(sampled_requests, num_requests)
+        self.maybe_oversample_requests(
+            sampled_requests, num_requests, request_id_prefix
+        )
         return sampled_requests

benchmarks/benchmark_serving.py

@@ -375,11 +375,12 @@ async def benchmark(
                     rps_change_events.append({"rps": rps_val, "timestamp": timestamp})
                 last_int_rps = current_int_rps
 
-        prompt, prompt_len, output_len, mm_content = (
+        prompt, prompt_len, output_len, mm_content, request_id = (
             request.prompt,
             request.prompt_len,
             request.expected_output_len,
             request.multi_modal_data,
+            request.request_id,
         )
         req_model_id, req_model_name = model_id, model_name
         if lora_modules:
@@ -397,6 +398,7 @@ async def benchmark(
             multi_modal_content=mm_content,
             ignore_eos=ignore_eos,
             extra_body=extra_body,
+            request_id=request_id,
         )
         task = limited_request_func(request_func_input=request_func_input, pbar=pbar)
         tasks.append(asyncio.create_task(task))
@@ -665,6 +667,7 @@ def main(args: argparse.Namespace):
             tokenizer=tokenizer,
             output_len=args.custom_output_len,
             skip_chat_template=args.custom_skip_chat_template,
+            request_id_prefix=args.request_id_prefix,
         )
 
     elif args.dataset_name == "sonnet":
@@ -678,6 +681,7 @@ def main(args: argparse.Namespace):
                 prefix_len=args.sonnet_prefix_len,
                 tokenizer=tokenizer,
                 return_prompt_formatted=False,
+                request_id_prefix=args.request_id_prefix,
             )
         else:
             assert tokenizer.chat_template or tokenizer.default_chat_template, (
@@ -690,6 +694,7 @@ def main(args: argparse.Namespace):
                 prefix_len=args.sonnet_prefix_len,
                 tokenizer=tokenizer,
                 return_prompt_formatted=True,
+                request_id_prefix=args.request_id_prefix,
             )
 
     elif args.dataset_name == "hf":
@@ -751,6 +756,7 @@ def main(args: argparse.Namespace):
             num_requests=args.num_prompts,
             tokenizer=tokenizer,
             output_len=args.hf_output_len,
+            request_id_prefix=args.request_id_prefix,
         )
 
     else:
@@ -762,10 +768,15 @@ def main(args: argparse.Namespace):
                 tokenizer=tokenizer,
                 num_requests=args.num_prompts,
                 output_len=args.sharegpt_output_len,
+                request_id_prefix=args.request_id_prefix,
             ),
             "burstgpt": lambda: BurstGPTDataset(
                 random_seed=args.seed, dataset_path=args.dataset_path
-            ).sample(tokenizer=tokenizer, num_requests=args.num_prompts),
+            ).sample(
+                tokenizer=tokenizer,
+                num_requests=args.num_prompts,
+                request_id_prefix=args.request_id_prefix,
+            ),
             "random": lambda: RandomDataset(dataset_path=args.dataset_path).sample(
                 tokenizer=tokenizer,
                 num_requests=args.num_prompts,
@@ -773,6 +784,7 @@ def main(args: argparse.Namespace):
                 input_len=args.random_input_len,
                 output_len=args.random_output_len,
                 range_ratio=args.random_range_ratio,
+                request_id_prefix=args.request_id_prefix,
             ),
         }
 
@@ -1118,6 +1130,13 @@ def create_argument_parser():
         "goodput, refer to DistServe paper: https://arxiv.org/pdf/2401.09670 "
         "and the blog: https://hao-ai-lab.github.io/blogs/distserve",
     )
+    parser.add_argument(
+        "--request-id-prefix",
+        type=str,
+        required=False,
+        default="benchmark-serving",
+        help="Specify the prefix of request id.",
+    )
 
     # group for dataset specific arguments
     custom_group = parser.add_argument_group("custom dataset options")

benchmarks/benchmark_throughput.py

@@ -96,7 +96,6 @@ def run_vllm(
         end = time.perf_counter()
     else:
         assert lora_requests is None, "BeamSearch API does not support LoRA"
-        prompts = [request.prompt for request in requests]
         # output_len should be the same for all requests.
         output_len = requests[0].expected_output_len
         for request in requests:
@@ -597,8 +596,8 @@ def validate_args(args):
     # https://github.com/vllm-project/vllm/issues/16222
     if args.data_parallel_size > 1:
         raise ValueError(
-            "Data parallel is not supported in offline benchmark, \
-            please use benchmark serving instead"
+            "Data parallel is not supported in offline benchmark, "
+            "please use benchmark serving instead"
         )
 
 

benchmarks/kernels/bench_block_fp8_gemm.py

+# SPDX-License-Identifier: Apache-2.0
+# SPDX-FileCopyrightText: Copyright contributors to the vLLM project
+
+import torch
+
+from vllm.model_executor.layers.quantization.utils.fp8_utils import (
+    w8a8_block_fp8_matmul,
+)
+from vllm.platforms import current_platform
+from vllm.triton_utils import triton as vllm_triton
+
+assert current_platform.is_cuda(), (
+    "Only support benchmarking w8a8 block fp8 kernel on CUDA device."
+)
+
+# DeepSeek-V3 weight shapes
+DEEPSEEK_V3_SHAPES = [
+    (512 + 64, 7168),
+    (2112, 7168),
+    ((128 + 64) * 128, 7168),
+    (128 * (128 + 128), 512),
+    (7168, 16384),
+    (7168, 18432),
+    (18432 * 2, 7168),
+    (24576, 1536),
+    (12288, 7168),
+    (4096, 7168),
+    (7168, 2048),
+]
+
+
+def build_w8a8_block_fp8_runner(M, N, K, block_size, device):
+    """Build runner function for w8a8 block fp8 matmul."""
+    factor_for_scale = 1e-2
+
+    fp8_info = torch.finfo(torch.float8_e4m3fn)
+    fp8_max, fp8_min = fp8_info.max, fp8_info.min
+
+    # Create random FP8 tensors
+    A_fp32 = (torch.rand(M, K, dtype=torch.float32, device=device) - 0.5) * 2 * fp8_max
+    A = A_fp32.clamp(min=fp8_min, max=fp8_max).to(torch.float8_e4m3fn)
+
+    B_fp32 = (torch.rand(N, K, dtype=torch.float32, device=device) - 0.5) * 2 * fp8_max
+    B = B_fp32.clamp(min=fp8_min, max=fp8_max).to(torch.float8_e4m3fn)
+
+    # Create scales
+    block_n, block_k = block_size[0], block_size[1]
+    n_tiles = (N + block_n - 1) // block_n
+    k_tiles = (K + block_k - 1) // block_k
+
+    As = torch.rand(M, k_tiles, dtype=torch.float32, device=device) * factor_for_scale
+    Bs = (
+        torch.rand(n_tiles, k_tiles, dtype=torch.float32, device=device)
+        * factor_for_scale
+    )
+
+    def run():
+        return w8a8_block_fp8_matmul(A, B, As, Bs, block_size, torch.bfloat16)
+
+    return run
+
+
+@vllm_triton.testing.perf_report(
+    vllm_triton.testing.Benchmark(
+        x_names=["batch_size"],
+        x_vals=[1, 16, 64, 128, 256, 512, 1024, 2048, 4096, 8192, 16384],
+        x_log=False,
+        line_arg="provider",
+        line_vals=["torch-bf16", "w8a8-block-fp8"],
+        line_names=["torch-bf16", "w8a8-block-fp8"],
+        ylabel="TFLOP/s (larger is better)",
+        plot_name="BF16 vs W8A8 Block FP8 GEMMs",
+        args={},
+    )
+)
+def benchmark_tflops(batch_size, provider, N, K, block_size=(128, 128)):
+    M = batch_size
+    device = "cuda"
+
+    quantiles = [0.5, 0.2, 0.8]
+
+    if provider == "torch-bf16":
+        a = torch.randn((M, K), device=device, dtype=torch.bfloat16)
+        b = torch.randn((N, K), device=device, dtype=torch.bfloat16)
+        ms, min_ms, max_ms = vllm_triton.testing.do_bench_cudagraph(
+            lambda: torch.nn.functional.linear(a, b), quantiles=quantiles
+        )
+    else:  # w8a8-block-fp8
+        run_w8a8 = build_w8a8_block_fp8_runner(M, N, K, block_size, device)
+        ms, min_ms, max_ms = vllm_triton.testing.do_bench_cudagraph(
+            lambda: run_w8a8(), quantiles=quantiles
+        )
+
+    to_tflops = lambda t_ms: (2 * M * N * K) * 1e-12 / (t_ms * 1e-3)
+    return to_tflops(ms), to_tflops(max_ms), to_tflops(min_ms)
+
+
+if __name__ == "__main__":
+    block_size = (128, 128)
+
+    for N, K in DEEPSEEK_V3_SHAPES:
+        print(f"\nBenchmarking DeepSeek-V3, N={N} K={K}")
+
+        print(f"TFLOP/s comparison (block_size={block_size}):")
+        benchmark_tflops.run(
+            print_data=True,
+            # show_plots=False,
+            # save_path=f"bench_w8a8_block_fp8_tflops_n{N}_k{K}",
+            N=N,
+            K=K,
+            block_size=block_size,
+        )
+
+    print("\nBenchmark finished!")

benchmarks/kernels/benchmark_grouped_gemm_cutlass.py

@@ -80,6 +80,11 @@ def bench_run(
         a, score, topk, renormalize=False
     )
 
+    ab_strides1 = torch.full((num_experts,), k, device="cuda", dtype=torch.int64)
+    ab_strides2 = torch.full((num_experts,), n, device="cuda", dtype=torch.int64)
+    c_strides1 = torch.full((num_experts,), 2 * n, device="cuda", dtype=torch.int64)
+    c_strides2 = torch.full((num_experts,), k, device="cuda", dtype=torch.int64)
+
     def run_triton_moe(
         a: torch.Tensor,
         w1: torch.Tensor,
@@ -111,6 +116,10 @@ def bench_run(
         w2: torch.Tensor,
         w1_scale: torch.Tensor,
         w2_scale: torch.Tensor,
+        ab_strides1: torch.Tensor,
+        ab_strides2: torch.Tensor,
+        c_strides1: torch.Tensor,
+        c_strides2: torch.Tensor,
         topk_weights: torch.Tensor,
         topk_ids: torch.Tensor,
         per_act_token: bool,
@@ -125,6 +134,10 @@ def bench_run(
                 topk_ids,
                 w1_scale,
                 w2_scale,
+                ab_strides1,
+                ab_strides2,
+                c_strides1,
+                c_strides2,
                 per_act_token,
                 a1_scale=None,
             )
@@ -136,6 +149,10 @@ def bench_run(
         w2_q: torch.Tensor,
         w1_scale: torch.Tensor,
         w2_scale: torch.Tensor,
+        ab_strides1: torch.Tensor,
+        ab_strides2: torch.Tensor,
+        c_strides1: torch.Tensor,
+        c_strides2: torch.Tensor,
         topk_weights: torch.Tensor,
         topk_ids: torch.Tensor,
     ):
@@ -150,6 +167,10 @@ def bench_run(
                 topk_ids,
                 w1_scale,
                 w2_scale,
+                ab_strides1,
+                ab_strides2,
+                c_strides1,
+                c_strides2,
                 per_act_token,
                 a1_scale=None,
             )
@@ -194,6 +215,10 @@ def bench_run(
             w2_q,
             w1_scale,
             w2_scale,
+            ab_strides1,
+            ab_strides2,
+            c_strides1,
+            c_strides2,
             topk_weights,
             topk_ids,
         )
@@ -231,6 +256,10 @@ def bench_run(
         "w1_scale": w1_scale,
         "w2_scale": w2_scale,
         "per_act_token": per_act_token,
+        "ab_strides1": ab_strides1,
+        "ab_strides2": ab_strides2,
+        "c_strides1": c_strides1,
+        "c_strides2": c_strides2,
         # cuda graph params
         "cutlass_graph": cutlass_graph,
         "triton_graph": triton_graph,
@@ -289,6 +318,10 @@ def bench_run(
         w2_q,
         w1_scale,
         w2_scale,
+        ab_strides1,
+        ab_strides2,
+        c_strides1,
+        c_strides2,
         topk_weights,
         topk_ids,
         per_act_token,
@@ -297,7 +330,7 @@ def bench_run(
 
     results.append(
         benchmark.Timer(
-            stmt="run_cutlass_moe(a, a_scale, w1_q, w2_q, w1_scale, w2_scale, topk_weights, topk_ids, per_act_token, num_runs)",  # noqa: E501
+            stmt="run_cutlass_moe(a, a_scale, w1_q, w2_q, w1_scale, w2_scale, ab_strides1, ab_strides2, c_strides1, c_strides2, topk_weights, topk_ids, per_act_token, num_runs)",  # noqa: E501
             globals=globals,
             label=label,
             sub_label=sub_label,

benchmarks/kernels/benchmark_machete.py

@@ -253,28 +253,7 @@ def marlin_create_bench_fn(bt: BenchmarkTensors) -> Callable:
     else:
         assert bt.a.dtype == torch.int8
         assert bt.wtype == scalar_types.uint4b8
-
-        if bt.w_ch_s is not None:
-            s_ch = bt.w_ch_s.to(torch.float32)
-        else:
-            s_ch = torch.ones(bt.w_ref.shape[1], dtype=torch.float32, device=device)
-
-        if bt.w_tok_s is not None:
-            s_tok = bt.w_tok_s.to(torch.float32)
-        else:
-            s_tok = torch.ones(bt.a.shape[0], dtype=torch.float32, device=device)
-
-        fn = lambda: ops.marlin_qqq_gemm(
-            a=bt.a,
-            b_q_weight=w_q,
-            s_group=w_s,
-            s_tok=s_tok,
-            s_ch=s_ch,
-            workspace=workspace.scratch,
-            size_m=bt.a.shape[0],
-            size_n=bt.w_ref.shape[1],
-            size_k=bt.w_ref.shape[0],
-        )
+        raise NotImplementedError("QQQ is not supported anymore")
 
     return fn
 
@@ -305,6 +284,25 @@ def machete_create_bench_fn(
     )
 
 
+def cutlass_w4a8_create_bench_fn(
+    bt: BenchmarkTensors, out_type=torch.dtype, schedule=None
+) -> Callable:
+    w_q = bt.w_q.t().contiguous().t()  # make col major
+    w_q = ops.cutlass_encode_and_reorder_int4b(w_q)
+    # expects fp8 scales
+    w_s = ops.cutlass_pack_scale_fp8(bt.w_g_s.to(torch.float8_e4m3fn))
+
+    return lambda: ops.cutlass_w4a8_mm(
+        a=bt.a,
+        b_q=w_q,
+        b_group_scales=w_s,
+        b_group_size=bt.group_size,
+        b_channel_scales=bt.w_ch_s,
+        a_token_scales=bt.w_tok_s,
+        maybe_schedule=schedule,
+    )
+
+
 # impl
 
 # bench
@@ -406,6 +404,20 @@ def bench(
         )
     )
 
+    # cutlass w4a8
+    if types.act_type == torch.float8_e4m3fn and group_size == 128:
+        timers.append(
+            bench_fns(
+                label,
+                sub_label,
+                f"cutlass w4a8 ({name_type_string})",
+                [
+                    cutlass_w4a8_create_bench_fn(bt, out_type=types.output_type)
+                    for bt in benchmark_tensors
+                ],
+            )
+        )
+
     if sweep_schedules:
         global _SWEEP_SCHEDULES_RESULTS
 

benchmarks/kernels/benchmark_moe.py

@@ -419,8 +419,10 @@ class BenchmarkWorker:
         )
         # NOTE(woosuk): The current naming convention uses w2.shape[2], which
         # is the intermediate size after silu_and_mul.
+        block_n = block_quant_shape[0] if block_quant_shape else None
+        block_k = block_quant_shape[1] if block_quant_shape else None
         op_config = get_moe_configs(
-            num_experts, shard_intermediate_size // 2, dtype_str
+            num_experts, shard_intermediate_size // 2, dtype_str, block_n, block_k
         )
         if op_config is None:
             config = get_default_config(
@@ -430,7 +432,7 @@ class BenchmarkWorker:
                 hidden_size,
                 topk,
                 dtype_str,
-                is_marlin=False,
+                block_quant_shape,
             )
         else:
             config = op_config[min(op_config.keys(), key=lambda x: abs(x - num_tokens))]

benchmarks/kernels/benchmark_silu_mul_fp8_quant.py

+#!/usr/bin/env python3
+# SPDX-License-Identifier: Apache-2.0
+# SPDX-FileCopyrightText: Copyright contributors to the vLLM project
+import time
+
+import torch
+
+from vllm.model_executor.layers.fused_moe.batched_deep_gemm_moe import (
+    silu_mul_fp8_quant_deep_gemm,
+)
+from vllm.platforms import current_platform
+
+
+def benchmark(E, T, H, G=128, runs=50):
+    current_platform.seed_everything(42)
+    y = torch.randn((E, T, 2 * H), dtype=torch.bfloat16, device="cuda")
+    tokens_per_expert = torch.randint(
+        T // 2, T, size=(E,), dtype=torch.int32, device="cuda"
+    )
+
+    # Warmup
+    for _ in range(10):
+        silu_mul_fp8_quant_deep_gemm(y, tokens_per_expert, group_size=G)
+        torch.cuda.synchronize()
+
+    # Benchmark
+    torch.cuda.synchronize()
+    start = time.perf_counter()
+    for _ in range(runs):
+        silu_mul_fp8_quant_deep_gemm(y, tokens_per_expert, group_size=G)
+    torch.cuda.synchronize()
+
+    avg_time = (time.perf_counter() - start) / runs * 1000
+
+    # Calculate actual work done (only count valid tokens)
+    actual_tokens = tokens_per_expert.sum().item()
+    actual_elements = actual_tokens * H
+
+    # GFLOPS: operations per element = exp + 3 muls + 1 div + quantization ops ≈ 8 ops
+    ops_per_element = 8
+    total_ops = actual_elements * ops_per_element
+    gflops = total_ops / (avg_time / 1000) / 1e9
+
+    # Memory bandwidth: bfloat16 inputs (2 bytes), fp8 output (1 byte), scales (4 bytes)
+    input_bytes = actual_tokens * 2 * H * 2  # 2*H bfloat16 inputs
+    output_bytes = actual_tokens * H * 1  # H fp8 outputs
+    scale_bytes = actual_tokens * (H // G) * 4  # scales in float32
+    total_bytes = input_bytes + output_bytes + scale_bytes
+    memory_bw = total_bytes / (avg_time / 1000) / 1e9
+
+    return avg_time, gflops, memory_bw
+
+
+configs = [
+    (8, 32, 1024),
+    (16, 64, 2048),
+    (32, 128, 4096),
+    # DeepSeekV3 Configs
+    (256, 16, 7168),
+    (256, 32, 7168),
+    (256, 64, 7168),
+    (256, 128, 7168),
+    (256, 256, 7168),
+    (256, 512, 7168),
+    (256, 1024, 7168),
+]
+
+print(f"GPU: {torch.cuda.get_device_name()}")
+print(f"{'Config':<20} {'Time(ms)':<10} {'GFLOPS':<10} {'GB/s':<10}")
+print("-" * 50)
+
+for E, T, H in configs:
+    try:
+        time_ms, gflops, gbps = benchmark(E, T, H)
+        print(f"E={E:3d},T={T:4d},H={H:4d} {time_ms:8.3f} {gflops:8.1f} {gbps:8.1f}")
+    except Exception:
+        print(f"E={E:3d},T={T:4d},H={H:4d} FAILED")

benchmarks/kernels/benchmark_trtllm_decode_attention.py

@@ -3,16 +3,17 @@
 
 import csv
 import os
-import random
 from datetime import datetime
+from typing import Optional
 
 import flashinfer
 import torch
 
-FLOAT32_BYTES = torch.finfo(torch.float).bits // 8
+from vllm.utils import round_up
 
-# KV Cache Layout for TRT-LLM
-# kv_cache_shape = (num_blocks, 2, num_kv_heads, page_size, head_dim)
+FLOAT32_BYTES = torch.finfo(torch.float).bits // 8
+FP8_DTYPE = torch.float8_e4m3fn
+FP4_DTYPE = torch.uint8
 
 
 def to_float8(x, dtype=torch.float8_e4m3fn):
@@ -26,149 +27,188 @@ def to_float8(x, dtype=torch.float8_e4m3fn):
 
 @torch.no_grad()
 def benchmark_decode(
-    num_seqs,
-    max_seq_len,
-    page_size=16,
-    dtype=torch.bfloat16,
-    kv_layout="HND",
-    num_kv_heads=8,
-    kv_cache_dtype="auto",
-    head_dim=128,
-    warmup=10,
-    trials=20,
+    dtype: torch.dtype,
+    quant_dtypes: tuple[
+        Optional[torch.dtype], Optional[torch.dtype], Optional[torch.dtype]
+    ],
+    batch_size: int,
+    max_seq_len: int,
+    num_heads: tuple[int, int] = (64, 8),
+    head_size: int = 128,
+    kv_layout: str = "HND",
+    block_size: int = 16,
+    warmup: int = 10,
+    trials: int = 20,
 ):
     torch.set_default_device("cuda")
-    device = "cuda"
     torch.manual_seed(0)
 
-    HEAD_GRP_SIZE = 8
-    MAX_SEQ_LEN = max_seq_len
-
-    # large number to reduce kv_cache reuse
-    NUM_BLOCKS = int(256000 / page_size)
-
-    workspace_buffer = torch.empty(1024 * 1024 * 1024, dtype=torch.int8, device=device)
+    q_quant_dtype, kv_quant_dtype, o_quant_dtype = quant_dtypes
+    q_quant_dtype = q_quant_dtype or dtype
+    kv_quant_dtype = kv_quant_dtype or dtype
+    o_quant_dtype = o_quant_dtype or dtype
 
-    # For decode, batch_size is num_decode_token
-    num_qo_heads = num_kv_heads * HEAD_GRP_SIZE
-    sm_scale = float(1.0 / (head_dim**0.5))
-    q = torch.randn(num_seqs, num_qo_heads, head_dim, device=device, dtype=dtype)
-    kv_lens = [random.randint(1, MAX_SEQ_LEN) for _ in range(num_seqs)]
+    num_qo_heads, num_kv_heads = num_heads
+    assert num_qo_heads % num_kv_heads == 0
 
-    max_kv_len = max(kv_lens)
-    kv_lens_tensor = torch.tensor(kv_lens, dtype=torch.int, device=device)
-    max_num_blocks_per_seq = (max_kv_len + page_size - 1) // page_size
+    sm_scale = float(1.0 / (head_size**0.5))
 
-    block_tables = torch.randint(
-        0, NUM_BLOCKS, (num_seqs, max_num_blocks_per_seq), dtype=torch.int32
-    )
-
-    kv_cache_shape = (NUM_BLOCKS, 2, num_kv_heads, page_size, head_dim)
-    kv_cache = torch.randn(size=kv_cache_shape, device=device, dtype=dtype)
+    # large number to reduce kv_cache reuse
+    NUM_BLOCKS = int(256000 / block_size)
+
+    kv_cache_shape = None
+    if kv_layout == "NHD":
+        kv_cache_shape = (NUM_BLOCKS, 2, block_size, num_kv_heads, head_size)
+    elif kv_layout == "HND":
+        kv_cache_shape = (NUM_BLOCKS, 2, num_kv_heads, block_size, head_size)
+    else:
+        raise ValueError(f"Invalid kv_layout: {kv_layout}")
+
+    # Always using 1.0 scale to reflect the real perf in benchmarking
+    q_scale = 1.0
+    ref_query = torch.randn(batch_size, num_qo_heads, head_size, dtype=dtype)
+    if q_quant_dtype == FP8_DTYPE:
+        query, _ = to_float8(ref_query)
+    else:
+        query = ref_query
+
+    kv_lens = torch.randint(1, max_seq_len, (batch_size,), dtype=torch.int32)
+    kv_lens[-1] = max_seq_len
+
+    seq_lens = kv_lens
+    max_seq_len = torch.max(seq_lens).item()
+
+    # Always using 1.0 scale to reflect the real perf in benchmarking
     k_scale = v_scale = 1.0
+    ref_kv_cache = torch.randn(kv_cache_shape, dtype=dtype)
+    if kv_quant_dtype == FP8_DTYPE:
+        kv_cache, _ = to_float8(ref_kv_cache)
+    else:
+        kv_cache = ref_kv_cache
 
-    if kv_cache_dtype.startswith("fp8"):
-        kv_cache, _ = to_float8(kv_cache)
-
-    output_trtllm = torch.empty(q.shape, dtype=dtype)
-
-    # Benchmark TRT decode
-    def trt_decode():
-        return flashinfer.decode.trtllm_batch_decode_with_kv_cache(
-            q,
-            kv_cache,
-            workspace_buffer,
-            block_tables,
-            kv_lens_tensor,
-            max_kv_len,
-            bmm1_scale=k_scale * sm_scale,
-            bmm2_scale=v_scale,
-            out=output_trtllm,
+    max_num_blocks_per_seq = (max_seq_len + block_size - 1) // block_size
+    block_tables = torch.randint(
+        0, NUM_BLOCKS, (batch_size, max_num_blocks_per_seq), dtype=torch.int32
     )
-
-    def time_fn(fn, warmup=10, trials=20):
-        torch.cuda.synchronize()
-        start = torch.cuda.Event(enable_timing=True)
-        end = torch.cuda.Event(enable_timing=True)
-        times = []
-        for i in range(warmup):
-            fn()
-        for i in range(trials):
-            start.record()
-            fn()
-            end.record()
-            torch.cuda.synchronize()
-            times.append(start.elapsed_time(end))  # ms
-        return sum(times) / len(times), torch.std(torch.tensor(times))
-
-    # TRT Decode
-    trt_mean, trt_std = time_fn(trt_decode)
-
     kv_indptr = [0]
     kv_indices = []
     kv_last_page_lens = []
-    for i in range(num_seqs):
-        seq_len = kv_lens[i]
+    for i in range(batch_size):
+        seq_len = seq_lens[i]
         assert seq_len > 0
-        num_blocks = (seq_len + page_size - 1) // page_size
+        num_blocks = (seq_len + block_size - 1) // block_size
         kv_indices.extend(block_tables[i, :num_blocks])
         kv_indptr.append(kv_indptr[-1] + num_blocks)
-        kv_last_page_len = seq_len % page_size
+        kv_last_page_len = seq_len % block_size
         if kv_last_page_len == 0:
-            kv_last_page_len = page_size
+            kv_last_page_len = block_size
         kv_last_page_lens.append(kv_last_page_len)
 
     kv_indptr = torch.tensor(kv_indptr, dtype=torch.int32)
     kv_indices = torch.tensor(kv_indices, dtype=torch.int32)
     kv_last_page_lens = torch.tensor(kv_last_page_lens, dtype=torch.int32)
-
-    output_baseline = torch.empty(q.shape, dtype=dtype)
+    workspace_buffer = torch.zeros(1024 * 1024 * 1024, dtype=torch.int8)
 
     wrapper = flashinfer.BatchDecodeWithPagedKVCacheWrapper(
         workspace_buffer,
         kv_layout,
-        use_tensor_cores=((num_qo_heads // num_kv_heads) > 4),
+        use_tensor_cores=True,
     )
-
     wrapper.plan(
         kv_indptr,
         kv_indices,
         kv_last_page_lens,
         num_qo_heads,
         num_kv_heads,
-        head_dim,
-        page_size,
+        head_size,
+        block_size,
         "NONE",
+        sm_scale=sm_scale,
         q_data_type=dtype,
-        kv_data_type=torch.float8_e4m3fn if kv_cache_dtype.startswith("fp8") else dtype,
+        kv_data_type=dtype,
     )
 
+    def time_fn(fn, warmup=10, trials=20):
+        torch.cuda.synchronize()
+        start = torch.cuda.Event(enable_timing=True)
+        end = torch.cuda.Event(enable_timing=True)
+        times = []
+        for i in range(warmup):
+            fn()
+        for i in range(trials):
+            start.record()
+            fn()
+            end.record()
+            torch.cuda.synchronize()
+            times.append(start.elapsed_time(end))  # ms
+        return sum(times) / len(times), torch.std(torch.tensor(times))
+
+    o_scale = 1.0
+    o_sf_scale = None
+    output_baseline = torch.empty(ref_query.shape, dtype=dtype)
+    if o_quant_dtype == FP4_DTYPE:
+        o_sf_scale = 500.0
+        output_trtllm = flashinfer.utils.FP4Tensor(
+            torch.empty(query.shape[:-1] + (query.shape[-1] // 2,), dtype=torch.uint8),
+            torch.empty(
+                (
+                    round_up(query.shape[0], 128),
+                    round_up(query.shape[1] * query.shape[2] // 16, 4),
+                ),
+                dtype=torch.float8_e4m3fn,
+            ),
+        )
+    else:
+        output_trtllm = torch.empty(query.shape, dtype=o_quant_dtype)
+
     def baseline_decode():
-        return wrapper.run(q, kv_cache, sm_scale, k_scale, v_scale, output_baseline)
+        return wrapper.run(
+            ref_query,
+            ref_kv_cache,
+            k_scale=k_scale,
+            v_scale=v_scale,
+            out=output_baseline,
+        )
+
+    def trtllm_decode():
+        return flashinfer.decode.trtllm_batch_decode_with_kv_cache(
+            query=query,
+            kv_cache=kv_cache,
+            workspace_buffer=workspace_buffer,
+            block_tables=block_tables,
+            seq_lens=seq_lens,
+            max_seq_len=max_seq_len,
+            bmm1_scale=q_scale * k_scale * sm_scale,
+            bmm2_scale=v_scale / o_scale,
+            o_sf_scale=o_sf_scale,
+            out=output_trtllm,
+        )
 
     baseline_mean, baseline_std = time_fn(baseline_decode)
+    trtllm_mean, trtllm_std = time_fn(trtllm_decode)
 
     # Calculate percentage speedup (positive means TRT is faster)
-    speedup_percent = (baseline_mean - trt_mean) / baseline_mean
+    speedup_percent = (baseline_mean - trtllm_mean) / baseline_mean
 
     print(
-        f"\t{num_seqs}\t{max_seq_len}\t{trt_mean:.3f}\t{trt_std.item():.3f}"
+        f"\t{batch_size}\t{max_seq_len}\t{trtllm_mean:.3f}\t{trtllm_std.item():.3f}"
         f"\t{baseline_mean:.3f}\t{baseline_std.item():.3f}\t{speedup_percent:.3f}"
     )
 
     # Return results for CSV writing
     return {
-        "num_seqs": num_seqs,
-        "trt_mean": trt_mean,
-        "trt_std": trt_std.item(),
+        "batch_size": batch_size,
+        "trtllm_mean": trtllm_mean,
+        "trtllm_std": trtllm_std.item(),
         "baseline_mean": baseline_mean,
         "baseline_std": baseline_std.item(),
         "speedup_percent": speedup_percent,
-        "q_dtype": str(dtype),
-        "kv_cache_dtype": kv_cache_dtype,
-        "page_size": page_size,
+        "q_dtype": str(q_quant_dtype),
+        "kv_cache_dtype": str(kv_quant_dtype),
+        "output_dtype": str(o_quant_dtype),
+        "block_size": block_size,
         "num_kv_heads": num_kv_heads,
-        "head_dim": head_dim,
+        "head_size": head_size,
         "max_seq_len": max_seq_len,
     }
 
@@ -180,17 +220,18 @@ def write_results_to_csv(results, filename=None):
         filename = f"flashinfer_trtllm_benchmark_{timestamp}.csv"
 
     fieldnames = [
-        "num_seqs",
-        "trt_mean",
-        "trt_std",
+        "batch_size",
+        "trtllm_mean",
+        "trtllm_std",
         "baseline_mean",
         "baseline_std",
         "speedup_percent",
         "q_dtype",
         "kv_cache_dtype",
-        "page_size",
+        "output_dtype",
+        "block_size",
         "num_kv_heads",
-        "head_dim",
+        "head_size",
         "max_seq_len",
     ]
 
@@ -209,43 +250,41 @@ def write_results_to_csv(results, filename=None):
 
 
 if __name__ == "__main__":
-    num_seqs = [1, 4, 8, 16, 32, 64, 128, 256]
+    batch_sizes = [1, 4, 8, 16, 32, 64, 128, 256]
     max_seq_lens = [1024, 2048, 4096, 8192, 16384, 32768, 65536, 131072]
     all_results = []
 
-    print(
-        "Running benchmark for q_dtype = bfloat16, kv_cache_dtype: bfloat16, "
-        "output_dtype: bfloat16"
-    )
-    print(
-        "\tnum_seqs\tmax_seq_len\ttrt_mean\ttrt_std\tbaseline_mean\t"
-        "baseline_std\tspeedup_percent"
-    )
-    for max_seq_len in max_seq_lens:
-        for bs in num_seqs:
-            result = benchmark_decode(
-                bs,
-                max_seq_len,
-                dtype=torch.bfloat16,
-                kv_cache_dtype="auto",
-            )
-            all_results.append(result)
+    dtype = torch.bfloat16
+    quant_dtypes = [
+        # (q_quant_dtype, kv_quant_dtype, o_quant_dtype)
+        (None, None, None),
+        (None, FP8_DTYPE, None),
+        (FP8_DTYPE, FP8_DTYPE, FP8_DTYPE),
+        (FP8_DTYPE, FP8_DTYPE, FP4_DTYPE),
+    ]
+
+    for quant_dtype in quant_dtypes:
+        q_quant_dtype, kv_quant_dtype, o_quant_dtype = quant_dtype
+        q_quant_dtype = q_quant_dtype or dtype
+        kv_quant_dtype = kv_quant_dtype or dtype
+        o_quant_dtype = o_quant_dtype or dtype
 
         print(
-        "Running benchmark for q_dtype = bfloat16, kv_cache_dtype: fp8, "
-        "output_dtype: bfloat16"
+            f"Running benchmark for q_dtype = {q_quant_dtype}, "
+            f"kv_cache_dtype: {kv_quant_dtype}, "
+            f"output_dtype: {o_quant_dtype}"
         )
         print(
-        "\tnum_seqs\tmax_seq_len\ttrt_mean\ttrt_std\tbaseline_mean\t"
+            "\tbatch_size\tmax_seq_len\ttrtllm_mean\ttrtllm_std\tbaseline_mean\t"
             "baseline_std\tspeedup_percent"
         )
         for max_seq_len in max_seq_lens:
-        for bs in num_seqs:
+            for bs in batch_sizes:
                 result = benchmark_decode(
-                bs,
-                max_seq_len,
-                dtype=torch.bfloat16,
-                kv_cache_dtype="fp8",
+                    dtype=dtype,
+                    quant_dtypes=quant_dtype,
+                    batch_size=bs,
+                    max_seq_len=max_seq_len,
                 )
                 all_results.append(result)