Docs: use custom directive to reference library versions

Fix important block from CONTRIBUTING.md

.azuredevops/components/AMDMIGraphX.yml

@@ -84,6 +84,8 @@ jobs:
     matrix:
       gfx942:
         JOB_GPU_TARGET: gfx942
+      gfx90a:
+        JOB_GPU_TARGET: gfx90a
   steps:
   - template: ${{ variables.CI_TEMPLATE_PATH }}/steps/dependencies-other.yml
     parameters:

.azuredevops/components/MIOpen.yml

@@ -67,6 +67,8 @@ jobs:
     matrix:
       gfx942:
         JOB_GPU_TARGET: gfx942
+      gfx90a:
+        JOB_GPU_TARGET: gfx90a
   steps:
   - template: ${{ variables.CI_TEMPLATE_PATH }}/steps/dependencies-other.yml
     parameters:

.azuredevops/components/MIVisionX.yml

@@ -77,6 +77,8 @@ jobs:
     matrix:
       gfx942:
         JOB_GPU_TARGET: gfx942
+      gfx90a:
+        JOB_GPU_TARGET: gfx90a
   steps:
   - template: ${{ variables.CI_TEMPLATE_PATH }}/steps/dependencies-other.yml
     parameters:

.azuredevops/components/ROCmValidationSuite.yml

@@ -67,6 +67,8 @@ jobs:
     matrix:
       gfx942:
         JOB_GPU_TARGET: gfx942
+      gfx90a:
+        JOB_GPU_TARGET: gfx90a
   steps:
   - template: ${{ variables.CI_TEMPLATE_PATH }}/steps/dependencies-other.yml
     parameters:

.azuredevops/components/Tensile.yml

@@ -87,7 +87,6 @@ jobs:
     parameters:
       aptPackages: ${{ parameters.aptPackages }}
       pipModules: ${{ parameters.pipModules }}
-      gpuTarget: $(JOB_GPU_TARGET)
 
 - job: Tensile_testing
   timeoutInMinutes: 90

.azuredevops/components/TransferBench.yml

@@ -42,6 +42,8 @@ jobs:
     matrix:
       gfx942:
         JOB_GPU_TARGET: gfx942
+      gfx90a:
+        JOB_GPU_TARGET: gfx90a
   steps:
   - template: ${{ variables.CI_TEMPLATE_PATH }}/steps/dependencies-other.yml
     parameters:

.azuredevops/components/composable_kernel.yml

@@ -48,6 +48,8 @@ jobs:
     matrix:
       gfx942:
         JOB_GPU_TARGET: gfx942
+      gfx90a:
+        JOB_GPU_TARGET: gfx90a
   steps:
   - template: ${{ variables.CI_TEMPLATE_PATH }}/steps/dependencies-other.yml
     parameters:

.azuredevops/components/hip-tests.yml

@@ -52,6 +52,8 @@ jobs:
     matrix:
       gfx942:
         JOB_GPU_TARGET: gfx942
+      gfx90a:
+        JOB_GPU_TARGET: gfx90a
   steps:
   - template: ${{ variables.CI_TEMPLATE_PATH }}/steps/dependencies-other.yml
     parameters:

.azuredevops/components/hipBLAS.yml

@@ -63,6 +63,8 @@ jobs:
     matrix:
       gfx942:
         JOB_GPU_TARGET: gfx942
+      gfx90a:
+        JOB_GPU_TARGET: gfx90a
   steps:
   - template: ${{ variables.CI_TEMPLATE_PATH }}/steps/dependencies-other.yml
     parameters:

.azuredevops/components/hipBLASLt.yml

@@ -72,6 +72,8 @@ jobs:
     matrix:
       gfx942:
         JOB_GPU_TARGET: gfx942
+      gfx90a:
+        JOB_GPU_TARGET: gfx90a
   steps:
   - template: ${{ variables.CI_TEMPLATE_PATH }}/steps/dependencies-other.yml
     parameters:
@@ -156,6 +158,7 @@ jobs:
         - deps
 
 - job: hipBLASLt_testing
+  timeoutInMinutes: 120
   dependsOn: hipBLASLt
   condition: and(succeeded(), eq(variables.ENABLE_GFX942_TESTS, 'true'), not(containsValue(split(variables.DISABLED_GFX942_TESTS, ','), variables['Build.DefinitionName'])))
   variables:

.azuredevops/components/hipCUB.yml

@@ -43,6 +43,8 @@ jobs:
     matrix:
       gfx942:
         JOB_GPU_TARGET: gfx942
+      gfx90a:
+        JOB_GPU_TARGET: gfx90a
   steps:
   - template: ${{ variables.CI_TEMPLATE_PATH }}/steps/dependencies-other.yml
     parameters:

.azuredevops/components/hipFFT.yml

@@ -54,6 +54,8 @@ jobs:
     matrix:
       gfx942:
         JOB_GPU_TARGET: gfx942
+      gfx90a:
+        JOB_GPU_TARGET: gfx90a
   steps:
   - template: ${{ variables.CI_TEMPLATE_PATH }}/steps/dependencies-other.yml
     parameters:

.azuredevops/components/hipRAND.yml

@@ -45,6 +45,8 @@ jobs:
     matrix:
       gfx942:
         JOB_GPU_TARGET: gfx942
+      gfx90a:
+        JOB_GPU_TARGET: gfx90a
   steps:
   - template: ${{ variables.CI_TEMPLATE_PATH }}/steps/dependencies-other.yml
     parameters:

.azuredevops/components/hipSOLVER.yml

@@ -57,6 +57,8 @@ jobs:
     matrix:
       gfx942:
         JOB_GPU_TARGET: gfx942
+      gfx90a:
+        JOB_GPU_TARGET: gfx90a
   steps:
   - template: ${{ variables.CI_TEMPLATE_PATH }}/steps/dependencies-other.yml
     parameters:

.azuredevops/components/hipSPARSE.yml

@@ -52,6 +52,8 @@ jobs:
     matrix:
       gfx942:
         JOB_GPU_TARGET: gfx942
+      gfx90a:
+        JOB_GPU_TARGET: gfx90a
   steps:
   - template: ${{ variables.CI_TEMPLATE_PATH }}/steps/dependencies-other.yml
     parameters:

.azuredevops/components/hipSPARSELt.yml

@@ -105,6 +105,7 @@ jobs:
         -DCMAKE_BUILD_TYPE=Release
         -DCMAKE_CXX_COMPILER=$(Agent.BuildDirectory)/rocm/llvm/bin/amdclang++
         -DCMAKE_C_COMPILER=$(Agent.BuildDirectory)/rocm/llvm/bin/amdclang
+        -DCMAKE_Fortran_COMPILER=f95
         -DAMDGPU_TARGETS=$(JOB_GPU_TARGET)
         -DTensile_LOGIC=
         -DTensile_CPU_THREADS=

.azuredevops/components/hipTensor.yml

@@ -42,6 +42,8 @@ jobs:
     matrix:
       gfx942:
         JOB_GPU_TARGET: gfx942
+      gfx90a:
+        JOB_GPU_TARGET: gfx90a
   steps:
   - template: ${{ variables.CI_TEMPLATE_PATH }}/steps/dependencies-other.yml
     parameters:

.azuredevops/components/hipfort.yml

@@ -51,6 +51,8 @@ jobs:
     matrix:
       gfx942:
         JOB_GPU_TARGET: gfx942
+      gfx90a:
+        JOB_GPU_TARGET: gfx90a
   steps:
   - template: ${{ variables.CI_TEMPLATE_PATH }}/steps/dependencies-other.yml
     parameters:

.azuredevops/components/llvm-project.yml

@@ -29,7 +29,7 @@ jobs:
     value: '$(Build.BinariesDirectory)/amdgcn/bitcode'
   - name: HIP_PATH
     value: '$(Agent.BuildDirectory)/rocm'
-  pool: ${{ variables.MEDIUM_BUILD_POOL }}
+  pool: ${{ variables.ULTRA_BUILD_POOL }}
   workspace:
     clean: all
   steps:
@@ -51,7 +51,7 @@ jobs:
       extraBuildFlags: >-
         -DCMAKE_PREFIX_PATH="$(Build.BinariesDirectory)/llvm;$(Build.BinariesDirectory)"
         -DCMAKE_BUILD_TYPE=Release
-        -DLLVM_ENABLE_PROJECTS=clang;lld;clang-tools-extra;mlir
+        -DLLVM_ENABLE_PROJECTS=clang;lld;clang-tools-extra;mlir;flang
         -DLLVM_ENABLE_RUNTIMES=compiler-rt;libunwind;libcxx;libcxxabi
         -DCLANG_ENABLE_AMDCLANG=ON
         -DLLVM_TARGETS_TO_BUILD=AMDGPU;X86
@@ -85,7 +85,7 @@ jobs:
       componentName: check-llvm
       testDir: 'llvm/build'
       testExecutable: './bin/llvm-lit'
-      testParameters: '-q --xunit-xml-output=llvm_test_output.xml ./test'
+      testParameters: '-q --xunit-xml-output=llvm_test_output.xml --filter-out="live-debug-values-spill-tracking" ./test'
       testOutputFile: llvm_test_output.xml
   - template: ${{ variables.CI_TEMPLATE_PATH }}/steps/test.yml
     parameters:

.azuredevops/components/omnitrace.yml

@@ -1,140 +0,0 @@
-# largely referenced from: https://github.com/ROCm/omnitrace/blob/main/.github/workflows/ubuntu-jammy.yml
-parameters:
-- name: checkoutRepo
-  type: string
-  default: 'self'
-- name: checkoutRef
-  type: string
-  default: ''
-- name: aptPackages
-  type: object
-  default:
-    - autoconf
-    - autotools-dev
-    - bison
-    - build-essential
-    - bzip2
-    - clang
-    - cmake
-    - environment-modules
-    - g++-12
-    - libdrm-dev
-    - libfabric-dev
-    - libiberty-dev
-    - libpapi-dev
-    - libpfm4-dev
-    - libtool
-    - libopenmpi-dev
-    - m4
-    - openmpi-bin
-    - software-properties-common
-    - python3-pip
-    - texinfo
-    - zlib1g-dev
-- name: pipModules
-  type: object
-  default:
-    - numpy
-    - perfetto
-    - dataclasses
-- name: rocmDependencies
-  type: object
-  default:
-    - aomp
-    - clr
-    - llvm-project
-    - rccl
-    - rocm-core
-    - rocm_smi_lib
-    - rocminfo
-    - ROCR-Runtime
-    - rocprofiler
-    - rocprofiler-register
-    - roctracer
-
-jobs:
-- job: omnitrace
-  variables:
-  - group: common
-  - template: /.azuredevops/variables-global.yml
-  pool: ${{ variables.MEDIUM_BUILD_POOL }}
-  workspace:
-    clean: all
-  strategy:
-    matrix:
-      gfx942:
-        JOB_GPU_TARGET: gfx942
-  steps:
-  - template: ${{ variables.CI_TEMPLATE_PATH }}/steps/dependencies-other.yml
-    parameters:
-      aptPackages: ${{ parameters.aptPackages }}
-      pipModules: ${{ parameters.pipModules }}
-  - template: ${{ variables.CI_TEMPLATE_PATH }}/steps/preamble.yml
-  - template: ${{ variables.CI_TEMPLATE_PATH }}/steps/checkout.yml
-    parameters:
-      checkoutRepo: ${{ parameters.checkoutRepo }}
-  - template: ${{ variables.CI_TEMPLATE_PATH }}/steps/dependencies-rocm.yml
-    parameters:
-      checkoutRef: ${{ parameters.checkoutRef }}
-      dependencyList: ${{ parameters.rocmDependencies }}
-      gpuTarget: $(JOB_GPU_TARGET)
-  - task: Bash@3
-    displayName: ROCm symbolic link
-    inputs:
-      targetType: inline
-      script: |
-        sudo rm -rf /opt/rocm
-        sudo ln -s $(Agent.BuildDirectory)/rocm /opt/rocm
-  - task: Bash@3
-    displayName: Add ROCm binaries to PATH
-    inputs:
-      targetType: inline
-      script: echo "##vso[task.prependpath]$(Agent.BuildDirectory)/rocm/bin"
-  - task: Bash@3
-    displayName: Add ROCm compilers to PATH
-    inputs:
-      targetType: inline
-      script: echo "##vso[task.prependpath]$(Agent.BuildDirectory)/rocm/llvm/bin"
-  - template: ${{ variables.CI_TEMPLATE_PATH }}/steps/build-cmake.yml
-    parameters:
-# build flags reference: https://rocm.docs.amd.com/projects/omnitrace/en/latest/install/install.html
-      extraBuildFlags: >-
-        -DCMAKE_PREFIX_PATH=$(Agent.BuildDirectory)/rocm
-        -DOMNITRACE_BUILD_TESTING=ON
-        -DOMNITRACE_BUILD_DYNINST=ON
-        -DOMNITRACE_BUILD_LIBUNWIND=ON
-        -DDYNINST_BUILD_TBB=ON
-        -DDYNINST_BUILD_ELFUTILS=ON
-        -DDYNINST_BUILD_LIBIBERTY=ON
-        -DDYNINST_BUILD_BOOST=ON
-        -DOMNITRACE_USE_PAPI=ON
-        -DOMNITRACE_USE_MPI=ON
-        -DAMDGPU_TARGETS=$(JOB_GPU_TARGET)
-      multithreadFlag: -- -j32
-  - task: Bash@3
-    displayName: Set up omnitrace env
-    inputs:
-      targetType: inline
-      script: source share/omnitrace/setup-env.sh
-      workingDirectory: build
-  - template: ${{ variables.CI_TEMPLATE_PATH }}/steps/test.yml
-    parameters:
-      componentName: omnitrace
-  - task: Bash@3
-    displayName: Remove ROCm binaries from PATH
-    condition: always()
-    inputs:
-      targetType: inline
-      script: echo "##vso[task.setvariable variable=PATH]$(echo $PATH | sed -e 's;:$(Agent.BuildDirectory)/rocm/bin;;' -e 's;^/;;' -e 's;/$;;')"
-  - task: Bash@3
-    displayName: Remove ROCm compilers from PATH
-    condition: always()
-    inputs:
-      targetType: inline
-      script: echo "##vso[task.setvariable variable=PATH]$(echo $PATH | sed -e 's;:$(Agent.BuildDirectory)/rocm/llvm/bin;;' -e 's;^/;;' -e 's;/$;;')"
-  - template: ${{ variables.CI_TEMPLATE_PATH }}/steps/manifest.yml
-    parameters:
-      gpuTarget: $(JOB_GPU_TARGET)
-  - template: ${{ variables.CI_TEMPLATE_PATH }}/steps/artifact-upload.yml
-    parameters:
-      gpuTarget: $(JOB_GPU_TARGET)

.azuredevops/components/rccl.yml

@@ -64,6 +64,8 @@ jobs:
     matrix:
       gfx942:
         JOB_GPU_TARGET: gfx942
+      gfx90a:
+        JOB_GPU_TARGET: gfx90a
   steps:
   - template: ${{ variables.CI_TEMPLATE_PATH }}/steps/dependencies-other.yml
     parameters:

.azuredevops/components/rdc.yml

@@ -65,6 +65,8 @@ jobs:
     matrix:
       gfx942:
         JOB_GPU_TARGET: gfx942
+      gfx90a:
+        JOB_GPU_TARGET: gfx90a
   steps:
   - template: ${{ variables.CI_TEMPLATE_PATH }}/steps/dependencies-other.yml
     parameters:

.azuredevops/components/rocAL.yml

@@ -73,6 +73,8 @@ jobs:
     matrix:
       gfx942:
         JOB_GPU_TARGET: gfx942
+      gfx90a:
+        JOB_GPU_TARGET: gfx90a
   steps:
   - task: Bash@3
     displayName: 'Register libjpeg-turbo packages'

.azuredevops/components/rocALUTION.yml

@@ -60,6 +60,8 @@ jobs:
     matrix:
       gfx942:
         JOB_GPU_TARGET: gfx942
+      gfx90a:
+        JOB_GPU_TARGET: gfx90a
   steps:
   - template: ${{ variables.CI_TEMPLATE_PATH }}/steps/dependencies-other.yml
     parameters:

.azuredevops/components/rocBLAS.yml

@@ -75,6 +75,8 @@ jobs:
     matrix:
       gfx942:
         JOB_GPU_TARGET: gfx942
+      gfx90a:
+        JOB_GPU_TARGET: gfx90a
   steps:
   - template: ${{ variables.CI_TEMPLATE_PATH }}/steps/dependencies-other.yml
     parameters:

.azuredevops/components/rocFFT.yml

@@ -55,6 +55,8 @@ jobs:
     matrix:
       gfx942:
         JOB_GPU_TARGET: gfx942
+      gfx90a:
+        JOB_GPU_TARGET: gfx90a
   steps:
   - template: ${{ variables.CI_TEMPLATE_PATH }}/steps/dependencies-other.yml
     parameters:

.azuredevops/components/rocJPEG.yml

@@ -47,6 +47,8 @@ jobs:
     matrix:
       gfx942:
         JOB_GPU_TARGET: gfx942
+      gfx90a:
+        JOB_GPU_TARGET: gfx90a
   steps:
   - template: ${{ variables.CI_TEMPLATE_PATH }}/steps/dependencies-other.yml
     parameters:

.azuredevops/components/rocPRIM.yml

@@ -42,6 +42,8 @@ jobs:
     matrix:
       gfx942:
         JOB_GPU_TARGET: gfx942
+      gfx90a:
+        JOB_GPU_TARGET: gfx90a
   steps:
   - template: ${{ variables.CI_TEMPLATE_PATH }}/steps/dependencies-other.yml
     parameters:

.azuredevops/components/rocPyDecode.yml

@@ -48,6 +48,8 @@ jobs:
     matrix:
       gfx942:
         JOB_GPU_TARGET: gfx942
+      gfx90a:
+        JOB_GPU_TARGET: gfx90a
   steps:
   - template: ${{ variables.CI_TEMPLATE_PATH }}/steps/dependencies-other.yml
     parameters:

.azuredevops/components/rocRAND.yml

@@ -45,6 +45,8 @@ jobs:
     matrix:
       gfx942:
         JOB_GPU_TARGET: gfx942
+      gfx90a:
+        JOB_GPU_TARGET: gfx90a
   steps:
   - template: ${{ variables.CI_TEMPLATE_PATH }}/steps/dependencies-other.yml
     parameters:

.azuredevops/components/rocSOLVER.yml

@@ -58,6 +58,8 @@ jobs:
     matrix:
       gfx942:
         JOB_GPU_TARGET: gfx942
+      gfx90a:
+        JOB_GPU_TARGET: gfx90a
   steps:
   - template: ${{ variables.CI_TEMPLATE_PATH }}/steps/dependencies-other.yml
     parameters:

.azuredevops/components/rocSPARSE.yml

@@ -56,6 +56,8 @@ jobs:
     matrix:
       gfx942:
         JOB_GPU_TARGET: gfx942
+      gfx90a:
+        JOB_GPU_TARGET: gfx90a
   steps:
   - template: ${{ variables.CI_TEMPLATE_PATH }}/steps/dependencies-other.yml
     parameters:

.azuredevops/components/rocThrust.yml

@@ -47,6 +47,8 @@ jobs:
     matrix:
       gfx942:
         JOB_GPU_TARGET: gfx942
+      gfx90a:
+        JOB_GPU_TARGET: gfx90a
   steps:
   - template: ${{ variables.CI_TEMPLATE_PATH }}/steps/dependencies-other.yml
     parameters:

.azuredevops/components/rocWMMA.yml

@@ -57,6 +57,8 @@ jobs:
     matrix:
       gfx942:
         JOB_GPU_TARGET: gfx942
+      gfx90a:
+        JOB_GPU_TARGET: gfx90a
   steps:
   - template: ${{ variables.CI_TEMPLATE_PATH }}/steps/dependencies-other.yml
     parameters:

.azuredevops/components/rocm-cmake.yml

@@ -8,7 +8,6 @@ parameters:
 - name: aptPackages
   type: object
   default:
-    - cmake
     - doxygen
     - doxygen-doc
     - ninja-build
@@ -18,7 +17,9 @@ parameters:
   type: object
   default:
     - cget
+    - cmake==3.20.5
     - ninja
+    - rocm-docs-core
 
 jobs:
 - job: rocm_cmake
@@ -33,21 +34,29 @@ jobs:
     parameters:
       aptPackages: ${{ parameters.aptPackages }}
       pipModules: ${{ parameters.pipModules }}
+  - task: Bash@3
+    displayName: Add CMake to PATH
+    inputs:
+      targetType: inline
+      script: echo "##vso[task.prependpath]$(python3 -m site --user-base)/bin"
   - template: ${{ variables.CI_TEMPLATE_PATH }}/steps/preamble.yml
   - template: ${{ variables.CI_TEMPLATE_PATH }}/steps/checkout.yml
     parameters:
       checkoutRepo: ${{ parameters.checkoutRepo }}
   - template: ${{ variables.CI_TEMPLATE_PATH }}/steps/build-cmake.yml
-# extra steps for ctest suite
-  - script: |
-      python -m pip install -r $(Build.SourcesDirectory)/docs/requirements.txt
-      python -m pip install -r $(Build.SourcesDirectory)/test/docsphinx/docs/.sphinx/requirements.txt
-      git config --global user.email "you@example.com"
-      git config --global user.name "Your Name"
-    displayName: "ctest setup"
+  - task: Bash@3
+    displayName: CTest setup
+    inputs:
+      targetType: inline
+      script: |
+        python -m pip install -r $(Build.SourcesDirectory)/docs/requirements.txt
+        python -m pip install -r $(Build.SourcesDirectory)/test/docsphinx/docs/.sphinx/requirements.txt
+        git config --global user.email "you@example.com"
+        git config --global user.name "Your Name"
   - template: ${{ variables.CI_TEMPLATE_PATH }}/steps/test.yml
     parameters:
       componentName: rocm-cmake
+      testParameters: '-E "pass-version-parent" -VV --output-on-failure --force-new-ctest-process --output-junit test_output.xml'
   - template: ${{ variables.CI_TEMPLATE_PATH }}/steps/manifest.yml
   - template: ${{ variables.CI_TEMPLATE_PATH }}/steps/artifact-upload.yml
   - template: ${{ variables.CI_TEMPLATE_PATH }}/steps/artifact-links.yml
@@ -56,4 +65,3 @@ jobs:
       aptPackages: ${{ parameters.aptPackages }}
       pipModules: ${{ parameters.pipModules }}
       environment: combined
-      gpuTarget: $(JOB_GPU_TARGET)

.azuredevops/components/rocm-examples.yml

@@ -75,6 +75,8 @@ jobs:
     matrix:
       gfx942:
         JOB_GPU_TARGET: gfx942
+      gfx90a:
+        JOB_GPU_TARGET: gfx90a
   steps:
   - template: ${{ variables.CI_TEMPLATE_PATH }}/steps/dependencies-other.yml
     parameters:

.azuredevops/components/rocm_smi_lib.yml

@@ -10,6 +10,7 @@ parameters:
   default:
     - cmake
     - libdrm-dev
+    - pkg-config
     - python3-pip
 
 jobs:
@@ -39,7 +40,6 @@ jobs:
   - template: ${{ variables.CI_TEMPLATE_PATH }}/steps/docker-container.yml
     parameters:
       aptPackages: ${{ parameters.aptPackages }}
-      gpuTarget: $(JOB_GPU_TARGET)
 
 - job: rocm_smi_lib_testing
   dependsOn: rocm_smi_lib

.azuredevops/components/rocprofiler-compute.yml

@@ -59,6 +59,8 @@ jobs:
     matrix:
       gfx942:
         JOB_GPU_TARGET: gfx942
+      gfx90a:
+        JOB_GPU_TARGET: gfx90a
   steps:
   - template: ${{ variables.CI_TEMPLATE_PATH }}/steps/dependencies-other.yml
     parameters:

.azuredevops/components/rocprofiler-sdk.yml

@@ -57,6 +57,8 @@ jobs:
     matrix:
       gfx942:
         JOB_GPU_TARGET: gfx942
+      gfx90a:
+        JOB_GPU_TARGET: gfx90a
   steps:
   - template: ${{ variables.CI_TEMPLATE_PATH }}/steps/dependencies-other.yml
     parameters:

.azuredevops/components/rocprofiler-systems.yml

@@ -72,6 +72,8 @@ jobs:
     matrix:
       gfx942:
         JOB_GPU_TARGET: gfx942
+      gfx90a:
+        JOB_GPU_TARGET: gfx90a
   steps:
   - template: ${{ variables.CI_TEMPLATE_PATH }}/steps/dependencies-other.yml
     parameters:

.azuredevops/components/rocprofiler.yml

@@ -57,6 +57,8 @@ jobs:
     matrix:
       gfx942:
         JOB_GPU_TARGET: gfx942
+      gfx90a:
+        JOB_GPU_TARGET: gfx90a
   steps:
   - template: ${{ variables.CI_TEMPLATE_PATH }}/steps/dependencies-other.yml
     parameters:

.azuredevops/components/rocr_debug_agent.yml

@@ -69,7 +69,6 @@ jobs:
   - template: ${{ variables.CI_TEMPLATE_PATH }}/steps/docker-container.yml
     parameters:
       aptPackages: ${{ parameters.aptPackages }}
-      gpuTarget: $(JOB_GPU_TARGET)
 
 - job: rocr_debug_agent_testing
   dependsOn: rocr_debug_agent

.azuredevops/components/roctracer.yml

@@ -11,6 +11,7 @@ parameters:
     - cmake
     - doxygen
     - graphviz
+    - libdrm-amdgpu-dev
     - ninja-build
     - python3-pip
 - name: pipModules
@@ -49,11 +50,14 @@ jobs:
     matrix:
       gfx942:
         JOB_GPU_TARGET: gfx942
+      gfx90a:
+        JOB_GPU_TARGET: gfx90a
   steps:
   - template: ${{ variables.CI_TEMPLATE_PATH }}/steps/dependencies-other.yml
     parameters:
       aptPackages: ${{ parameters.aptPackages }}
       pipModules: ${{ parameters.pipModules }}
+      registerROCmPackages: true
   - template: ${{ variables.CI_TEMPLATE_PATH }}/steps/preamble.yml
   - template: ${{ variables.CI_TEMPLATE_PATH }}/steps/checkout.yml
     parameters:
@@ -85,6 +89,7 @@ jobs:
       aptPackages: ${{ parameters.aptPackages }}
       pipModules: ${{ parameters.pipModules }}
       gpuTarget: $(JOB_GPU_TARGET)
+      registerROCmPackages: true
 
 - job: roctracer_testing
   dependsOn: roctracer
@@ -104,6 +109,8 @@ jobs:
   - template: ${{ variables.CI_TEMPLATE_PATH }}/steps/dependencies-other.yml
     parameters:
       aptPackages: ${{ parameters.aptPackages }}
+      pipModules: ${{ parameters.pipModules }}
+      registerROCmPackages: true
   - template: ${{ variables.CI_TEMPLATE_PATH }}/steps/preamble.yml
   - template: ${{ variables.CI_TEMPLATE_PATH }}/steps/local-artifact-download.yml
     parameters:
@@ -128,3 +135,4 @@ jobs:
       pipModules: ${{ parameters.pipModules }}
       environment: test
       gpuTarget: $(JOB_GPU_TARGET)
+      registerROCmPackages: true

.azuredevops/components/rpp.yml

@@ -57,6 +57,8 @@ jobs:
     matrix:
       gfx942:
         JOB_GPU_TARGET: gfx942
+      gfx90a:
+        JOB_GPU_TARGET: gfx90a
   steps:
   - template: ${{ variables.CI_TEMPLATE_PATH }}/steps/dependencies-other.yml
     parameters:

.azuredevops/templates/steps/dependencies-rocm.yml

@@ -222,13 +222,13 @@ parameters:
       hasGpuTarget: false
     rocm-examples:
       pipelineId: $(ROCM_EXAMPLES_PIPELINE_ID)
-      stagingBranch: develop
-      mainlineBranch: develop
+      stagingBranch: amd-staging
+      mainlineBranch: amd-mainline
       hasGpuTarget: true
     rocminfo:
       pipelineId: $(ROCMINFO_PIPELINE_ID)
       stagingBranch: amd-staging
-      mainlineBranch: amd-master
+      mainlineBranch: amd-mainline
       hasGpuTarget: false
     rocMLIR:
       pipelineId: $(ROCMLIR_PIPELINE_ID)
@@ -262,7 +262,7 @@ parameters:
       hasGpuTarget: true
     rocprofiler-compute:
       pipelineId: $(ROCPROFILER_COMPUTE_PIPELINE_ID)
-      stagingBranch: amd-staging
+      stagingBranch: develop
       mainlineBranch: amd-mainline
       hasGpuTarget: true
     rocprofiler-register:

.azuredevops/templates/steps/test.yml

@@ -33,7 +33,6 @@ parameters:
     - aomp
     - HIPIFY
     - MIVisionX
-    - rocm-cmake
     - rocm_smi_lib
     - rocprofiler-sdk
     - roctracer

.azuredevops/variables-global.yml

@@ -28,13 +28,13 @@ variables:
 - name: GFX942_TEST_POOL
   value: gfx942_test_pool
 - name: LATEST_RELEASE_VERSION
-  value: 6.3.2
+  value: 6.3.3
 - name: REPO_RADEON_VERSION
-  value: 6.3.2
+  value: 6.3.3
 - name: NEXT_RELEASE_VERSION
   value: 6.4.0
 - name: LATEST_RELEASE_TAG
-  value: rocm-6.3.2
+  value: rocm-6.3.3
 - name: AMDMIGRAPHX_GFX942_TEST_PIPELINE_ID
   value: 197
 - name: AMDMIGRAPHX_PIPELINE_ID

.gitignore

@@ -11,6 +11,7 @@ _toc.yml
 docBin/
 _doxygen/
 _readthedocs/
+__pycache__/
 
 # avoid duplicating contributing.md due to conf.py
 docs/CHANGELOG.md

.wordlist.txt

@@ -117,6 +117,7 @@ FX
 Filesystem
 FindDb
 Flang
+FluxBenchmark
 Fortran
 Fuyu
 GALB
@@ -131,6 +132,7 @@ GDS
 GEMM
 GEMMs
 GFortran
+GFXIP
 Gemma
 GiB
 GIM
@@ -316,6 +318,7 @@ PipelineParallel
 PnP
 PowerEdge
 PowerShell
+Pretraining
 Profiler's
 PyPi
 Pytest
@@ -478,6 +481,7 @@ ZenDNN
 accuracies
 activations
 addr
+ai
 alloc
 allocatable
 allocator
@@ -543,6 +547,7 @@ cTDP
 dataset
 datasets
 dataspace
+datatemplate
 datatype
 datatypes
 dbgapi
@@ -571,6 +576,7 @@ el
 embeddings
 enablement
 encodings
+endfor
 endpgm
 enqueue
 env
@@ -691,6 +697,7 @@ pageable
 pallas
 parallelization
 parallelizing
+param
 parameterization
 passthrough
 perfcounter
@@ -715,6 +722,7 @@ preprocessing
 preprocessor
 prequantized
 prerequisites
+pretraining
 profiler
 profilers
 protobuf
@@ -807,6 +815,7 @@ supercomputing
 symlink
 symlinks
 sys
+tabindex
 td
 tensorfloat
 th
@@ -852,6 +861,7 @@ vectorizes
 virtualize
 virtualized
 vjxb
+vllm
 voxel
 walkthrough
 walkthroughs

CONTRIBUTING.md

@@ -66,11 +66,10 @@ project-specific steps. Refer to each repository's PR process for any additional
   during our release cycle, as coordinated by the maintainer
 * We'll inform you once your change is committed
 
-:::{important}
-By creating a PR, you agree to allow your contribution to be licensed under the
-terms of the LICENSE.txt file in the corresponding repository. Different repositories may use different
-licenses.
-:::
+> [!IMPORTANT]
+> By creating a PR, you agree to allow your contribution to be licensed under the
+> terms of the LICENSE.txt file in the corresponding repository. Different repositories may use different
+> licenses.
 
 You can look up each license on the [ROCm licensing](https://rocm.docs.amd.com/en/latest/about/license.html) page.
 

README.md

@@ -116,7 +116,7 @@ bash install-prerequisites.sh
 # For ubuntu22.04 system
 cd ROCm/tools/rocm-build/docker/ubuntu22
 cp * /tmp && cd /tmp
-bash install-prerequisities.sh
+bash install-prerequisites.sh
 # For ubuntu24.04 system
 cd ROCm/tools/rocm-build/docker/ubuntu24
 cp * /tmp && cd /tmp

docs/compatibility/ml-compatibility/jax-compatibility.rst

@@ -4,6 +4,8 @@
    :description: JAX compatibility
    :keywords: GPU, JAX compatibility
 
+.. version-set:: rocm_version latest
+
 *******************************************************************************
 JAX compatibility
 *******************************************************************************
@@ -119,7 +121,8 @@ Critical ROCm libraries for JAX
 
 The functionality of JAX with ROCm is determined by its underlying library
 dependencies. These critical ROCm components affect the capabilities,
-performance, and feature set available to developers.
+performance, and feature set available to developers. The versions described
+are available in ROCm :version:`rocm_version`.
 
 .. list-table::
     :header-rows: 1
@@ -129,7 +132,7 @@ performance, and feature set available to developers.
       - Purpose
       - Used in
     * - `hipBLAS <https://github.com/ROCm/hipBLAS>`_
-      - 2.3.0
+      - :version-ref:`hipBLAS rocm_version`
       - Provides GPU-accelerated Basic Linear Algebra Subprograms (BLAS) for
         matrix and vector operations.
       - Matrix multiplication in ``jax.numpy.matmul``, ``jax.lax.dot`` and
@@ -138,7 +141,7 @@ performance, and feature set available to developers.
         ``jax.numpy.einsum`` with matrix-multiplication patterns algebra
         operations.
     * - `hipBLASLt <https://github.com/ROCm/hipBLASLt>`_
-      - 0.10.0
+      - :version-ref:`hipBLASLt rocm_version`
       - hipBLASLt is an extension of hipBLAS, providing additional
         features like epilogues fused into the matrix multiplication kernel or
         use of integer tensor cores.
@@ -147,7 +150,7 @@ performance, and feature set available to developers.
         operations, mixed-precision support, and hardware-specific
         optimizations.
     * - `hipCUB <https://github.com/ROCm/hipCUB>`_
-      - 3.3.0
+      - :version-ref:`hipCUB rocm_version`
       - Provides a C++ template library for parallel algorithms for reduction,
         scan, sort and select.
       - Reduction functions (``jax.numpy.sum``, ``jax.numpy.mean``,
@@ -155,23 +158,23 @@ performance, and feature set available to developers.
         (``jax.numpy.cumsum``, ``jax.numpy.cumprod``) and sorting
         (``jax.numpy.sort``, ``jax.numpy.argsort``).
     * - `hipFFT <https://github.com/ROCm/hipFFT>`_
-      - 1.0.17
+      - :version-ref:`hipFFT rocm_version`
       - Provides GPU-accelerated Fast Fourier Transform (FFT) operations.
       - Used in functions like ``jax.numpy.fft``.
     * - `hipRAND <https://github.com/ROCm/hipRAND>`_
-      - 2.11.0
+      - :version-ref:`hipRAND rocm_version`
       - Provides fast random number generation for GPUs.
       - The ``jax.random.uniform``, ``jax.random.normal``,
         ``jax.random.randint`` and ``jax.random.split``.
     * - `hipSOLVER <https://github.com/ROCm/hipSOLVER>`_
-      - 2.3.0
+      - :version-ref:`hipSOLVER rocm_version`
       - Provides GPU-accelerated solvers for linear systems, eigenvalues, and
         singular value decompositions (SVD).
       - Solving linear systems (``jax.numpy.linalg.solve``), matrix
         factorizations, SVD (``jax.numpy.linalg.svd``) and eigenvalue problems
         (``jax.numpy.linalg.eig``).
     * - `hipSPARSE <https://github.com/ROCm/hipSPARSE>`_
-      - 3.1.2
+      - :version-ref:`hipSPARSE rocm_version`
       - Accelerates operations on sparse matrices, such as sparse matrix-vector
         or matrix-matrix products.
       - Sparse matrix multiplication (``jax.numpy.matmul``), sparse
@@ -179,28 +182,28 @@ performance, and feature set available to developers.
         (``jax.experimental.sparse.dot``), sparse linear system solvers and
         sparse data handling.
     * - `hipSPARSELt <https://github.com/ROCm/hipSPARSELt>`_
-      - 0.2.2
+      - :version-ref:`hipSPARSELt rocm_version`
       - Accelerates operations on sparse matrices, such as sparse matrix-vector
         or matrix-matrix products.
       - Sparse matrix multiplication (``jax.numpy.matmul``), sparse
         matrix-vector and matrix-matrix products
         (``jax.experimental.sparse.dot``) and sparse linear system solvers.
     * - `MIOpen <https://github.com/ROCm/MIOpen>`_
-      - 3.3.0
+      - :version-ref:`MIOpen rocm_version`
       - Optimized for deep learning primitives such as convolutions, pooling,
         normalization, and activation functions.
       - Speeds up convolutional neural networks (CNNs), recurrent neural
         networks (RNNs), and other layers. Used in operations like
         ``jax.nn.conv``, ``jax.nn.relu``, and ``jax.nn.batch_norm``.
     * - `RCCL <https://github.com/ROCm/rccl>`_
-      - 2.21.5
+      - :version-ref:`RCCL rocm_version`
       - Optimized for multi-GPU communication for operations like  all-reduce,
         broadcast, and scatter.
       - Distribute computations across multiple GPU with ``pmap`` and
         ``jax.distributed``. XLA automatically uses rccl when executing
         operations across multiple GPUs on AMD hardware.
     * - `rocThrust <https://github.com/ROCm/rocThrust>`_
-      - 3.3.0
+      - :version-ref:`rocThrust rocm_version`
       - Provides a C++ template library for parallel algorithms like sorting,
         reduction, and scanning.
       - Reduction operations like ``jax.numpy.sum``, ``jax.pmap`` for

docs/compatibility/ml-compatibility/pytorch-compatibility.rst

@@ -4,6 +4,8 @@
     :description: PyTorch compatibility
     :keywords: GPU, PyTorch compatibility
 
+.. version-set:: rocm_version latest
+
 ********************************************************************************
 PyTorch compatibility
 ********************************************************************************
@@ -56,7 +58,7 @@ Docker image compatibility
 
 AMD validates and publishes ready-made `PyTorch images <https://hub.docker.com/r/rocm/pytorch>`_
 with ROCm backends on Docker Hub. The following Docker image tags and
-associated inventories are validated for `ROCm 6.3.0 <https://repo.radeon.com/rocm/apt/6.3/>`_.
+associated inventories are validated for `ROCm 6.3.3 <https://repo.radeon.com/rocm/apt/6.3.3/>`_.
 Click the |docker-icon| icon to view the image on Docker Hub.
 
 .. list-table:: PyTorch Docker image components
@@ -77,26 +79,26 @@ Click the |docker-icon| icon to view the image on Docker Hub.
 
     * - .. raw:: html
 
-           <a href="https://hub.docker.com/layers/rocm/pytorch/rocm6.3_ubuntu24.04_py3.12_pytorch_release_2.4.0/images/sha256-98ddf20333bd01ff749b8092b1190ee369a75d3b8c71c2fac80ffdcb1a98d529?context=explore"><i class="fab fa-docker fa-lg"></i></a>
+           <a href="https://hub.docker.com/layers/rocm/pytorch/rocm6.3.3_ubuntu24.04_py3.12_pytorch_release_2.4.0/images/sha256-6c798857b2c9526b44ba535710b93a1737546acea79b53a93c646195c272f1d5"><i class="fab fa-docker fa-lg"></i></a>
 
       - `2.4.0 <https://github.com/ROCm/pytorch/tree/release/2.4>`_
       - 24.04
-      - `3.12 <https://www.python.org/downloads/release/python-3128/>`_
+      - `3.12.9 <https://www.python.org/downloads/release/python-3129/>`_
       - `1.4.0 <https://github.com/ROCm/apex/tree/release/1.4.0>`_
       - `0.19.0 <https://github.com/pytorch/vision/tree/v0.19.0>`_
       - `2.13.0 <https://github.com/tensorflow/tensorboard/tree/2.13.0>`_
       - `master <https://bitbucket.org/icl/magma/src/master/>`_
       - `1.10.0 <https://github.com/openucx/ucx/tree/v1.10.0>`_
-      - `4.0.7 <https://github.com/open-mpi/ompi/tree/v4.0.7>`_
+      - `4.0.3 <https://github.com/open-mpi/ompi/tree/v4.0.3>`_
       - `5.3-1.0.5.0 <https://content.mellanox.com/ofed/MLNX_OFED-5.3-1.0.5.0/MLNX_OFED_LINUX-5.3-1.0.5.0-ubuntu20.04-x86_64.tgz>`_
 
     * - .. raw:: html
 
-           <a href="https://hub.docker.com/layers/rocm/pytorch/rocm6.3_ubuntu22.04_py3.10_pytorch_release_2.4.0/images/sha256-402c9b4f1a6b5a81c634a1932b56cbe01abb699cfcc7463d226276997c6cf8ea?context=explore"><i class="fab fa-docker fa-lg"></i></a>
+           <a href="https://hub.docker.com/layers/rocm/pytorch/rocm6.3.3_ubuntu22.04_py3.10_pytorch_release_2.4.0/images/sha256-a09b21248133876fc8912a5ff4e6ee2c8d62b14120313e426b3dadda5702713d"><i class="fab fa-docker fa-lg"></i></a>
 
       - `2.4.0 <https://github.com/ROCm/pytorch/tree/release/2.4>`_
       - 22.04
-      - `3.10 <https://www.python.org/downloads/release/python-31016/>`_
+      - `3.10.16 <https://www.python.org/downloads/release/python-31016/>`_
       - `1.4.0 <https://github.com/ROCm/apex/tree/release/1.4.0>`_
       - `0.19.0 <https://github.com/pytorch/vision/tree/v0.19.0>`_
       - `2.13.0 <https://github.com/tensorflow/tensorboard/tree/2.13.0>`_
@@ -107,11 +109,11 @@ Click the |docker-icon| icon to view the image on Docker Hub.
 
     * - .. raw:: html
 
-           <a href="https://hub.docker.com/layers/rocm/pytorch/rocm6.3_ubuntu22.04_py3.9_pytorch_release_2.4.0/images/sha256-e0608b55d408c3bfe5c19fdd57a4ced3e0eb3a495b74c309980b60b156c526dd?context=explore"><i class="fab fa-docker fa-lg"></i></a>
+           <a href="https://hub.docker.com/layers/rocm/pytorch/rocm6.3.3_ubuntu22.04_py3.9_pytorch_release_2.4.0/images/sha256-963187534467f0f9da77996762fc1d112a6faa5372277c348a505533e7876ec8"><i class="fab fa-docker fa-lg"></i></a>
 
       - `2.4.0 <https://github.com/ROCm/pytorch/tree/release/2.4>`_
       - 22.04
-      - `3.9.18 <https://www.python.org/downloads/release/python-3918/>`_
+      - `3.9.21 <https://www.python.org/downloads/release/python-3921/>`_
       - `1.4.0 <https://github.com/ROCm/apex/tree/release/1.4.0>`_
       - `0.19.0 <https://github.com/pytorch/vision/tree/v0.19.0>`_
       - `2.13.0 <https://github.com/tensorflow/tensorboard/tree/2.13.0>`_
@@ -122,11 +124,11 @@ Click the |docker-icon| icon to view the image on Docker Hub.
 
     * - .. raw:: html
 
-           <a href="https://hub.docker.com/layers/rocm/pytorch/rocm6.3_ubuntu22.04_py3.10_pytorch_release_2.3.0/images/sha256-652cf25263d05b1de548222970aeb76e60b12de101de66751264709c0d0ff9d8?context=explore"><i class="fab fa-docker fa-lg"></i></a>
+           <a href="https://hub.docker.com/layers/rocm/pytorch/rocm6.3.3_ubuntu22.04_py3.10_pytorch_release_2.3.0/images/sha256-952f2621bd2bf3078bef19061e05b209105a82a7908e7e6cdf85014938a4d93a"><i class="fab fa-docker fa-lg"></i></a>
 
       - `2.3.0 <https://github.com/ROCm/pytorch/tree/release/2.3>`_
       - 22.04
-      - `3.10 <https://www.python.org/downloads/release/python-31016/>`_
+      - `3.10.16 <https://www.python.org/downloads/release/python-31016/>`_
       - `1.3.0 <https://github.com/ROCm/apex/tree/release/1.3.0>`_
       - `0.18.0 <https://github.com/pytorch/vision/tree/v0.18.0>`_
       - `2.13.0 <https://github.com/tensorflow/tensorboard/tree/2.13.0>`_
@@ -137,7 +139,7 @@ Click the |docker-icon| icon to view the image on Docker Hub.
 
     * - .. raw:: html
 
-           <a href="https://hub.docker.com/layers/rocm/pytorch/rocm6.3_ubuntu22.04_py3.10_pytorch_release_2.2.1/images/sha256-051976f26beab8f9aa65d999e3ad546c027b39240a0cc3ee81b114a9024f2912?context=explore"><i class="fab fa-docker fa-lg"></i></a>
+           <a href="https://hub.docker.com/layers/rocm/pytorch/rocm6.3.3_ubuntu22.04_py3.10_pytorch_release_2.2.1/images/sha256-a2fe20e170feb9e05da3e5728bb98e40d08567e137be8e6ba797962ed2852608"><i class="fab fa-docker fa-lg"></i></a>
 
       - `2.2.1 <https://github.com/ROCm/pytorch/tree/release/2.2>`_
       - 22.04
@@ -152,7 +154,7 @@ Click the |docker-icon| icon to view the image on Docker Hub.
 
     * - .. raw:: html
 
-           <a href="https://hub.docker.com/layers/rocm/pytorch/rocm6.3_ubuntu20.04_py3.9_pytorch_release_2.2.1/images/sha256-88c839a364d109d3748c100385bfa100d28090d25118cc723fd0406390ab2f7e?context=explore"><i class="fab fa-docker fa-lg"></i></a>
+           <a href="https://hub.docker.com/layers/rocm/pytorch/rocm6.3.3_ubuntu20.04_py3.9_pytorch_release_2.2.1/images/sha256-7f231937c897cca5f89e360be33c70a2017d60f62d1fbe81292be48c15fe345b"><i class="fab fa-docker fa-lg"></i></a>
 
       - `2.2.1 <https://github.com/ROCm/pytorch/tree/release/2.2>`_
       - 20.04
@@ -167,14 +169,14 @@ Click the |docker-icon| icon to view the image on Docker Hub.
 
     * - .. raw:: html
 
-           <a href="https://hub.docker.com/layers/rocm/pytorch/rocm6.3_ubuntu22.04_py3.9_pytorch_release_1.13.1/images/sha256-994424ed07a63113f79dd9aa72159124c00f5fbfe18127151e6658f7d0b6f821?context=explore"><i class="fab fa-docker fa-lg"></i></a>
+           <a href="https://hub.docker.com/layers/rocm/pytorch/rocm6.3.3_ubuntu22.04_py3.9_pytorch_release_1.13.1/images/sha256-616a47758004f91951e2da6c1fe291f903de65a7b2318d4b18359b48fe3032f4"><i class="fab fa-docker fa-lg"></i></a>
 
       - `1.13.1 <https://github.com/ROCm/pytorch/tree/release/1.13>`_
       - 22.04
       - `3.9.21 <https://www.python.org/downloads/release/python-3921/>`_
       - `1.0.0 <https://github.com/ROCm/apex/tree/release/1.0.0>`_
       - `0.14.0 <https://github.com/pytorch/vision/tree/v0.14.0>`_
-      - `2.18.0 <https://github.com/tensorflow/tensorboard/tree/2.18>`_
+      - `2.19.0 <https://github.com/tensorflow/tensorboard/tree/2.19>`_
       - `master <https://bitbucket.org/icl/magma/src/master/>`_
       - `1.14.1 <https://github.com/openucx/ucx/tree/v1.14.1>`_
       - `4.1.5 <https://github.com/open-mpi/ompi/tree/v4.1.5>`_
@@ -182,7 +184,7 @@ Click the |docker-icon| icon to view the image on Docker Hub.
 
     * - .. raw:: html
 
-           <a href="https://hub.docker.com/layers/rocm/pytorch/rocm6.3_ubuntu20.04_py3.9_pytorch_release_1.13.1/images/sha256-7b8139fe40a9aeb4bca3aecd15c22c1fa96e867d93479fa3a24fdeeeeafa1219?context=explore"><i class="fab fa-docker fa-lg"></i></a>
+           <a href="https://hub.docker.com/layers/rocm/pytorch/rocm6.3.3_ubuntu20.04_py3.9_pytorch_release_1.13.1/images/sha256-a2cfb365aea58b84595e241ffdb0d5ef3e6566e98c10b5499f4aa29983a74ea2"><i class="fab fa-docker fa-lg"></i></a>
 
       - `1.13.1 <https://github.com/ROCm/pytorch/tree/release/1.13>`_
       - 20.04
@@ -200,7 +202,8 @@ Critical ROCm libraries for PyTorch
 
 The functionality of PyTorch with ROCm is determined by its underlying library
 dependencies. These critical ROCm components affect the capabilities,
-performance, and feature set available to developers.
+performance, and feature set available to developers. The versions described
+are available in ROCm :version:`rocm_version`.
 
 .. list-table::
     :header-rows: 1
@@ -210,28 +213,28 @@ performance, and feature set available to developers.
       - Purpose
       - Used in
     * - `Composable Kernel <https://github.com/ROCm/composable_kernel>`_
-      - 1.1.0
+      - :version-ref:`"Composable Kernel" rocm_version`
       - Enables faster execution of core operations like matrix multiplication
         (GEMM), convolutions and transformations.
       - Speeds up ``torch.permute``, ``torch.view``, ``torch.matmul``,
         ``torch.mm``, ``torch.bmm``, ``torch.nn.Conv2d``, ``torch.nn.Conv3d``
         and ``torch.nn.MultiheadAttention``.
     * - `hipBLAS <https://github.com/ROCm/hipBLAS>`_
-      - 2.3.0
+      - :version-ref:`hipBLAS rocm_version`
       - Provides GPU-accelerated Basic Linear Algebra Subprograms (BLAS) for
         matrix and vector operations.
       - Supports operations like matrix multiplication, matrix-vector products,
         and tensor contractions. Utilized in both dense and batched linear
         algebra operations.
     * - `hipBLASLt <https://github.com/ROCm/hipBLASLt>`_
-      - 0.10.0
+      - :version-ref:`hipBLASLt rocm_version`
       - hipBLASLt is an extension of the hipBLAS library, providing additional
         features like epilogues fused into the matrix multiplication kernel or
         use of integer tensor cores.
       - It accelerates operations like ``torch.matmul``, ``torch.mm``, and the
         matrix multiplications used in convolutional and linear layers.
     * - `hipCUB <https://github.com/ROCm/hipCUB>`_
-      - 3.3.0
+      - :version-ref:`hipCUB rocm_version`
       - Provides a C++ template library for parallel algorithms for reduction,
         scan, sort and select.
       - Supports operations like ``torch.sum``, ``torch.cumsum``, ``torch.sort``
@@ -239,93 +242,93 @@ performance, and feature set available to developers.
         irregular shapes often involve scanning, sorting, and filtering, which
         hipCUB handles efficiently.
     * - `hipFFT <https://github.com/ROCm/hipFFT>`_
-      - 1.0.17
+      - :version-ref:`hipFFT rocm_version`
       - Provides GPU-accelerated Fast Fourier Transform (FFT) operations.
       - Used in functions like the ``torch.fft`` module.
     * - `hipRAND <https://github.com/ROCm/hipRAND>`_
-      - 2.11.0
+      - :version-ref:`hipRAND rocm_version`
       - Provides fast random number generation for GPUs.
       - The ``torch.rand``, ``torch.randn`` and stochastic layers like
         ``torch.nn.Dropout``.
     * - `hipSOLVER <https://github.com/ROCm/hipSOLVER>`_
-      - 2.3.0
+      - :version-ref:`hipSOLVER rocm_version`
       - Provides GPU-accelerated solvers for linear systems, eigenvalues, and
         singular value decompositions (SVD).
       - Supports functions like ``torch.linalg.solve``,
         ``torch.linalg.eig``, and ``torch.linalg.svd``.
     * - `hipSPARSE <https://github.com/ROCm/hipSPARSE>`_
-      - 3.1.2
+      - :version-ref:`hipSPARSE rocm_version`
       - Accelerates operations on sparse matrices, such as sparse matrix-vector
         or matrix-matrix products.
       - Sparse tensor operations ``torch.sparse``.
     * - `hipSPARSELt <https://github.com/ROCm/hipSPARSELt>`_
-      - 0.2.2
+      - :version-ref:`hipSPARSELt rocm_version`
       - Accelerates operations on sparse matrices, such as sparse matrix-vector
         or matrix-matrix products.
       - Sparse tensor operations ``torch.sparse``.
     * - `hipTensor <https://github.com/ROCm/hipTensor>`_
-      - 1.4.0
+      - :version-ref:`hipTensor rocm_version`
       - Optimizes for high-performance tensor operations, such as contractions.
       - Accelerates tensor algebra, especially in deep learning and scientific
         computing.
     * - `MIOpen <https://github.com/ROCm/MIOpen>`_
-      - 3.3.0
+      - :version-ref:`MIOpen rocm_version`
       - Optimizes deep learning primitives such as convolutions, pooling,
         normalization, and activation functions.
       - Speeds up convolutional neural networks (CNNs), recurrent neural
         networks (RNNs), and other layers. Used in operations like
         ``torch.nn.Conv2d``, ``torch.nn.ReLU``, and ``torch.nn.LSTM``.
     * - `MIGraphX <https://github.com/ROCm/AMDMIGraphX>`_
-      - 2.11.0
+      - :version-ref:`MIGraphX rocm_version`
       - Adds graph-level optimizations, ONNX models and mixed precision support
         and enable Ahead-of-Time (AOT) Compilation.
       - Speeds up inference models and executes ONNX models for
         compatibility with other frameworks.
         ``torch.nn.Conv2d``, ``torch.nn.ReLU``, and ``torch.nn.LSTM``.
     * - `MIVisionX <https://github.com/ROCm/MIVisionX>`_
-      - 3.1.0
+      - :version-ref:`MIVisionX rocm_version`
       - Optimizes acceleration for computer vision and AI workloads like
         preprocessing, augmentation, and inferencing.
       - Faster data preprocessing and augmentation pipelines for datasets like
         ImageNet or COCO and easy to integrate into PyTorch's ``torch.utils.data``
         and ``torchvision`` workflows.
     * - `rocAL <https://github.com/ROCm/rocAL>`_
-      - 2.1.0
+      - :version-ref:`rocAL rocm_version`
       - Accelerates the data pipeline by offloading intensive preprocessing and
         augmentation tasks. rocAL is part of MIVisionX.
       - Easy to integrate into PyTorch's ``torch.utils.data`` and
         ``torchvision`` data load workloads.
     * - `RCCL <https://github.com/ROCm/rccl>`_
-      - 2.21.5
+      - :version-ref:`RCCL rocm_version`
       - Optimizes for multi-GPU communication for operations like AllReduce and
         Broadcast.
       - Distributed data parallel training (``torch.nn.parallel.DistributedDataParallel``).
         Handles communication in multi-GPU setups.
     * - `rocDecode <https://github.com/ROCm/rocDecode>`_
-      - 0.8.0
+      - :version-ref:`rocDecode rocm_version`
       - Provides hardware-accelerated data decoding capabilities, particularly
         for image, video, and other dataset formats.
       - Can be integrated in ``torch.utils.data``, ``torchvision.transforms``
         and ``torch.distributed``.
     * - `rocJPEG <https://github.com/ROCm/rocJPEG>`_
-      - 0.6.0
+      - :version-ref:`rocJPEG rocm_version`
       - Provides hardware-accelerated JPEG image decoding and encoding.
       - GPU accelerated ``torchvision.io.decode_jpeg`` and
         ``torchvision.io.encode_jpeg`` and can be integrated in
         ``torch.utils.data`` and ``torchvision``.
     * - `RPP <https://github.com/ROCm/RPP>`_
-      - 1.9.1
+      - :version-ref:`RPP rocm_version`
       - Speeds up data augmentation, transformation, and other preprocessing steps.
       - Easy to integrate into PyTorch's ``torch.utils.data`` and
         ``torchvision`` data load workloads.
     * - `rocThrust <https://github.com/ROCm/rocThrust>`_
-      - 3.3.0
+      - :version-ref:`rocThrust rocm_version`
       - Provides a C++ template library for parallel algorithms like sorting,
         reduction, and scanning.
       - Utilized in backend operations for tensor computations requiring
         parallel processing.
     * - `rocWMMA <https://github.com/ROCm/rocWMMA>`_
-      - 1.6.0
+      - :version-ref:`rocWMMA rocm_version`
       - Accelerates warp-level matrix-multiply and matrix-accumulate to speed up matrix
         multiplication (GEMM) and accumulation operations with mixed precision
         support.

docs/compatibility/ml-compatibility/tensorflow-compatibility.rst

@@ -4,6 +4,8 @@
     :description: TensorFlow compatibility
     :keywords: GPU, TensorFlow compatibility
 
+.. version-set:: rocm_version latest
+
 *******************************************************************************
 TensorFlow compatibility
 *******************************************************************************
@@ -54,7 +56,7 @@ Docker image compatibility
 AMD validates and publishes ready-made `TensorFlow images
 <https://hub.docker.com/r/rocm/tensorflow>`_ with ROCm backends on
 Docker Hub. The following Docker image tags and associated inventories are
-validated for `ROCm 6.3.1 <https://repo.radeon.com/rocm/apt/6.3.1/>`_. Click
+validated for `ROCm 6.3.3 <https://repo.radeon.com/rocm/apt/6.3.3/>`_. Click
 the |docker-icon| icon to view the image on Docker Hub.
 
 .. list-table:: TensorFlow Docker image components
@@ -68,47 +70,47 @@ the |docker-icon| icon to view the image on Docker Hub.
 
     * - .. raw:: html
 
-           <a href="https://hub.docker.com/layers/rocm/tensorflow/rocm6.3.1-py3.12-tf2.17.0-dev/images/sha256-804121ee4985718277ba7dcec53c57bdade130a1ef42f544b6c48090ad379c17"><i class="fab fa-docker fa-lg"></i> rocm/tensorflow</a>
+           <a href="https://hub.docker.com/layers/rocm/tensorflow/rocm6.3.3-py3.12-tf2.17-dev/images/sha256-fd2653f436880366cc874aa24264ca9dabd892d76ccb63fb807debba459bcaaf"><i class="fab fa-docker fa-lg"></i> rocm/tensorflow</a>
 
       - `tensorflow-rocm 2.17.0 <https://repo.radeon.com/rocm/manylinux/rocm-rel-6.3/tensorflow_rocm-2.17.0-cp312-cp312-manylinux_2_28_x86_64.whl>`__
       - dev
-      - `Python 3.12 <https://www.python.org/downloads/release/python-3124/>`_
+      - `Python 3.12.4 <https://www.python.org/downloads/release/python-3124/>`_
       - `TensorBoard 2.17.1 <https://github.com/tensorflow/tensorboard/tree/2.17.1>`_
 
     * - .. raw:: html
 
-           <a href="https://hub.docker.com/layers/rocm/tensorflow/rocm6.3.1-py3.10-tf2.17.0-dev/images/sha256-776837ffa945913f6c466bfe477810a11453d21d5b6afb200be1c36e48fbc08e"><i class="fab fa-docker fa-lg"></i> rocm/tensorflow</a>
+           <a href="https://hub.docker.com/layers/rocm/tensorflow/rocm6.3.3-py3.10-tf2.17-dev/images/sha256-8a5eb7443798935dd269575e2abae847b702e1dfb06766ab84f081a6314d8b95"><i class="fab fa-docker fa-lg"></i> rocm/tensorflow</a>
 
       - `tensorflow-rocm 2.17.0 <https://repo.radeon.com/rocm/manylinux/rocm-rel-6.3/tensorflow_rocm-2.17.0-cp310-cp310-manylinux_2_28_x86_64.whl>`__
       - dev
-      - `Python 3.10 <https://www.python.org/downloads/release/python-31012/>`_
-      - `TensorBoard 2.17.0 <https://github.com/tensorflow/tensorboard/tree/2.17.0>`_
+      - `Python 3.10.16 <https://www.python.org/downloads/release/python-31016/>`_
+      - `TensorBoard 2.17.1 <https://github.com/tensorflow/tensorboard/tree/2.17.1>`_
 
     * - .. raw:: html
 
-           <a href="https://hub.docker.com/layers/rocm/tensorflow/rocm6.3.1-py3.12-tf2.16.2-dev/images/sha256-c793e1483e30809c3c28fc5d7805bedc033c73da224f839fff370717cb100944"><i class="fab fa-docker fa-lg"></i> rocm/tensorflow</a>
+           <a href="https://hub.docker.com/layers/rocm/tensorflow/rocm6.3.3-py3.12-tf2.16-dev/images/sha256-8fc939b10cdd6d2b11407474880d4c8ab2b52ab6e2d1743c921fc2adbfd0422f"><i class="fab fa-docker fa-lg"></i> rocm/tensorflow</a>
 
       - `tensorflow-rocm 2.16.2 <https://repo.radeon.com/rocm/manylinux/rocm-rel-6.3/tensorflow_rocm-2.16.2-cp312-cp312-manylinux_2_28_x86_64.whl>`__
       - dev
-      - `Python 3.12 <https://www.python.org/downloads/release/python-3124/>`_
+      - `Python 3.12.4 <https://www.python.org/downloads/release/python-3124/>`_
       - `TensorBoard 2.16.2 <https://github.com/tensorflow/tensorboard/tree/2.16.2>`_
 
     * - .. raw:: html
 
-           <a href="https://hub.docker.com/layers/rocm/tensorflow/rocm6.3.1-py3.10-tf2.16.0-dev/images/sha256-263e78414ae85d7bcd52a025a94131d0a279872a45ed632b9165336dfdcd4443"><i class="fab fa-docker fa-lg"></i> rocm/tensorflow</a>
+           <a href="https://hub.docker.com/layers/rocm/tensorflow/rocm6.3.3-py3.10-tf2.16-dev/images/sha256-a4cc6ab23d59fdf5459ceac1f0a603e6c16ae7f885d30e42c0c2b3ac60c2ad10"><i class="fab fa-docker fa-lg"></i> rocm/tensorflow</a>
 
       - `tensorflow-rocm 2.16.2 <https://repo.radeon.com/rocm/manylinux/rocm-rel-6.3/tensorflow_rocm-2.16.2-cp310-cp310-manylinux_2_28_x86_64.whl>`__
       - dev
-      - `Python 3.10 <https://www.python.org/downloads/release/python-31012/>`_
+      - `Python 3.10.16 <https://www.python.org/downloads/release/python-31016/>`_
       - `TensorBoard 2.16.2 <https://github.com/tensorflow/tensorboard/tree/2.16.2>`_
 
     * - .. raw:: html
 
-           <a href="https://hub.docker.com/layers/rocm/tensorflow/rocm6.3.1-py3.10-tf2.15.0-dev/images/sha256-479046a8477ca701a9494a813ab17e8ab4f6baa54641e65dc8d07629f1e6a880"><i class="fab fa-docker fa-lg"></i> rocm/tensorflow</a>
+           <a href="https://hub.docker.com/layers/rocm/tensorflow/rocm6.3.3-py3.10-tf2.15-dev/images/sha256-60887c488421184adcb60b9ed4f72a8bd7bdb64d238e50943ca7cbde38e4aa48"><i class="fab fa-docker fa-lg"></i> rocm/tensorflow</a>
 
       - `tensorflow-rocm 2.15.1 <https://repo.radeon.com/rocm/manylinux/rocm-rel-6.3/tensorflow_rocm-2.15.1-cp310-cp310-manylinux_2_28_x86_64.whl>`_
       - dev
-      - `Python 3.10 <https://www.python.org/downloads/release/python-31012/>`_
+      - `Python 3.10.16 <https://www.python.org/downloads/release/python-31016/>`_
       - `TensorBoard 2.15.2 <https://github.com/tensorflow/tensorboard/tree/2.15.2>`_
 
 Critical ROCm libraries for TensorFlow
@@ -117,7 +119,8 @@ Critical ROCm libraries for TensorFlow
 TensorFlow depends on multiple components and the supported features of those
 components can affect the TensorFlow ROCm supported feature set. The versions
 in the following table refer to the first TensorFlow version where the ROCm
-library was introduced as a dependency.
+library was introduced as a dependency. The versions described
+are available in ROCm :version:`rocm_version`.
 
 .. list-table::
     :widths: 25, 10, 35, 30
@@ -128,43 +131,43 @@ library was introduced as a dependency.
       - Purpose
       - Used in
     * - `hipBLAS <https://github.com/ROCm/hipBLAS>`_
-      - 2.3.0
+      - :version-ref:`hipBLAS rocm_version`
       - Provides GPU-accelerated Basic Linear Algebra Subprograms (BLAS) for
         matrix and vector operations.
       - Accelerates operations like ``tf.matmul``, ``tf.linalg.matmul``, and
         other matrix multiplications commonly used in neural network layers.
     * - `hipBLASLt <https://github.com/ROCm/hipBLASLt>`_
-      - 0.10.0
+      - :version-ref:`hipBLASLt rocm_version`
       - Extends hipBLAS with additional optimizations like fused kernels and
         integer tensor cores.
       - Optimizes matrix multiplications and linear algebra operations used in
         layers like dense, convolutional, and RNNs in TensorFlow.
     * - `hipCUB <https://github.com/ROCm/hipCUB>`_
-      - 3.3.0
+      - :version-ref:`hipCUB rocm_version`
       - Provides a C++ template library for parallel algorithms for reduction,
         scan, sort and select.
       - Supports operations like ``tf.reduce_sum``, ``tf.cumsum``, ``tf.sort``
         and other tensor operations in TensorFlow, especially those involving
         scanning, sorting, and filtering.
     * - `hipFFT <https://github.com/ROCm/hipFFT>`_
-      - 1.0.17
+      - :version-ref:`hipFFT rocm_version`
       - Accelerates Fast Fourier Transforms (FFT) for signal processing tasks.
       - Used for operations like signal processing, image filtering, and
         certain types of neural networks requiring FFT-based transformations.
     * - `hipSOLVER <https://github.com/ROCm/hipSOLVER>`_
-      - 2.3.0
+      - :version-ref:`hipSOLVER rocm_version`
       - Provides GPU-accelerated direct linear solvers for dense and sparse
         systems.
       - Optimizes linear algebra functions such as solving systems of linear
         equations, often used in optimization and training tasks.
     * - `hipSPARSE <https://github.com/ROCm/hipSPARSE>`_
-      - 3.1.2
+      - :version-ref:`hipSPARSE rocm_version`
       - Optimizes sparse matrix operations for efficient computations on sparse
         data.
       - Accelerates sparse matrix operations in models with sparse weight
         matrices or activations, commonly used in neural networks.
     * - `MIOpen <https://github.com/ROCm/MIOpen>`_
-      - 3.3.0
+      - :version-ref:`MIOpen rocm_version`
       - Provides optimized deep learning primitives such as convolutions,
         pooling,
         normalization, and activation functions.
@@ -172,13 +175,13 @@ library was introduced as a dependency.
         in TensorFlow for layers like ``tf.nn.conv2d``, ``tf.nn.relu``, and
         ``tf.nn.lstm_cell``.
     * - `RCCL <https://github.com/ROCm/rccl>`_
-      - 2.21.5
+      - :version-ref:`RCCL rocm_version`
       - Optimizes for multi-GPU communication for operations like AllReduce and
         Broadcast.
       - Distributed data parallel training (``tf.distribute.MirroredStrategy``).
         Handles communication in multi-GPU setups.
     * - `rocThrust <https://github.com/ROCm/rocThrust>`_
-      - 3.3.0
+      - :version-ref:`rocThrust rocm_version`
       - Provides a C++ template library for parallel algorithms like sorting,
         reduction, and scanning.
       - Reduction operations like ``tf.reduce_sum``, ``tf.cumsum`` for computing

docs/conceptual/gpu-arch.md

@@ -32,7 +32,7 @@ architecture.
 
 * [AMD Instinct™ MI250 microarchitecture](./gpu-arch/mi250.md)
 * [AMD Instinct MI200/CDNA2 ISA](https://www.amd.com/system/files/TechDocs/instinct-mi200-cdna2-instruction-set-architecture.pdf)
-* [White paper](https://www.amd.com/system/files/documents/amd-cdna2-white-paper.pdf)
+* [White paper](https://www.amd.com/content/dam/amd/en/documents/instinct-business-docs/white-papers/amd-cdna2-white-paper.pdf)
 * [Performance counters](./gpu-arch/mi300-mi200-performance-counters.rst)
 
 :::
@@ -45,7 +45,7 @@ architecture.
 
 * [AMD Instinct™ MI100 microarchitecture](./gpu-arch/mi100.md)
 * [AMD Instinct MI100/CDNA1 ISA](https://www.amd.com/system/files/TechDocs/instinct-mi100-cdna1-shader-instruction-set-architecture%C2%A0.pdf)
-* [White paper](https://www.amd.com/system/files/documents/amd-cdna-whitepaper.pdf)
+* [White paper](https://www.amd.com/content/dam/amd/en/documents/instinct-business-docs/white-papers/amd-cdna-white-paper.pdf)
 
 :::
 
@@ -55,7 +55,6 @@ architecture.
 * [AMD RDNA3 ISA](https://www.amd.com/system/files/TechDocs/rdna3-shader-instruction-set-architecture-feb-2023_0.pdf)
 * [AMD RDNA2 ISA](https://www.amd.com/system/files/TechDocs/rdna2-shader-instruction-set-architecture.pdf)
 * [AMD RDNA ISA](https://www.amd.com/system/files/TechDocs/rdna-shader-instruction-set-architecture.pdf)
-* [AMD RDNA Architecture White Paper](https://www.amd.com/system/files/documents/rdna-whitepaper.pdf)
 
 :::
 

docs/conf.py

@@ -6,6 +6,8 @@
 
 import os
 import shutil
+import sys
+from pathlib import Path
 
 shutil.copy2("../RELEASE.md", "./about/release-notes.md")
 
@@ -49,6 +51,9 @@ article_pages = [
 
     {"file": "how-to/rocm-for-ai/training/index", "os": ["linux"]},
     {"file": "how-to/rocm-for-ai/training/train-a-model", "os": ["linux"]},
+    {"file": "how-to/rocm-for-ai/training/prerequisite-system-validation", "os": ["linux"]},
+    {"file": "how-to/rocm-for-ai/training/benchmark-docker/megatron-lm", "os": ["linux"]},
+    {"file": "how-to/rocm-for-ai/training/benchmark-docker/pytorch-training", "os": ["linux"]},
     {"file": "how-to/rocm-for-ai/training/scale-model-training", "os": ["linux"]},
 
     {"file": "how-to/rocm-for-ai/fine-tuning/index", "os": ["linux"]},
@@ -63,7 +68,7 @@ article_pages = [
     {"file": "how-to/rocm-for-ai/inference/llm-inference-frameworks", "os": ["linux"]},
     {"file": "how-to/rocm-for-ai/inference/vllm-benchmark", "os": ["linux"]},
     {"file": "how-to/rocm-for-ai/inference/deploy-your-model", "os": ["linux"]},
-    
+
     {"file": "how-to/rocm-for-ai/inference-optimization/index", "os": ["linux"]},
     {"file": "how-to/rocm-for-ai/inference-optimization/model-quantization", "os": ["linux"]},
     {"file": "how-to/rocm-for-ai/inference-optimization/model-acceleration-libraries", "os": ["linux"]},
@@ -86,11 +91,16 @@ article_pages = [
 
 external_toc_path = "./sphinx/_toc.yml"
 
-extensions = ["rocm_docs", "sphinx_reredirects", "sphinx_sitemap"]
+# Add the _extensions directory to Python's search path
+sys.path.append(str(Path(__file__).parent / 'extension'))
+
+extensions = ["rocm_docs", "sphinx_reredirects", "sphinx_sitemap", "sphinxcontrib.datatemplates", "version-ref"]
+
+compatibility_matrix_file = str(Path(__file__).parent / 'compatibility/compatibility-matrix-historical-6.0.csv')
 
 external_projects_current_project = "rocm"
 
-# Uncomment if facing rate limit exceed issue with local build 
+# Uncomment if facing rate limit exceed issue with local build
 # external_projects_remote_repository = ""
 
 html_baseurl = os.environ.get("READTHEDOCS_CANONICAL_URL", "https://rocm-stg.amd.com/")
@@ -101,8 +111,9 @@ if os.environ.get("READTHEDOCS", "") == "True":
 html_theme = "rocm_docs_theme"
 html_theme_options = {"flavor": "rocm-docs-home"}
 
-html_static_path = ["sphinx/static/css"]
-html_css_files = ["rocm_custom.css", "rocm_rn.css"]
+html_static_path = ["sphinx/static/css", "extension/how-to/rocm-for-ai/inference"]
+html_css_files = ["rocm_custom.css", "rocm_rn.css", "vllm-benchmark.css"]
+html_js_files = ["vllm-benchmark.js"]
 
 html_title = "ROCm Documentation"
 

docs/data/how-to/rocm-for-ai/inference/vllm-benchmark-models.yaml

@@ -0,0 +1,153 @@
+vllm_benchmark:
+  unified_docker:
+    latest:
+      pull_tag: rocm/vllm:rocm6.3.1_mi300_ubuntu22.04_py3.12_vllm_0.6.6
+      docker_hub_url: https://hub.docker.com/layers/rocm/vllm/rocm6.3.1_mi300_ubuntu22.04_py3.12_vllm_0.6.6/images/sha256-9a12ef62bbbeb5a4c30a01f702c8e025061f575aa129f291a49fbd02d6b4d6c9
+      rocm_version: 6.3.1
+      vllm_version: 0.6.6
+      pytorch_version: 2.7.0 (2.7.0a0+git3a58512)
+  model_groups:
+    - group: Llama
+      tag: llama
+      models:
+      - model: Llama 3.1 8B
+        mad_tag: pyt_vllm_llama-3.1-8b
+        model_repo: meta-llama/Llama-3.1-8B-Instruct
+        url: https://huggingface.co/meta-llama/Llama-3.1-8B
+        precision: float16
+      - model: Llama 3.1 70B
+        mad_tag: pyt_vllm_llama-3.1-70b
+        model_repo: meta-llama/Llama-3.1-70B-Instruct
+        url: https://huggingface.co/meta-llama/Llama-3.1-70B-Instruct
+        precision: float16
+      - model: Llama 3.1 405B
+        mad_tag: pyt_vllm_llama-3.1-405b
+        model_repo: meta-llama/Llama-3.1-405B-Instruct
+        url: https://huggingface.co/meta-llama/Llama-3.1-405B-Instruct
+        precision: float16
+      - model: Llama 3.2 11B Vision
+        mad_tag: pyt_vllm_llama-3.2-11b-vision-instruct
+        model_repo: meta-llama/Llama-3.2-11B-Vision-Instruct
+        url: https://huggingface.co/meta-llama/Llama-3.2-11B-Vision-Instruct
+        precision: float16
+      - model: Llama 2 7B
+        mad_tag: pyt_vllm_llama-2-7b
+        model_repo: meta-llama/Llama-2-7b-chat-hf
+        url: https://huggingface.co/meta-llama/Llama-2-7b-chat-hf
+        precision: float16
+      - model: Llama 2 70B
+        mad_tag: pyt_vllm_llama-2-70b
+        model_repo: meta-llama/Llama-2-70b-chat-hf
+        url: https://huggingface.co/meta-llama/Llama-2-70b-chat-hf
+        precision: float16
+      - model: Llama 3.1 70B FP8
+        mad_tag: pyt_vllm_llama-3.1-70b_fp8
+        model_repo: amd/Llama-3.1-70B-Instruct-FP8-KV
+        url: https://huggingface.co/amd/Llama-3.1-70B-Instruct-FP8-KV
+        precision: float8
+      - model: Llama 3.1 405B FP8
+        mad_tag: pyt_vllm_llama-3.1-405b_fp8
+        model_repo: amd/Llama-3.1-405B-Instruct-FP8-KV
+        url: https://huggingface.co/amd/Llama-3.1-405B-Instruct-FP8-KV
+        precision: float8
+    - group: Mistral
+      tag: mistral
+      models:
+      - model: Mixtral MoE 8x7B
+        mad_tag: pyt_vllm_mixtral-8x7b
+        model_repo: mistralai/Mixtral-8x7B-Instruct-v0.1
+        url: https://huggingface.co/mistralai/Mixtral-8x7B-Instruct-v0.1
+        precision: float16
+      - model: Mixtral MoE 8x22B
+        mad_tag: pyt_vllm_mixtral-8x22b
+        model_repo: mistralai/Mixtral-8x22B-Instruct-v0.1
+        url: https://huggingface.co/mistralai/Mixtral-8x22B-Instruct-v0.1
+        precision: float16
+      - model: Mistral 7B
+        mad_tag: pyt_vllm_mistral-7b
+        model_repo: mistralai/Mistral-7B-Instruct-v0.3
+        url: https://huggingface.co/mistralai/Mistral-7B-Instruct-v0.3
+        precision: float16
+      - model: Mixtral MoE 8x7B FP8
+        mad_tag: pyt_vllm_mixtral-8x7b_fp8
+        model_repo: amd/Mixtral-8x7B-Instruct-v0.1-FP8-KV
+        url: https://huggingface.co/amd/Mixtral-8x7B-Instruct-v0.1-FP8-KV
+        precision: float8
+      - model: Mixtral MoE 8x22B FP8
+        mad_tag: pyt_vllm_mixtral-8x22b_fp8
+        model_repo: amd/Mixtral-8x22B-Instruct-v0.1-FP8-KV
+        url: https://huggingface.co/amd/Mixtral-8x22B-Instruct-v0.1-FP8-KV
+        precision: float8
+      - model: Mistral 7B FP8
+        mad_tag: pyt_vllm_mistral-7b_fp8
+        model_repo: amd/Mistral-7B-v0.1-FP8-KV
+        url: https://huggingface.co/amd/Mistral-7B-v0.1-FP8-KV
+        precision: float8
+    - group: Qwen
+      tag: qwen
+      models:
+      - model: Qwen2 7B
+        mad_tag: pyt_vllm_qwen2-7b
+        model_repo: Qwen/Qwen2-7B-Instruct
+        url: https://huggingface.co/Qwen/Qwen2-7B-Instruct
+        precision: float16
+      - model: Qwen2 72B
+        mad_tag: pyt_vllm_qwen2-72b
+        model_repo: Qwen/Qwen2-72B-Instruct
+        url: https://huggingface.co/Qwen/Qwen2-72B-Instruct
+        precision: float16
+    - group: JAIS
+      tag: jais
+      models:
+      - model: JAIS 13B
+        mad_tag: pyt_vllm_jais-13b
+        model_repo: core42/jais-13b-chat
+        url: https://huggingface.co/core42/jais-13b-chat
+        precision: float16
+      - model: JAIS 30B
+        mad_tag: pyt_vllm_jais-30b
+        model_repo: core42/jais-30b-chat-v3
+        url: https://huggingface.co/core42/jais-30b-chat-v3
+        precision: float16
+    - group: DBRX
+      tag: dbrx
+      models:
+      - model: DBRX Instruct
+        mad_tag: pyt_vllm_dbrx-instruct
+        model_repo: databricks/dbrx-instruct
+        url: https://huggingface.co/databricks/dbrx-instruct
+        precision: float16
+      - model: DBRX Instruct FP8
+        mad_tag: pyt_vllm_dbrx_fp8
+        model_repo: amd/dbrx-instruct-FP8-KV
+        url: https://huggingface.co/amd/dbrx-instruct-FP8-KV
+        precision: float8
+    - group: Gemma
+      tag: gemma
+      models:
+      - model: Gemma 2 27B
+        mad_tag: pyt_vllm_gemma-2-27b
+        model_repo: google/gemma-2-27b
+        url: https://huggingface.co/google/gemma-2-27b
+        precision: float16
+    - group: Cohere
+      tag: cohere
+      models:
+      - model: C4AI Command R+ 08-2024
+        mad_tag: pyt_vllm_c4ai-command-r-plus-08-2024
+        model_repo: CohereForAI/c4ai-command-r-plus-08-2024
+        url: https://huggingface.co/CohereForAI/c4ai-command-r-plus-08-2024
+        precision: float16
+      - model: C4AI Command R+ 08-2024 FP8
+        mad_tag: pyt_vllm_command-r-plus_fp8
+        model_repo: amd/c4ai-command-r-plus-FP8-KV
+        url: https://huggingface.co/amd/c4ai-command-r-plus-FP8-KV
+        precision: float8
+    - group: DeepSeek
+      tag: deepseek
+      models:
+      - model: DeepSeek MoE 16B
+        mad_tag: pyt_vllm_deepseek-moe-16b-chat
+        model_repo: deepseek-ai/deepseek-moe-16b-chat
+        url: https://huggingface.co/deepseek-ai/deepseek-moe-16b-chat
+        precision: float16

docs/extension/how-to/rocm-for-ai/inference/vllm-benchmark.js

@@ -0,0 +1,212 @@
+function ready(proc) {
+  // Check if page is loaded. If so, init.
+  if (document.readyState !== "loading") {
+    proc();
+  } else {
+    // Otherwise, wait for DOMContentLoaded event.
+    document.addEventListener("DOMContentLoaded", proc);
+  }
+}
+
+ready(() => {
+  const ModelPicker = {
+    // Selector strings for DOM elements
+    SELECTORS: {
+      CONTAINER: "#vllm-benchmark-ud-params-picker",
+      MODEL_GROUP_BTN: 'div[data-param-k="model-group"][data-param-v]',
+      MODEL_PARAM_BTN: 'div[data-param-k="model"][data-param-v]',
+      MODEL_DOC: "div.model-doc",
+    },
+    CSS_CLASSES: {
+      HIDDEN: "hidden",
+    },
+    ATTRIBUTES: {
+      PARAM_KEY: "data-param-k", // URL search parameter key (i.e., "model")
+      PARAM_VALUE: "data-param-v", // URL search param value (e.g., "pyt_vllm_llama-3.1-8b", "pyt_vllm_llama-3.1-70b") -- these are MAD model tags
+      PARAM_GROUP: "data-param-group", // Model group (e.g., "llama", "mistral")
+      PARAM_STATE: "data-param-state", // Selection state
+    },
+
+    // Cache DOM elements
+    elements: {
+      container: null,
+      modelGroups: null,
+      modelParams: null,
+      modelDocs: null,
+    },
+
+    data: {
+      availableModels: new Set(),
+      modelsByGroup: new Map(),
+      modelToGroupMap: new Map(),
+      formattedModelClassMap: new Map(), //TODO
+    },
+
+    init() {
+      this.elements.container = document.querySelector(
+        this.SELECTORS.CONTAINER,
+      );
+      if (!this.elements.container) return;
+
+      this.cacheDOMElements();
+      if (!this.validateElements()) return;
+
+      this.buildModelData();
+      this.bindEvents();
+      this.initializeState();
+    },
+
+    cacheDOMElements() {
+      const { CONTAINER, MODEL_GROUP_BTN, MODEL_PARAM_BTN, MODEL_DOC } =
+        this.SELECTORS;
+      this.elements = {
+        container: document.querySelector(CONTAINER),
+        modelGroups: document.querySelectorAll(MODEL_GROUP_BTN),
+        modelParams: document.querySelectorAll(MODEL_PARAM_BTN),
+        modelDocs: document.querySelectorAll(MODEL_DOC),
+      };
+    },
+
+    validateElements() {
+      const { modelGroups, modelParams } = this.elements;
+      if (!modelGroups.length || !modelParams.length) {
+        console.warn("Model picker is missing required elements");
+        return false;
+      }
+      return true;
+    },
+
+    buildModelData() {
+      const { PARAM_VALUE, PARAM_GROUP } = this.ATTRIBUTES;
+
+      this.elements.modelParams.forEach((model) => {
+        const modelTag = model.getAttribute(PARAM_VALUE);
+        const groupTag = model.getAttribute(PARAM_GROUP);
+
+        if (!modelTag || !groupTag) return;
+
+        this.data.availableModels.add(modelTag);
+        this.data.modelToGroupMap.set(modelTag, groupTag);
+
+        // FIXME: this is because Sphinx auto-formats class names to use dashes
+        this.data.formattedModelClassMap.set(
+          modelTag,
+          modelTag.replace(/[^a-zA-Z0-9]/g, "-"),
+        );
+
+        if (!this.data.modelsByGroup.has(groupTag)) {
+          this.data.modelsByGroup.set(groupTag, []);
+        }
+        this.data.modelsByGroup.get(groupTag).push(modelTag);
+      });
+    },
+
+    // Event listeners for user interactions
+    bindEvents() {
+      const handleInteraction = (event) => {
+        const target = event.target.closest(`[${this.ATTRIBUTES.PARAM_KEY}]`);
+        if (!target) return;
+
+        const paramType = target.getAttribute(this.ATTRIBUTES.PARAM_KEY);
+        const paramValue = target.getAttribute(this.ATTRIBUTES.PARAM_VALUE);
+
+        if (paramType === "model") {
+          const groupTag = target.getAttribute(this.ATTRIBUTES.PARAM_GROUP);
+          if (groupTag) this.updateUI(paramValue, groupTag);
+        } else if (paramType === "model-group") {
+          const firstModelInGroup = this.data.modelsByGroup.get(paramValue)
+            ?.[0];
+          if (firstModelInGroup) this.updateUI(firstModelInGroup, paramValue);
+        }
+      };
+
+      this.elements.container.addEventListener("click", handleInteraction);
+      this.elements.container.addEventListener("keydown", (event) => {
+        if (event.key === "Enter" || event.key === " ") {
+          event.preventDefault();
+          handleInteraction(event);
+        }
+      });
+    },
+
+    // Update the page based on the selected model
+    updateUI(modelTag, groupTag) {
+      const validModel = this.setModelSearchParam(modelTag);
+
+      // Update model group buttons
+      this.elements.modelGroups.forEach((group) => {
+        const isSelected =
+          group.getAttribute(this.ATTRIBUTES.PARAM_VALUE) === groupTag;
+        group.setAttribute(
+          this.ATTRIBUTES.PARAM_STATE,
+          isSelected ? "selected" : "",
+        );
+        group.setAttribute("aria-selected", isSelected.toString());
+      });
+
+      // Update model buttons
+      this.elements.modelParams.forEach((model) => {
+        const isInSelectedGroup =
+          model.getAttribute(this.ATTRIBUTES.PARAM_GROUP) === groupTag;
+        const isSelectedModel =
+          model.getAttribute(this.ATTRIBUTES.PARAM_VALUE) === validModel;
+
+        model.classList.toggle(this.CSS_CLASSES.HIDDEN, !isInSelectedGroup);
+        model.setAttribute(
+          this.ATTRIBUTES.PARAM_STATE,
+          isSelectedModel ? "selected" : "",
+        );
+        model.setAttribute("aria-selected", isSelectedModel.toString());
+      });
+
+      // Update visibility of doc sections
+      const formattedClass = this.data.formattedModelClassMap.get(validModel);
+      if (formattedClass) {
+        this.elements.modelDocs.forEach((doc) => {
+          doc.classList.toggle(
+            this.CSS_CLASSES.HIDDEN,
+            !doc.classList.contains(formattedClass),
+          );
+        });
+      }
+    },
+
+    // Get the current model from the URL search parameters.
+    getModelSearchParam() {
+      return new URLSearchParams(location.search).get("model");
+    },
+
+    // Set the model in the URL search parameters, or fallback to the first available one.
+    setModelSearchParam(modelTag) {
+      const defaultModel = [...this.data.availableModels][0];
+      const model = this.data.availableModels.has(modelTag)
+        ? modelTag
+        : defaultModel;
+
+      const searchParams = new URLSearchParams(location.search);
+      searchParams.set("model", model);
+
+      history.replaceState(
+        {},
+        "",
+        `${location.pathname}?${searchParams.toString()}`,
+      );
+      return model;
+    },
+
+    // Initialize the UI state based on the current URL search parameter or default values.
+    initializeState() {
+      const currentModel = this.getModelSearchParam();
+      const validModel = this.setModelSearchParam(currentModel);
+
+      const initialGroup = this.data.modelToGroupMap.get(validModel) ??
+        [...this.data.modelsByGroup.keys()][0];
+
+      if (initialGroup) {
+        this.updateUI(validModel, initialGroup);
+      }
+    },
+  };
+
+  ModelPicker.init();
+});

docs/extension/version-ref.py

@@ -0,0 +1,266 @@
+from docutils import nodes
+from docutils.parsers.rst import Directive
+from sphinx.util import logging
+import csv
+from io import StringIO
+import re
+import shlex
+
+logger = logging.getLogger(__name__)
+
+class VersionReference(nodes.Inline, nodes.TextElement):
+    """Represents an inline version reference."""
+    pass
+
+class VersionSetDirective(Directive):
+    """Directive for setting version references within a page scope."""
+    required_arguments = 2  # name and value
+    optional_arguments = 0
+
+    def run(self):
+        env = self.state.document.settings.env
+        if not hasattr(env, 'doc_version_refs'):
+            env.doc_version_refs = {}
+        current_doc = env.docname
+        if current_doc not in env.doc_version_refs:
+            env.doc_version_refs[current_doc] = {}
+
+        name, value = self.arguments
+        if name.lower() == 'latest':
+            logger.warning('Cannot override the "latest" keyword with version-set')
+            return []
+
+        # Handle 'latest' value by getting the actual version
+        if value.lower() == 'latest':
+            data = getattr(env, 'compatibility_matrix', None)
+            if data:
+                latest_version = get_latest_rocm_version(data)
+                if latest_version:
+                    value = latest_version
+
+        env.doc_version_refs[current_doc][name] = value
+        return []
+
+def clean_library_name(name):
+    """Extract library name from RST formatting."""
+    # Handle :doc: format
+    doc_match = re.search(r':doc:`([^<]+)(?:\s+<[^>]+>)?`', name)
+    if doc_match:
+        return doc_match.group(1).strip()
+
+    # Handle other link formats
+    link_match = re.search(r'`([^<]+)(?:\s+<[^>]+>)?`_?', name)
+    if link_match:
+        return link_match.group(1).strip()
+
+    return name.strip()
+
+def get_latest_rocm_version(data):
+    """Get the latest ROCm version from the matrix headers."""
+    if not data or len(data) == 0:
+        return None
+
+    # Get all column names except 'ROCm Version'
+    columns = [col for col in data[0].keys() if col != 'ROCm Version']
+    # Return the first column name (assumed to be the latest version)
+    return columns[0] if columns else None
+
+def version_role(name, rawtext, text, lineno, inliner, options={}, content=[]):
+    """
+    Role function to print version value.
+    Usage: :version:`version_name`
+    """
+    try:
+        version_name = text.strip()
+        env = inliner.document.settings.env
+
+        if hasattr(env, 'doc_version_refs'):
+            current_doc = env.docname
+            if current_doc in env.doc_version_refs:
+                doc_refs = env.doc_version_refs[current_doc]
+                if version_name in doc_refs:
+                    version = doc_refs[version_name]
+                    node = nodes.Text(version)
+                    return [node], []
+
+        msg = inliner.reporter.warning(
+            f'No version defined for name {version_name}',
+            line=lineno
+        )
+        return [], [msg]
+
+    except Exception as e:
+        msg = inliner.reporter.error(
+            f'Error looking up version: {str(e)}',
+            line=lineno
+        )
+        prb = inliner.problematic(rawtext, rawtext, msg)
+        return [prb], [msg]
+
+def version_ref_role(name, rawtext, text, lineno, inliner, options={}, content=[]):
+    """
+    Role function for version references.
+    Usage: :version-ref:`library_name release`
+           :version-ref:`"library name" release`
+           :version-ref:`library_name latest`
+           :version-ref:`rocm latest`
+    """
+    try:
+        # Parse the text - handle both quoted and unquoted formats
+        if '"' in text:
+            parts = shlex.split(text)
+        else:
+            parts = text.split()
+
+        if len(parts) != 2:
+            msg = inliner.reporter.error(
+                'Version reference must be in format "library_name release" or "\\"library name\\" release"',
+                line=lineno
+            )
+            prb = inliner.problematic(rawtext, rawtext, msg)
+            return [prb], [msg]
+
+        library_name, release = parts
+        env = inliner.document.settings.env
+
+        # Check if release is a version reference in current document
+        if hasattr(env, 'doc_version_refs'):
+            current_doc = env.docname
+            if current_doc in env.doc_version_refs:
+                doc_refs = env.doc_version_refs[current_doc]
+                if release in doc_refs:
+                    release = doc_refs[release]
+
+        # Handle special case for "rocm latest"
+        if library_name.lower() == 'rocm' and release.lower() == 'latest':
+            data = getattr(env, 'compatibility_matrix', None)
+            if not data:
+                raise ValueError("Compatibility matrix not found in environment")
+
+            latest_version = get_latest_rocm_version(data)
+            if latest_version:
+                node = VersionReference()
+                node += nodes.Text(latest_version)
+                return [node], []
+            else:
+                msg = inliner.reporter.warning(
+                    'No ROCm versions found in compatibility matrix',
+                    line=lineno
+                )
+                return [], [msg]
+
+        version = lookup_version(inliner, library_name, release)
+
+        if version:
+            node = VersionReference()
+            node += nodes.Text(version)
+            return [node], []
+        else:
+            msg = inliner.reporter.warning(
+                f'No version found for library {library_name} in release {release}',
+                line=lineno
+            )
+            return [], [msg]
+
+    except Exception as e:
+        msg = inliner.reporter.error(
+            f'Error looking up version: {str(e)}',
+            line=lineno
+        )
+        prb = inliner.problematic(rawtext, rawtext, msg)
+        return [prb], [msg]
+
+def lookup_version(inliner, library_name, release):
+    """Look up the version in the compatibility matrix."""
+    env = inliner.document.settings.env
+    data = getattr(env, 'compatibility_matrix', None)
+
+    if not data:
+        raise ValueError("Compatibility matrix not found in environment")
+
+    # Handle the 'latest' keyword
+    if release.lower() == 'latest':
+        latest_version = get_latest_rocm_version(data)
+        if not latest_version:
+            return None
+        release = latest_version
+
+    # For ROCm, check if the version exists in column headers
+    if library_name.lower() == 'rocm':
+        columns = [col for col in data[0].keys() if col != 'ROCm Version']
+        if release in columns:
+            return release
+        return None
+
+    # Find the library version
+    for row in data:
+        row_lib_name = clean_library_name(row['ROCm Version'])
+        if row_lib_name == library_name:
+            # Get the version, removing any whitespace
+            version = row.get(release, '').strip()
+            if version:
+                return version
+
+    # If not found, try a case-insensitive search
+    for row in data:
+        row_lib_name = clean_library_name(row['ROCm Version'])
+        if row_lib_name.lower() == library_name.lower():
+            version = row.get(release, '').strip()
+            if version:
+                return version
+
+    return None
+
+def visit_version_reference(self, node):
+    self.body.append(f'<span class="version-reference">')
+
+def depart_version_reference(self, node):
+    self.body.append('</span>')
+
+def load_compatibility_matrix(app):
+    """Load the compatibility matrix content from CSV."""
+    if not app.config.compatibility_matrix_file:
+        logger.warning('No compatibility matrix file configured')
+        return
+
+    try:
+        with open(app.config.compatibility_matrix_file, 'r', encoding='utf-8') as f:
+            reader = csv.DictReader(f)
+            app.env.compatibility_matrix = list(reader)
+            logger.info('Successfully loaded compatibility matrix')
+
+            # Debug: print first few rows with their library names
+            for row in list(app.env.compatibility_matrix)[:5]:
+                if 'ROCm Version' in row:
+                    lib_name = clean_library_name(row['ROCm Version'])
+                    logger.debug(f"Loaded library: {lib_name}")
+
+    except Exception as e:
+        logger.error(f'Error loading compatibility matrix: {str(e)}')
+
+def purge_version_refs(app, env, docname):
+    """Remove version references for a document when it is purged"""
+    if hasattr(env, 'doc_version_refs'):
+        if docname in env.doc_version_refs:
+            del env.doc_version_refs[docname]
+
+def setup(app):
+    app.add_node(VersionReference,
+                 html=(visit_version_reference, depart_version_reference))
+    app.add_role('version-ref', version_ref_role)
+    app.add_role('version', version_role)
+    app.add_directive('version-set', VersionSetDirective)
+
+    # Add a config value for the compatibility matrix file path
+    app.add_config_value('compatibility_matrix_file', None, 'env')
+
+    # Connect to the builder-inited event to load the matrix
+    app.connect('builder-inited', load_compatibility_matrix)
+
+    # Connect to env-purge-doc event to clean up document-specific version refs
+    app.connect('env-purge-doc', purge_version_refs)
+
+    return {
+        'parallel_read_safe': True,
+        'parallel_write_safe': True,
+    }

docs/how-to/rocm-for-ai/inference/vllm-benchmark.rst

@@ -9,422 +9,266 @@ LLM inference performance validation on AMD Instinct MI300X
 
 .. _vllm-benchmark-unified-docker:
 
-The `ROCm vLLM Docker <https://hub.docker.com/r/rocm/vllm/tags>`_ image offers
-a prebuilt, optimized environment for validating large language model (LLM)
-inference performance on the AMD Instinct™ MI300X accelerator. This ROCm vLLM
-Docker image integrates vLLM and PyTorch tailored specifically for the MI300X
-accelerator and includes the following components:
+.. datatemplate:yaml:: /data/how-to/rocm-for-ai/inference/vllm-benchmark-models.yaml
 
-* `ROCm 6.3.1 <https://github.com/ROCm/ROCm>`_
+   {% set unified_docker = data.vllm_benchmark.unified_docker.latest %}
+   {% set model_groups = data.vllm_benchmark.model_groups %}
 
-* `vLLM 0.6.6 <https://docs.vllm.ai/en/latest>`_
+   The `ROCm vLLM Docker <{{ unified_docker.docker_hub_url }}>`_ image offers
+   a prebuilt, optimized environment for validating large language model (LLM)
+   inference performance on the AMD Instinct™ MI300X accelerator. This ROCm vLLM
+   Docker image integrates vLLM and PyTorch tailored specifically for the MI300X
+   accelerator and includes the following components:
 
-* `PyTorch 2.7.0 (2.7.0a0+git3a58512) <https://github.com/pytorch/pytorch>`_
+   * `ROCm {{ unified_docker.rocm_version }} <https://github.com/ROCm/ROCm>`_
 
-With this Docker image, you can quickly validate the expected inference
-performance numbers for the MI300X accelerator. This topic also provides tips on
-optimizing performance with popular AI models. For more information, see the lists of
-:ref:`available models for MAD-integrated benchmarking <vllm-benchmark-mad-models>`
-and :ref:`standalone benchmarking <vllm-benchmark-standalone-options>`.
+   * `vLLM {{ unified_docker.vllm_version }} <https://docs.vllm.ai/en/latest>`_
 
-.. _vllm-benchmark-vllm:
+   * `PyTorch {{ unified_docker.pytorch_version }} <https://github.com/pytorch/pytorch>`_
 
-.. note::
+   With this Docker image, you can quickly validate the expected inference
+   performance numbers for the MI300X accelerator. This topic also provides tips on
+   optimizing performance with popular AI models.
 
-   vLLM is a toolkit and library for LLM inference and serving. AMD implements
-   high-performance custom kernels and modules in vLLM to enhance performance.
-   See :ref:`fine-tuning-llms-vllm` and :ref:`mi300x-vllm-optimization` for
-   more information.
+   .. _vllm-benchmark-available-models:
 
-Getting started
-===============
+   Available models
+   ================
 
-Use the following procedures to reproduce the benchmark results on an
-MI300X accelerator with the prebuilt vLLM Docker image.
+   .. raw:: html
 
-.. _vllm-benchmark-get-started:
+      <div id="vllm-benchmark-ud-params-picker" class="container-fluid">
+        <div class="row">
+          <div class="col-2 me-2 model-param-head">Model</div>
+          <div class="row col-10">
+   {% for model_group in model_groups %}
+            <div class="col-3 model-param" data-param-k="model-group" data-param-v="{{ model_group.tag }}" tabindex="0">{{ model_group.group }}</div>
+   {% endfor %}
+          </div>
+        </div>
 
-1. Disable NUMA auto-balancing.
+        <div class="row mt-1">
+          <div class="col-2 me-2 model-param-head">Model variant</div>
+          <div class="row col-10">
+   {% for model_group in model_groups %}
+      {% set models = model_group.models %}
+      {% for model in models %}
+         {% if models|length % 3 == 0 %}
+            <div class="col-4 model-param" data-param-k="model" data-param-v="{{ model.mad_tag }}" data-param-group="{{ model_group.tag }}" tabindex="0">{{ model.model }}</div>
+         {% else %}
+            <div class="col-6 model-param" data-param-k="model" data-param-v="{{ model.mad_tag }}" data-param-group="{{ model_group.tag }}" tabindex="0">{{ model.model }}</div>
+         {% endif %}
+      {% endfor %}
+   {% endfor %}
+          </div>
+        </div>
+      </div>
 
-   To optimize performance, disable automatic NUMA balancing. Otherwise, the GPU
-   might hang until the periodic balancing is finalized. For more information,
-   see :ref:`AMD Instinct MI300X system optimization <mi300x-disable-numa>`.
+   .. _vllm-benchmark-vllm:
 
-   .. code-block:: shell
+   {% for model_group in model_groups %}
+      {% for model in model_group.models %}
 
-      # disable automatic NUMA balancing
-      sh -c 'echo 0 > /proc/sys/kernel/numa_balancing'
-      # check if NUMA balancing is disabled (returns 0 if disabled)
-      cat /proc/sys/kernel/numa_balancing
-      0
+   .. container:: model-doc {{model.mad_tag}}
 
-2. Download the :ref:`ROCm vLLM Docker image <vllm-benchmark-unified-docker>`.
+      .. note::
 
-   Use the following command to pull the Docker image from Docker Hub.
+         See the `{{ model.model }} model card on Hugging Face <{{ model.url }}>`_ to learn more about your selected model.
+         Some models require access authorization prior to use via an external license agreement through a third party.
 
-   .. code-block:: shell
+      {% endfor %}
+   {% endfor %}
 
-      docker pull rocm/vllm:rocm6.3.1_mi300_ubuntu22.04_py3.12_vllm_0.6.6
 
-Once the setup is complete, choose between two options to reproduce the
-benchmark results:
+   .. note::
 
--  :ref:`MAD-integrated benchmarking <vllm-benchmark-mad>`
+      vLLM is a toolkit and library for LLM inference and serving. AMD implements
+      high-performance custom kernels and modules in vLLM to enhance performance.
+      See :ref:`fine-tuning-llms-vllm` and :ref:`mi300x-vllm-optimization` for
+      more information.
 
--  :ref:`Standalone benchmarking <vllm-benchmark-standalone>`
+   Getting started
+   ===============
 
-.. _vllm-benchmark-mad:
+   Use the following procedures to reproduce the benchmark results on an
+   MI300X accelerator with the prebuilt vLLM Docker image.
 
-MAD-integrated benchmarking
-===========================
+   .. _vllm-benchmark-get-started:
 
-Clone the ROCm Model Automation and Dashboarding (`<https://github.com/ROCm/MAD>`__) repository to a local
-directory and install the required packages on the host machine.
+   1. Disable NUMA auto-balancing.
 
-.. code-block:: shell
+      To optimize performance, disable automatic NUMA balancing. Otherwise, the GPU
+      might hang until the periodic balancing is finalized. For more information,
+      see :ref:`AMD Instinct MI300X system optimization <mi300x-disable-numa>`.
 
-   git clone https://github.com/ROCm/MAD
-   cd MAD
-   pip install -r requirements.txt
+      .. code-block:: shell
 
-Use this command to run a performance benchmark test of the Llama 3.1 8B model
-on one GPU with ``float16`` data type in the host machine.
+         # disable automatic NUMA balancing
+         sh -c 'echo 0 > /proc/sys/kernel/numa_balancing'
+         # check if NUMA balancing is disabled (returns 0 if disabled)
+         cat /proc/sys/kernel/numa_balancing
+         0
 
-.. code-block:: shell
+   2. Download the `ROCm vLLM Docker image <{{ unified_docker.docker_hub_url }}>`_.
 
-   export MAD_SECRETS_HFTOKEN="your personal Hugging Face token to access gated models"
-   python3 tools/run_models.py --tags pyt_vllm_llama-3.1-8b --keep-model-dir --live-output --timeout 28800
+      Use the following command to pull the Docker image from Docker Hub.
 
-ROCm MAD launches a Docker container with the name
-``container_ci-pyt_vllm_llama-3.1-8b``. The latency and throughput reports of the
-model are collected in the following path: ``~/MAD/reports_float16/``.
+      .. code-block:: shell
 
-Although the following models are preconfigured to collect latency and
-throughput performance data, you can also change the benchmarking parameters.
-Refer to the :ref:`Standalone benchmarking <vllm-benchmark-standalone>` section.
+         docker pull {{ unified_docker.pull_tag }}
 
-.. _vllm-benchmark-mad-models:
+   Benchmarking
+   ============
 
-Available models
-----------------
+   Once the setup is complete, choose between two options to reproduce the
+   benchmark results:
 
-.. list-table::
-   :header-rows: 1
-   :widths: 2, 3
+   .. _vllm-benchmark-mad:
 
-   * - Model name
-     - Tag
+   {% for model_group in model_groups %}
+      {% for model in model_group.models %}
 
-   * - `Llama 3.1 8B <https://huggingface.co/meta-llama/Llama-3.1-8B>`_
-     - ``pyt_vllm_llama-3.1-8b``
+   .. container:: model-doc {{model.mad_tag}}
 
-   * - `Llama 3.1 70B <https://huggingface.co/meta-llama/Llama-3.1-70B-Instruct>`_
-     - ``pyt_vllm_llama-3.1-70b``
+      .. tab-set::
 
-   * - `Llama 3.1 405B <https://huggingface.co/meta-llama/Llama-3.1-405B-Instruct>`_
-     - ``pyt_vllm_llama-3.1-405b``
+         .. tab-item:: MAD-integrated benchmarking
 
-   * - `Llama 3.2 11B Vision <https://huggingface.co/meta-llama/Llama-3.2-11B-Vision-Instruct>`_
-     - ``pyt_vllm_llama-3.2-11b-vision-instruct``
+            Clone the ROCm Model Automation and Dashboarding (`<https://github.com/ROCm/MAD>`__) repository to a local
+            directory and install the required packages on the host machine.
 
-   * - `Llama 2 7B <https://huggingface.co/meta-llama/Llama-2-7b-chat-hf>`_
-     - ``pyt_vllm_llama-2-7b``
+            .. code-block:: shell
 
-   * - `Llama 2 70B <https://huggingface.co/meta-llama/Llama-2-70b-chat-hf>`_
-     - ``pyt_vllm_llama-2-70b``
+               git clone https://github.com/ROCm/MAD
+               cd MAD
+               pip install -r requirements.txt
 
-   * - `Mixtral MoE 8x7B <https://huggingface.co/mistralai/Mixtral-8x7B-Instruct-v0.1>`_
-     - ``pyt_vllm_mixtral-8x7b``
+            Use this command to run the performance benchmark test on the `{{model.model}} <{{ model.url }}>`_ model
+            using one GPU with the ``{{model.precision}}`` data type on the host machine.
 
-   * - `Mixtral MoE 8x22B <https://huggingface.co/mistralai/Mixtral-8x22B-Instruct-v0.1>`_
-     - ``pyt_vllm_mixtral-8x22b``
+            .. code-block:: shell
 
-   * - `Mistral 7B <https://huggingface.co/mistralai/Mistral-7B-Instruct-v0.3>`_
-     - ``pyt_vllm_mistral-7b``
+               export MAD_SECRETS_HFTOKEN="your personal Hugging Face token to access gated models"
+               python3 tools/run_models.py --tags {{model.mad_tag}} --keep-model-dir --live-output --timeout 28800
 
-   * - `Qwen2 7B <https://huggingface.co/Qwen/Qwen2-7B-Instruct>`_
-     - ``pyt_vllm_qwen2-7b``
+            MAD launches a Docker container with the name
+            ``container_ci-{{model.mad_tag}}``. The latency and throughput reports of the
+            model are collected in the following path: ``~/MAD/reports_{{model.precision}}/``.
 
-   * - `Qwen2 72B <https://huggingface.co/Qwen/Qwen2-72B-Instruct>`_
-     - ``pyt_vllm_qwen2-72b``
+            Although the :ref:`available models <vllm-benchmark-available-models>` are preconfigured
+            to collect latency and throughput performance data, you can also change the benchmarking
+            parameters. See the standalone benchmarking tab for more information.
 
-   * - `JAIS 13B <https://huggingface.co/core42/jais-13b-chat>`_
-     - ``pyt_vllm_jais-13b``
+         .. tab-item:: Standalone benchmarking
 
-   * - `JAIS 30B <https://huggingface.co/core42/jais-30b-chat-v3>`_
-     - ``pyt_vllm_jais-30b``
+            Run the vLLM benchmark tool independently by starting the
+            `Docker container <https://hub.docker.com/layers/rocm/vllm/rocm6.3.1_mi300_ubuntu22.04_py3.12_vllm_0.6.6/images/sha256-9a12ef62bbbeb5a4c30a01f702c8e025061f575aa129f291a49fbd02d6b4d6c9>`_
+            as shown in the following snippet.
 
-   * - `DBRX Instruct <https://huggingface.co/databricks/dbrx-instruct>`_
-     - ``pyt_vllm_dbrx-instruct``
+            .. code-block::
 
-   * - `Gemma 2 27B <https://huggingface.co/google/gemma-2-27b>`_
-     - ``pyt_vllm_gemma-2-27b``
+               docker pull rocm/vllm:rocm6.3.1_mi300_ubuntu22.04_py3.12_vllm_0.6.6
+               docker run -it --device=/dev/kfd --device=/dev/dri --group-add video --shm-size 16G --security-opt seccomp=unconfined --security-opt apparmor=unconfined --cap-add=SYS_PTRACE -v $(pwd):/workspace --env HUGGINGFACE_HUB_CACHE=/workspace --name vllm_v0.6.6 rocm/vllm:rocm6.3.1_mi300_ubuntu22.04_py3.12_vllm_0.6.6
 
-   * - `C4AI Command R+ 08-2024 <https://huggingface.co/CohereForAI/c4ai-command-r-plus-08-2024>`_
-     - ``pyt_vllm_c4ai-command-r-plus-08-2024``
+            In the Docker container, clone the ROCm MAD repository and navigate to the
+            benchmark scripts directory at ``~/MAD/scripts/vllm``.
 
-   * - `DeepSeek MoE 16B <https://huggingface.co/deepseek-ai/deepseek-moe-16b-chat>`_
-     - ``pyt_vllm_deepseek-moe-16b-chat``
+            .. code-block::
 
-   * - `Llama 3.1 70B FP8 <https://huggingface.co/amd/Llama-3.1-70B-Instruct-FP8-KV>`_
-     - ``pyt_vllm_llama-3.1-70b_fp8``
+               git clone https://github.com/ROCm/MAD
+               cd MAD/scripts/vllm
 
-   * - `Llama 3.1 405B FP8 <https://huggingface.co/amd/Llama-3.1-405B-Instruct-FP8-KV>`_
-     - ``pyt_vllm_llama-3.1-405b_fp8``
+            To start the benchmark, use the following command with the appropriate options.
 
-   * - `Mixtral MoE 8x7B FP8 <https://huggingface.co/amd/Mixtral-8x7B-Instruct-v0.1-FP8-KV>`_
-     - ``pyt_vllm_mixtral-8x7b_fp8``
+            .. code-block::
 
-   * - `Mixtral MoE 8x22B FP8 <https://huggingface.co/amd/Mixtral-8x22B-Instruct-v0.1-FP8-KV>`_
-     - ``pyt_vllm_mixtral-8x22b_fp8``
+               ./vllm_benchmark_report.sh -s $test_option -m {{model.model_repo}} -g $num_gpu -d {{model.precision}}
 
-   * - `Mistral 7B FP8 <https://huggingface.co/amd/Mistral-7B-v0.1-FP8-KV>`_
-     - ``pyt_vllm_mistral-7b_fp8``
+            .. list-table::
+               :header-rows: 1
+               :align: center
 
-   * - `DBRX Instruct FP8 <https://huggingface.co/amd/dbrx-instruct-FP8-KV>`_
-     - ``pyt_vllm_dbrx_fp8``
+               * - Name
+                 - Options
+                 - Description
 
-   * - `C4AI Command R+ 08-2024 FP8 <https://huggingface.co/amd/c4ai-command-r-plus-FP8-KV>`_
-     - ``pyt_vllm_command-r-plus_fp8``
+               * - ``$test_option``
+                 - latency
+                 - Measure decoding token latency
 
-.. _vllm-benchmark-standalone:
+               * -
+                 - throughput
+                 - Measure token generation throughput
 
-Standalone benchmarking
-=======================
+               * -
+                 - all
+                 - Measure both throughput and latency
 
-You can run the vLLM benchmark tool independently by starting the
-`Docker container <https://hub.docker.com/layers/rocm/vllm/rocm6.3.1_mi300_ubuntu22.04_py3.12_vllm_0.6.6/images/sha256-9a12ef62bbbeb5a4c30a01f702c8e025061f575aa129f291a49fbd02d6b4d6c9>`_
-as shown in the following snippet.
+               * - ``$num_gpu``
+                 - 1 or 8
+                 - Number of GPUs
 
-.. code-block::
+               * - ``$datatype``
+                 - ``float16`` or ``float8``
+                 - Data type
 
-   docker pull rocm/vllm:rocm6.3.1_mi300_ubuntu22.04_py3.12_vllm_0.6.6
-   docker run -it --device=/dev/kfd --device=/dev/dri --group-add video --shm-size 16G --security-opt seccomp=unconfined --security-opt apparmor=unconfined --cap-add=SYS_PTRACE -v $(pwd):/workspace --env HUGGINGFACE_HUB_CACHE=/workspace --name vllm_v0.6.6 rocm/vllm:rocm6.3.1_mi300_ubuntu22.04_py3.12_vllm_0.6.6
+            .. note::
 
-In the Docker container, clone the ROCm MAD repository and navigate to the
-benchmark scripts directory at ``~/MAD/scripts/vllm``.
+               The input sequence length, output sequence length, and tensor parallel (TP) are
+               already configured. You don't need to specify them with this script.
 
-.. code-block::
+            .. note::
 
-   git clone https://github.com/ROCm/MAD
-   cd MAD/scripts/vllm
+               If you encounter the following error, pass your access-authorized Hugging
+               Face token to the gated models.
 
-Command
--------
+               .. code-block::
 
-To start the benchmark, use the following command with the appropriate options.
-See :ref:`Options <vllm-benchmark-standalone-options>` for the list of
-options and their descriptions.
+                  OSError: You are trying to access a gated repo.
 
-.. code-block:: shell
+                  # pass your HF_TOKEN
+                  export HF_TOKEN=$your_personal_hf_token
 
-   ./vllm_benchmark_report.sh -s $test_option -m $model_repo -g $num_gpu -d $datatype
+            Here are some examples of running the benchmark with various options.
 
-See the :ref:`examples <vllm-benchmark-run-benchmark>` for more information.
+            * Latency benchmark
 
-.. note::
+              Use this command to benchmark the latency of the {{model.model}} model on eight GPUs with the ``{{model.precision}}`` data type.
 
-   The input sequence length, output sequence length, and tensor parallel (TP) are
-   already configured. You don't need to specify them with this script.
+              .. code-block::
 
-.. note::
+                 ./vllm_benchmark_report.sh -s latency -m {{model.model_repo}} -g 8 -d {{model.precision}}
 
-   If you encounter the following error, pass your access-authorized Hugging
-   Face token to the gated models.
+              Find the latency report at ``./reports_{{model.precision}}_vllm_rocm{{unified_docker.rocm_version}}/summary/{{model.model_repo.split('/', 1)[1] if '/' in model.model_repo else model.model_repo}}_latency_report.csv``.
 
-   .. code-block:: shell
+            * Throughput benchmark
 
-      OSError: You are trying to access a gated repo.
+              Use this command to throughput the latency of the {{model.model}} model on eight GPUs with the ``{{model.precision}}`` data type.
 
-      # pass your HF_TOKEN
-      export HF_TOKEN=$your_personal_hf_token
+              .. code-block:: shell
 
-.. _vllm-benchmark-standalone-options:
+                 ./vllm_benchmark_report.sh -s latency -m {{model.model_repo}} -g 8 -d {{model.precision}}
 
-Options and available models
-----------------------------
+              Find the throughput report at ``./reports_{{model.precision}}_vllm_rocm{{unified_docker.rocm_version}}/summary/{{model.model_repo.split('/', 1)[1] if '/' in model.model_repo else model.model_repo}}_throughput_report.csv``.
 
-.. list-table::
-   :header-rows: 1
-   :align: center
+            .. raw:: html
 
-   * - Name
-     - Options
-     - Description
+               <style>
+               mjx-container[jax="CHTML"][display="true"] {
+                  text-align: left;
+                  margin: 0;
+               }
+               </style>
 
-   * - ``$test_option``
-     - latency
-     - Measure decoding token latency
+            .. note::
 
-   * -
-     - throughput
-     - Measure token generation throughput
+               Throughput is calculated as:
 
-   * -
-     - all
-     - Measure both throughput and latency
+               - .. math:: throughput\_tot = requests \times (\mathsf{\text{input lengths}} + \mathsf{\text{output lengths}}) / elapsed\_time
 
-   * - ``$model_repo``
-     - ``meta-llama/Llama-3.1-8B-Instruct``
-     - `Llama 3.1 8B <https://huggingface.co/meta-llama/Llama-3.1-8B>`_
-
-   * - (``float16``)
-     - ``meta-llama/Llama-3.1-70B-Instruct``
-     - `Llama 3.1 70B <https://huggingface.co/meta-llama/Llama-3.1-70B-Instruct>`_
-
-   * -
-     - ``meta-llama/Llama-3.1-405B-Instruct``
-     - `Llama 3.1 405B <https://huggingface.co/meta-llama/Llama-3.1-405B-Instruct>`_
-
-   * -
-     - ``meta-llama/Llama-3.2-11B-Vision-Instruct``
-     - `Llama 3.2 11B Vision <https://huggingface.co/meta-llama/Llama-3.2-11B-Vision-Instruct>`_
-
-   * -
-     - ``meta-llama/Llama-2-7b-chat-hf``
-     - `Llama 2 7B <https://huggingface.co/meta-llama/Llama-2-7b-chat-hf>`_
-
-   * -
-     - ``meta-llama/Llama-2-70b-chat-hf``
-     - `Llama 2 7B <https://huggingface.co/meta-llama/Llama-2-70b-chat-hf>`_
-
-   * -
-     - ``mistralai/Mixtral-8x7B-Instruct-v0.1``
-     - `Mixtral MoE 8x7B <https://huggingface.co/mistralai/Mixtral-8x7B-Instruct-v0.1>`_
-
-   * -
-     - ``mistralai/Mixtral-8x22B-Instruct-v0.1``
-     - `Mixtral MoE 8x22B <https://huggingface.co/mistralai/Mixtral-8x22B-Instruct-v0.1>`_
-
-   * -
-     - ``mistralai/Mistral-7B-Instruct-v0.3``
-     - `Mistral 7B <https://huggingface.co/mistralai/Mistral-7B-Instruct-v0.3>`_
-
-   * -
-     - ``Qwen/Qwen2-7B-Instruct``
-     - `Qwen2 7B <https://huggingface.co/Qwen/Qwen2-7B-Instruct>`_
-
-   * -
-     - ``Qwen/Qwen2-72B-Instruct``
-     - `Qwen2 72B <https://huggingface.co/Qwen/Qwen2-72B-Instruct>`_
-
-   * -
-     - ``core42/jais-13b-chat``
-     - `JAIS 13B <https://huggingface.co/core42/jais-13b-chat>`_
-
-   * -
-     - ``core42/jais-30b-chat-v3``
-     - `JAIS 30B <https://huggingface.co/core42/jais-30b-chat-v3>`_
-
-   * -
-     - ``databricks/dbrx-instruct``
-     - `DBRX Instruct <https://huggingface.co/databricks/dbrx-instruct>`_
-
-   * -
-     - ``google/gemma-2-27b``
-     - `Gemma 2 27B <https://huggingface.co/google/gemma-2-27b>`_
-
-   * -
-     - ``CohereForAI/c4ai-command-r-plus-08-2024``
-     - `C4AI Command R+ 08-2024 <https://huggingface.co/CohereForAI/c4ai-command-r-plus-08-2024>`_
-
-   * -
-     - ``deepseek-ai/deepseek-moe-16b-chat``
-     - `DeepSeek MoE 16B <https://huggingface.co/deepseek-ai/deepseek-moe-16b-chat>`_
-
-   * - ``$model_repo``
-     - ``amd/Llama-3.1-70B-Instruct-FP8-KV``
-     - `Llama 3.1 70B FP8 <https://huggingface.co/amd/Llama-3.1-70B-Instruct-FP8-KV>`_
-
-   * - (``float8``)
-     - ``amd/Llama-3.1-405B-Instruct-FP8-KV``
-     - `Llama 3.1 405B FP8 <https://huggingface.co/amd/Llama-3.1-405B-Instruct-FP8-KV>`_
-
-   * -
-     - ``amd/Mixtral-8x7B-Instruct-v0.1-FP8-KV``
-     - `Mixtral MoE 8x7B FP8 <https://huggingface.co/amd/Mixtral-8x7B-Instruct-v0.1-FP8-KV>`_
-
-   * -
-     - ``amd/Mixtral-8x22B-Instruct-v0.1-FP8-KV``
-     - `Mixtral MoE 8x22B FP8 <https://huggingface.co/amd/Mixtral-8x22B-Instruct-v0.1-FP8-KV>`_
-
-   * -
-     - ``amd/Mistral-7B-v0.1-FP8-KV``
-     - `Mistral 7B FP8 <https://huggingface.co/amd/Mistral-7B-v0.1-FP8-KV>`_
-
-   * -
-     - ``amd/dbrx-instruct-FP8-KV``
-     - `DBRX Instruct FP8 <https://huggingface.co/amd/dbrx-instruct-FP8-KV>`_
-
-   * -
-     - ``amd/c4ai-command-r-plus-FP8-KV``
-     - `C4AI Command R+ 08-2024 FP8 <https://huggingface.co/amd/c4ai-command-r-plus-FP8-KV>`_
-
-   * - ``$num_gpu``
-     - 1 or 8
-     - Number of GPUs
-
-   * - ``$datatype``
-     - ``float16`` or ``float8``
-     - Data type
-
-.. _vllm-benchmark-run-benchmark:
-
-Running the benchmark on the MI300X accelerator
------------------------------------------------
-
-Here are some examples of running the benchmark with various options.
-See :ref:`Options <vllm-benchmark-standalone-options>` for the list of
-options and their descriptions.
-
-Example 1: latency benchmark
-^^^^^^^^^^^^^^^^^^^^^^^^^^^^
- 
-Use this command to benchmark the latency of the Llama 3.1 70B model on eight GPUs with the ``float16`` and ``float8`` data types.
-
-.. code-block::
-
-   ./vllm_benchmark_report.sh -s latency -m meta-llama/Llama-3.1-70B-Instruct -g 8 -d float16
-   ./vllm_benchmark_report.sh -s latency -m amd/Llama-3.1-70B-Instruct-FP8-KV -g 8 -d float8
-
-Find the latency reports at:
-
-- ``./reports_float16/summary/Llama-3.1-70B-Instruct_latency_report.csv``
-
-- ``./reports_float8/summary/Llama-3.1-70B-Instruct-FP8-KV_latency_report.csv``
-
-Example 2: throughput benchmark
-^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
-
-Use this command to benchmark the throughput of the Llama 3.1 70B model on eight GPUs with the ``float16`` and ``float8`` data types.
-
-.. code-block:: shell
-
-   ./vllm_benchmark_report.sh -s throughput -m meta-llama/Llama-3.1-70B-Instruct -g 8 -d float16
-   ./vllm_benchmark_report.sh -s throughput -m amd/Llama-3.1-70B-Instruct-FP8-KV -g 8 -d float8
-
-Find the throughput reports at:
-
-- ``./reports_float16/summary/Llama-3.1-70B-Instruct_throughput_report.csv``
-
-- ``./reports_float8/summary/Llama-3.1-70B-Instruct-FP8-KV_throughput_report.csv``
-
-.. raw:: html
-
-   <style>
-   mjx-container[jax="CHTML"][display="true"] {
-       text-align: left;
-       margin: 0;
-   }
-   </style>
-
-.. note::
-
-   Throughput is calculated as:
-
-   - .. math:: throughput\_tot = requests \times (\mathsf{\text{input lengths}} + \mathsf{\text{output lengths}}) / elapsed\_time
-
-   - .. math:: throughput\_gen = requests \times \mathsf{\text{output lengths}} / elapsed\_time
+               - .. math:: throughput\_gen = requests \times \mathsf{\text{output lengths}} / elapsed\_time
+      {% endfor %}
+   {% endfor %}
 
 Further reading
 ===============
@@ -446,33 +290,3 @@ Further reading
 
 - To learn how to fine-tune LLMs, see
   :doc:`Fine-tuning LLMs <../fine-tuning/index>`.
-
-Previous versions
-=================
-
-This table lists previous versions of the ROCm vLLM Docker image for inference
-performance validation. For detailed information about available models for
-benchmarking, see the version-specific documentation.
-
-.. list-table::
-   :header-rows: 1
-   :stub-columns: 1
-
-   * - ROCm version
-     - vLLM version
-     - PyTorch version
-     - Resources
-
-   * - 6.2.1
-     - 0.6.4
-     - 2.5.0
-     - 
-       * `Documentation <https://rocm.docs.amd.com/en/docs-6.3.0/how-to/performance-validation/mi300x/vllm-benchmark.html>`_
-       * `Docker Hub <https://hub.docker.com/layers/rocm/vllm/rocm6.2_mi300_ubuntu20.04_py3.9_vllm_0.6.4/images/sha256-ccbb74cc9e7adecb8f7bdab9555f7ac6fc73adb580836c2a35ca96ff471890d8>`_
-
-   * - 6.2.0
-     - 0.4.3
-     - 2.4.0
-     -
-       * `Documentation <https://rocm.docs.amd.com/en/docs-6.2.0/how-to/performance-validation/mi300x/vllm-benchmark.html>`_
-       * `Docker Hub <https://hub.docker.com/layers/rocm/vllm/rocm6.2_mi300_ubuntu22.04_py3.9_vllm_7c5fd50/images/sha256-9e4dd4788a794c3d346d7d0ba452ae5e92d39b8dfac438b2af8efdc7f15d22c0>`_

docs/how-to/rocm-for-ai/training/benchmark-docker/megatron-lm.rst

@@ -0,0 +1,547 @@
+:orphan:
+
+.. meta::
+   :description: How to train a model using Megatron-LM for ROCm.
+   :keywords: ROCm, AI, LLM, train, Megatron-LM, megatron, Llama, tutorial, docker, torch
+
+******************************************
+Training a model with Megatron-LM for ROCm
+******************************************
+
+The Megatron-LM framework for ROCm is a specialized fork of the robust Megatron-LM,
+designed to enable efficient training of large-scale language models on AMD
+GPUs. By leveraging AMD Instinct™ MI300X series accelerators, Megatron-LM delivers
+enhanced scalability, performance, and resource utilization for AI workloads.
+It is purpose-built to support models like Llama 2, Llama 3, Llama 3.1, and
+DeepSeek, enabling developers to train next-generation AI models more
+efficiently. See the GitHub repository at `<https://github.com/ROCm/Megatron-LM>`__.
+
+AMD provides a ready-to-use Docker image for MI300X accelerators containing
+essential components, including PyTorch, ROCm libraries, and Megatron-LM
+utilities. It contains the following software components to accelerate training
+workloads:
+
++--------------------------+--------------------------------+
+| Software component       | Version                        |
++==========================+================================+
+| ROCm                     | 6.3.0                          |
++--------------------------+--------------------------------+
+| PyTorch                  | 2.7.0a0+git637433              |
++--------------------------+--------------------------------+
+| Python                   | 3.10                           |
++--------------------------+--------------------------------+
+| Transformer Engine       | 1.11                           |
++--------------------------+--------------------------------+
+| Flash Attention          | 3.0.0                          |
++--------------------------+--------------------------------+
+| hipBLASLt                | git258a2162                    |
++--------------------------+--------------------------------+
+| Triton                   | 3.1                            |
++--------------------------+--------------------------------+
+
+Supported features and models
+=============================
+
+Megatron-LM provides the following key features to train large language models efficiently:
+
+- Transformer Engine (TE)
+
+- APEX
+
+- GEMM tuning
+
+- Torch.compile
+
+- 3D parallelism: TP + SP + CP
+
+- Distributed optimizer
+
+- Flash Attention (FA) 3
+
+- Fused kernels
+
+- Pre-training
+
+.. _amd-megatron-lm-model-support:
+
+The following models are pre-optimized for performance on the AMD Instinct MI300X accelerator.
+
+* Llama 2 7B
+
+* Llama 2 70B
+
+* Llama 3 8B
+
+* Llama 3 70B
+
+* Llama 3.1 8B
+
+* Llama 3.1 70B
+
+* DeepSeek-V2-Lite
+
+.. note::
+
+   Some models, such as Llama 3, require an external license agreement through
+   a third party (for example, Meta).
+
+System validation
+=================
+
+If you have already validated your system settings, skip this step. Otherwise,
+complete the :ref:`system validation and optimization steps <train-a-model-system-validation>`
+to set up your system before starting training.
+
+Disable NUMA auto-balancing
+---------------------------
+
+Generally, application performance can benefit from disabling NUMA auto-balancing. However,
+it might be detrimental to performance with certain types of workloads.
+
+Run the command ``cat /proc/sys/kernel/numa_balancing`` to check your current NUMA (Non-Uniform
+Memory Access) settings. Output ``0`` indicates this setting is disabled. If there is no output or
+the output is ``1``, run the following command to disable NUMA auto-balancing.
+
+.. code-block:: shell
+
+   sudo sh -c 'echo 0 > /proc/sys/kernel/numa_balancing'
+
+See :ref:`mi300x-disable-numa` for more information.
+
+.. _mi300x-amd-megatron-lm-training:
+
+Environment setup
+=================
+
+The pre-built ROCm Megatron-LM environment allows users to quickly validate system performance, conduct
+training benchmarks, and achieve superior performance for models like Llama 3.1, Llama 2, and DeepSeek V2.
+
+Use the following instructions to set up the environment, configure the script to train models, and
+reproduce the benchmark results on the MI300X accelerators with the AMD Megatron-LM Docker
+image.
+
+.. _amd-megatron-lm-requirements:
+ 
+Download the Docker image
+-------------------------
+
+1. Use the following command to pull the Docker image from Docker Hub.
+
+   .. code-block:: shell
+
+      docker pull rocm/megatron-lm:v25.3
+
+2. Launch the Docker container.
+
+   .. code-block:: shell
+
+      docker run -it --device /dev/dri --device /dev/kfd --network host --ipc host --group-add video --cap-add SYS_PTRACE --security-opt seccomp=unconfined --privileged -v $HOME:$HOME -v  $HOME/.ssh:/root/.ssh --shm-size 64G --name megatron_training_env rocm/megatron-lm:v25.3
+
+3. Use these commands if you exit the ``megatron_training_env`` container and need to return to it.
+
+   .. code-block:: shell
+
+      docker start megatron_training_env
+      docker exec -it megatron_training_env bash
+
+The Docker container includes a pre-installed, verified version of Megatron-LM from the `release branch <https://github.com/ROCm/Megatron-LM/tree/megatron_release_v25.3>`_.
+
+.. _amd-megatron-lm-environment-setup:
+
+Configuration scripts
+---------------------
+
+.. tab-set::
+
+   .. tab-item:: Llama
+      :sync: llama
+
+      If you're working with Llama 2 7B or Llama 2 70 B, use the ``train_llama2.sh`` configuration
+      script in the ``examples/llama`` directory of
+      `<https://github.com/ROCm/Megatron-LM/tree/megatron_release_v25.3/examples/llama>`__.
+      Likewise, if you're working with Llama 3 or Llama 3.1, then use ``train_llama3.sh`` and update
+      the configuration script accordingly.
+
+   .. tab-item:: DeepSeek V2
+      :sync: deepseek
+
+      Use the ``train_deepseek_v2.sh`` configuration script in the ``examples/deepseek_v2``
+      directory of
+      `<https://github.com/ROCm/Megatron-LM/tree/megatron_release_v25.3/examples/deepseek_v2>`__
+      and update the configuration script accordingly.
+
+Network interface
+^^^^^^^^^^^^^^^^^
+
+.. tab-set::
+
+   .. tab-item:: Llama
+      :sync: llama
+
+      To avoid connectivity issues in multi-node deployments, ensure the correct network interface
+      is set in your training scripts.
+
+      1. Run the following command (outside the container) to find the active network interface on your system.
+
+         .. code-block:: shell
+
+            ip a
+
+      2. Update the ``NCCL_SOCKET_IFNAME`` and ``GLOO_SOCKET_IFNAME`` variables with your system’s network interface. For
+         example:
+
+         .. code-block:: shell
+
+            export NCCL_SOCKET_IFNAME=ens50f0np0
+
+            export GLOO_SOCKET_IFNAME=ens50f0np0
+
+Dataset options
+^^^^^^^^^^^^^^^
+
+.. tab-set::
+
+   .. tab-item:: Llama
+      :sync: llama
+
+      You can use either mock data or real data for training.
+
+      * Mock data can be useful for testing and validation. Use the ``MOCK_DATA`` variable to toggle between mock and real data. The default
+        value is ``1`` for enabled.
+
+        .. code-block:: bash
+
+           MOCK_DATA=1
+
+      * If you're using a real dataset, update the ``DATA_PATH`` variable to point to the location of your dataset.
+
+        .. code-block:: bash
+
+           MOCK_DATA=0
+
+           DATA_PATH=${DATA_PATH:-"/data/bookcorpus_text_sentence"}  # Change to where your dataset is stored
+
+        Ensure that the files are accessible inside the Docker container.
+
+   .. tab-item:: DeepSeek V2
+      :sync: deepseek
+
+      If you don't already have the dataset, download the DeepSeek dataset using the following
+      commands:
+
+      .. code-block:: shell
+
+         mkdir deepseek-datasets
+         cd deepseek-datasets
+         wget https://atp-modelzoo-wlcb-pai.oss-cn-wulanchabu.aliyuncs.com/release/models/pai-megatron-patch/deepseek-datasets/SlimPajama.json
+         wget https://atp-modelzoo-wlcb-pai.oss-cn-wulanchabu.aliyuncs.com/release/models/pai-megatron-patch/deepseek-datasets/alpaca_zh-train.json
+         wget https://atp-modelzoo-wlcb-pai.oss-cn-wulanchabu.aliyuncs.com/release/models/pai-megatron-patch/deepseek-datasets/alpaca_zh-valid.json
+         wget https://atp-modelzoo-wlcb-pai.oss-cn-wulanchabu.aliyuncs.com/release/models/pai-megatron-patch/deepseek-datasets/mmap_deepseekv2_datasets_text_document.bin
+         wget https://atp-modelzoo-wlcb-pai.oss-cn-wulanchabu.aliyuncs.com/release/models/pai-megatron-patch/deepseek-datasets/mmap_deepseekv2_datasets_text_document.idx
+
+      You can use either mock data or real data for training.
+
+      * Mock data can be useful for testing and validation. Use the ``MOCK_DATA`` variable to toggle between mock and real data. The default
+        value is ``1`` for enabled.
+
+        .. code-block:: bash
+
+           MOCK_DATA=1
+
+      * If you're using a real dataset, update the ``DATA_DIR`` variable to point to the location of your dataset.
+
+        .. code-block:: bash
+
+           MOCK_DATA=0
+
+           DATA_DIR="/root/data/deepseek-datasets"  # Change to where your dataset is stored
+
+        Ensure that the files are accessible inside the Docker container.
+
+Tokenizer
+^^^^^^^^^
+
+Tokenization is the process of converting raw text into tokens that can be processed by the model. For Llama
+models, this typically involves sub-word tokenization, where words are broken down into smaller units based on
+a fixed vocabulary. The tokenizer is trained along with the model on a large corpus of text, and it learns a
+fixed vocabulary that can represent a wide range of text from different domains. This allows Llama models to
+handle a variety of input sequences, including unseen words or domain-specific terms.
+
+.. tab-set::
+
+   .. tab-item:: Llama
+      :sync: llama
+
+      To train any of the Llama 2 models that :ref:`this Docker image supports <amd-megatron-lm-model-support>`, use the ``Llama2Tokenizer``.
+
+      To train any of Llama 3 and Llama 3.1 models that this Docker image supports, use the ``HuggingFaceTokenizer``.
+      Set the Hugging Face model link in the ``TOKENIZER_MODEL`` variable.
+
+      For example, if you're using the Llama 3.1 8B model:
+
+      .. code-block:: shell
+
+         TOKENIZER_MODEL=meta-llama/Llama-3.1-8B
+
+   .. tab-item:: DeepSeek V2
+      :sync: deepseek
+
+      To train any of the DeepSeek V2 models that :ref:`this Docker image supports <amd-megatron-lm-model-support>`, use the ``DeepSeekV2Tokenizer``.
+
+Multi-node training
+^^^^^^^^^^^^^^^^^^^
+
+.. tab-set::
+
+   .. tab-item:: Llama
+      :sync: llama
+
+      If you're running multi-node training, update the following environment variables. They can
+      also be passed as command line arguments.
+
+      * Change ``localhost`` to the master node's hostname:
+
+        .. code-block:: shell
+
+           MASTER_ADDR="${MASTER_ADDR:-localhost}"
+
+      * Set the number of nodes you want to train on (for instance, ``2``, ``4``, ``8``):
+
+        .. code-block:: shell
+
+           NNODES="${NNODES:-1}"
+
+      * Set the rank of each node (0 for master, 1 for the first worker node, and so on):
+
+        .. code-block:: shell
+
+           NODE_RANK="${NODE_RANK:-0}"
+
+      * Set ``DATA_CACHE_PATH`` to a common directory accessible by all the nodes (for example, an
+        NFS directory) for multi-node runs:
+
+        .. code-block:: shell
+
+           DATA_CACHE_PATH=/root/cache # Set to a common directory for multi-node runs
+
+      * For multi-node runs, make sure the correct network drivers are installed on the nodes. If
+        inside a Docker, either install the drivers inside the Docker container or pass the network
+        drivers from the host while creating the Docker container.
+
+Start training on AMD Instinct accelerators
+===========================================
+
+The prebuilt Megatron-LM with ROCm training environment allows users to quickly validate
+system performance, conduct training benchmarks, and achieve superior
+performance for models like Llama 3.1 and Llama 2. This container should not be
+expected to provide generalized performance across all training workloads. You
+can expect the container to perform in the model configurations described in
+the following section, but other configurations are not validated by AMD.
+
+Use the following instructions to set up the environment, configure the script
+to train models, and reproduce the benchmark results on MI300X series
+accelerators with the AMD Megatron-LM Docker image.
+
+.. tab-set::
+
+   .. tab-item:: Llama
+      :sync: llama
+
+      .. tab-set::
+
+         .. tab-item:: Single node training
+            :sync: single-node
+
+            To run training on a single node, navigate to the Megatron-LM folder and use the
+            following command:
+
+            .. code-block:: shell
+
+               TEE_OUTPUT=1 MBS=2 BS=128 TP=1 TE_FP8=1 SEQ_LENGTH=8192 MODEL_SIZE=8 bash examples/llama/train_llama3.sh
+
+         .. tab-item:: Multi-node training
+            :sync: multi-node
+
+            To run training on multiple nodes, launch the Docker container on each node. For example, for a two node setup (``NODE0`` as the master node), use these commands.
+
+            * On the master node ``NODE0``:
+
+              .. code-block:: shell
+
+                 TEE_OUTPUT=1 MBS=2 BS=256 TP=1 TE_FP8=1 SEQ_LENGTH=8192 MODEL_SIZE=8 MASTER_ADDR=IP_NODE0 NNODES=2 NODE_RANK=0 bash examples/llama/train_llama3.sh
+
+            * On the worker node ``NODE1``:
+
+              .. code-block:: shell
+
+                 TEE_OUTPUT=1 MBS=2 BS=256 TP=1 TE_FP8=1 SEQ_LENGTH=8192 MODEL_SIZE=8 MASTER_ADDR=IP_NODE0 NNODES=2 NODE_RANK=1 bash examples/llama/train_llama3.sh
+
+
+   .. tab-item:: DeepSeek V2
+      :sync: deepseek
+
+      To run the training on a single node, go to ``/Megatron-LM`` folder and use the following command:
+
+      .. code-block:: shell
+
+         cd /workspace/Megatron-LM
+         GEMM_TUNING=1 PR=bf16 MBS=4 AC=none bash examples/deepseek_v2/train_deepseekv2.sh
+
+Key options
+-----------
+
+.. _amd-megatron-lm-benchmark-test-vars:
+
+The benchmark tests support the following sets of variables:
+
+.. tab-set::
+
+   .. tab-item:: Llama
+      :sync: llama
+
+      ``TEE_OUTPUT``
+        ``1`` to enable training logs or ``0`` to disable.
+
+      ``TE_FP8``
+        ``0`` for BP16 (default) or ``1`` for FP8 GEMMs.
+
+      ``GEMM_TUNING``
+        ``1`` to enable GEMM tuning, which boosts performance by using the best GEMM kernels.
+
+      ``USE_FLASH_ATTN``
+        ``1`` to enable Flash Attention.
+
+      ``ENABLE_PROFILING``
+        ``1`` to enable PyTorch profiling for performance analysis.
+
+      ``transformer-impl``
+        ``transformer_engine`` to use the Transformer Engine (TE) or ``local`` to disable TE.
+
+      ``MODEL_SIZE``
+        ``8B`` or ``70B`` for Llama 3 and 3.1. ``7B`` or ``70B`` for Llama 2.
+
+      ``TOTAL_ITERS``
+        The total number of iterations -- ``10`` by default.
+
+      ``MOCK_DATA``
+        ``1`` to use mock data or ``0`` to use real data provided by you.
+
+      ``MBS``
+        Micro batch size.
+
+      ``BS``
+        Global batch size.
+
+      ``TP``
+        Tensor parallel (``1``, ``2``, ``4``, ``8``).
+
+      ``SEQ_LENGTH``
+        Input sequence length.
+
+   .. tab-item:: DeepSeek V2
+      :sync: deepseek
+
+      ``PR``
+        Precision for training. ``bf16`` for BF16 (default) or ``fp8`` for FP8 GEMMs.
+
+      ``GEMM_TUNING``
+        ``1`` to enable GEMM tuning, which boosts performance by using the best GEMM kernels.
+
+      ``TOTAL_ITERS``
+        The total number of iterations -- ``10`` by default.
+
+      ``MOCK_DATA``
+        ``1`` to use mock data or ``0`` to use real data provided by you.
+
+      ``MBS``
+        Micro batch size.
+
+      ``GBS``
+        Global batch size.
+
+Benchmarking examples
+---------------------
+
+.. tab-set::
+
+   .. tab-item:: Llama
+      :sync: llama
+
+      .. tab-set::
+
+         .. tab-item:: Single node training
+            :sync: single-node
+
+            Use this command to run training with Llama 2 7B model on a single node. You can specify MBS, BS, FP,
+            datatype, and so on.
+
+            .. code-block:: bash
+
+               TEE_OUTPUT=1 MBS=5 BS=120 TP=8 TE_FP8=0 NO_TORCH_COMPILE=1
+               SEQ_LENGTH=4096 bash examples/llama/train_llama2.sh
+
+            You can find the training logs at the location defined in ``$TRAIN_LOG`` in the :ref:`configuration script <amd-megatron-lm-environment-setup>`.
+
+            See the sample output:
+
+            .. image:: ../../../../data/how-to/rocm-for-ai/llama2-7b-training-log-sample.png
+               :width: 800
+
+         .. tab-item:: Multi-node training
+            :sync: multi-node
+
+            Launch the Docker container on each node.
+
+            In this example, run training with Llama 2 7B model on 2 nodes with specific MBS, BS, FP, datatype, and
+            so on.
+
+            On the master node:
+
+            .. code-block:: bash
+
+               TEE_OUTPUT=1 MBS=4 BS=64 TP=8 TE_FP8=0 NO_TORCH_COMPILE=1
+               SEQ_LENGTH=4096 bash examples/llama/train_llama2.sh
+
+            On the worker node:
+
+            .. code-block:: bash
+
+               TEE_OUTPUT=1 MBS=4 BS=64 TP=8 TE_FP8=0 NO_TORCH_COMPILE=1
+               SEQ_LENGTH=4096 bash examples/llama/train_llama2.sh
+
+            You can find the training logs at the location defined in ``$TRAIN_LOG`` in the :ref:`configuration script <amd-megatron-lm-environment-setup>`.
+
+            Sample output for 2-node training:
+
+            Master node:
+
+            .. image:: ../../../../data/how-to/rocm-for-ai/2-node-training-master.png
+               :width: 800
+
+            Worker node:
+
+            .. image:: ../../../../data/how-to/rocm-for-ai/2-node-training-worker.png
+               :width: 800
+
+Previous versions
+=================
+
+This table lists previous versions of the ROCm Megatron-LM Docker image for training
+performance validation. For detailed information about available models for
+benchmarking, see the version-specific documentation.
+
+.. list-table::
+   :header-rows: 1
+   :stub-columns: 1
+
+   * - ROCm version
+     - Megatron-LM version
+     - PyTorch version
+     - Resources
+
+   * - 6.1
+     - 24.12-dev
+     - 2.4.0
+     - 
+       * `Documentation <https://rocm.docs.amd.com/en/docs-6.3.0/how-to/rocm-for-ai/train-a-model.html>`_
+       * `Docker Hub <https://hub.docker.com/layers/rocm/megatron-lm/24.12-dev/images/sha256-5818c50334ce3d69deeeb8f589d83ec29003817da34158ebc9e2d112b929bf2e>`_

docs/how-to/rocm-for-ai/training/benchmark-docker/pytorch-training.rst

@@ -0,0 +1,341 @@
+:orphan:
+
+.. meta::
+   :description: How to train a model using PyTorch for ROCm.
+   :keywords: ROCm, AI, LLM, train, PyTorch, torch, Llama, flux, tutorial, docker
+
+**************************************
+Training a model with PyTorch for ROCm
+**************************************
+
+PyTorch is an open-source machine learning framework that is widely used for
+model training with GPU-optimized components for transformer-based models.
+
+The PyTorch for ROCm training Docker (``rocm/pytorch-training:v25.3``) image
+provides a prebuilt optimized environment for fine-tuning and pretraining a
+model on AMD Instinct MI325X and MI300X accelerators. It includes the following
+software components to accelerate training workloads:
+
++--------------------------+--------------------------------+
+| Software component       | Version                        |
++==========================+================================+
+| ROCm                     | 6.3.0                          |
++--------------------------+--------------------------------+
+| PyTorch                  | 2.7.0a0+git637433              |
++--------------------------+--------------------------------+
+| Python                   | 3.10                           |
++--------------------------+--------------------------------+
+| Transformer Engine       | 1.11                           |
++--------------------------+--------------------------------+
+| Flash Attention          | 3.0.0                          |
++--------------------------+--------------------------------+
+| hipBLASLt                | git258a2162                    |
++--------------------------+--------------------------------+
+| Triton                   | 3.1                            |
++--------------------------+--------------------------------+
+
+.. _amd-pytorch-training-model-support:
+
+Supported models
+================
+
+The following models are pre-optimized for performance on the AMD Instinct MI300X accelerator.
+
+* Llama 3.1 8B
+
+* Llama 3.1 70B
+
+* FLUX.1-dev
+
+.. note::
+
+   Only these models are supported in the following steps.
+
+   Some models, such as Llama 3, require an external license agreement through
+   a third party (for example, Meta).
+
+System validation
+=================
+
+If you have already validated your system settings, skip this step. Otherwise,
+complete the :ref:`system validation and optimization steps <train-a-model-system-validation>`
+to set up your system before starting training.
+
+Disable NUMA auto-balancing
+---------------------------
+
+Generally, application performance can benefit from disabling NUMA auto-balancing. However,
+it might be detrimental to performance with certain types of workloads.
+
+Run the command ``cat /proc/sys/kernel/numa_balancing`` to check your current NUMA (Non-Uniform
+Memory Access) settings. Output ``0`` indicates this setting is disabled. If there is no output or
+the output is ``1``, run the following command to disable NUMA auto-balancing.
+
+.. code-block:: shell
+
+   sudo sh -c 'echo 0 > /proc/sys/kernel/numa_balancing'
+
+See :ref:`mi300x-disable-numa` for more information.
+
+Environment setup
+=================
+
+This Docker image is optimized for specific model configurations outlined
+below. Performance can vary for other training workloads, as AMD 
+doesn’t validate configurations and run conditions outside those described.
+
+Download the Docker image
+-------------------------
+
+1. Use the following command to pull the Docker image from Docker Hub.
+
+   .. code-block:: shell
+
+      docker pull rocm/pytorch-training:v25.3
+
+2. Run the Docker container.
+
+   .. code-block:: shell
+
+      docker run -it --device /dev/dri --device /dev/kfd --network host --ipc host --group-add video --cap-add SYS_PTRACE --security-opt seccomp=unconfined --privileged -v $HOME:$HOME -v  $HOME/.ssh:/root/.ssh --shm-size 64G --name training_env rocm/pytorch-training:v25.3
+
+3. Use these commands if you exit the ``training_env`` container and need to return to it.
+
+   .. code-block:: shell
+
+      docker start training_env
+      docker exec -it training_env bash
+
+4. In the Docker container, clone the `<https://github.com/ROCm/MAD>`__ repository and navigate to the benchmark scripts directory.
+
+   .. code-block:: shell
+
+      git clone https://github.com/ROCm/MAD
+      cd MAD/scripts/pytorch-train
+
+Prepare training datasets and dependencies
+------------------------------------------
+
+The following benchmarking examples may require downloading models and datasets
+from Hugging Face. To ensure successful access to gated repos, set your
+``HF_TOKEN``.
+
+Run the setup script to install libraries and datasets needed for benchmarking.
+
+.. code-block:: shell
+
+   ./pytorch_benchmark_setup.sh
+
+``pytorch_benchmark_setup.sh`` installs the following libraries:
+
+.. list-table::
+   :header-rows: 1
+
+   * - Library
+     - Benchmark model
+     - Reference
+
+   * - ``accelerate``
+     - Llama 3.1 8B, FLUX
+     - `Hugging Face Accelerate <https://huggingface.co/docs/accelerate/en/index>`_
+
+   * - ``datasets``
+     - Llama 3.1 8B, 70B, FLUX
+     - `Hugging Face Datasets <https://huggingface.co/docs/datasets/v3.2.0/en/index>`_ 3.2.0
+
+   * - ``torchdata``
+     - Llama 3.1 70B
+     - `TorchData <https://pytorch.org/data/beta/index.html>`_
+
+   * - ``tomli``
+     - Llama 3.1 70B
+     - `Tomli <https://pypi.org/project/tomli/>`_
+
+   * - ``tiktoken``
+     - Llama 3.1 70B
+     - `tiktoken <https://github.com/openai/tiktoken>`_
+
+   * - ``blobfile``
+     - Llama 3.1 70B
+     - `blobfile <https://pypi.org/project/blobfile/>`_
+
+   * - ``tabulate``
+     - Llama 3.1 70B
+     - `tabulate <https://pypi.org/project/tabulate/>`_
+
+   * - ``wandb``
+     - Llama 3.1 70B
+     - `Weights & Biases <https://github.com/wandb/wandb>`_
+
+   * - ``sentencepiece``
+     - Llama 3.1 70B, FLUX
+     - `SentencePiece <https://github.com/google/sentencepiece>`_ 0.2.0
+
+   * - ``tensorboard``
+     - Llama 3.1 70 B, FLUX
+     - `TensorBoard <https://www.tensorflow.org/tensorboard>`_ 2.18.0
+
+   * - ``csvkit``
+     - FLUX
+     - `csvkit <https://csvkit.readthedocs.io/en/latest/>`_ 2.0.1
+
+   * - ``deepspeed``
+     - FLUX
+     - `DeepSpeed <https://github.com/deepspeedai/DeepSpeed>`_ 0.16.2
+
+   * - ``diffusers``
+     - FLUX
+     - `Hugging Face Diffusers <https://huggingface.co/docs/diffusers/en/index>`_ 0.31.0
+
+   * - ``GitPython``
+     - FLUX
+     - `GitPython <https://github.com/gitpython-developers/GitPython>`_ 3.1.44
+
+   * - ``opencv-python-headless``
+     - FLUX
+     - `opencv-python-headless <https://pypi.org/project/opencv-python-headless/>`_ 4.10.0.84
+
+   * - ``peft``
+     - FLUX
+     - `PEFT <https://huggingface.co/docs/peft/en/index>`_ 0.14.0
+
+   * - ``protobuf``
+     - FLUX
+     - `Protocol Buffers <https://github.com/protocolbuffers/protobuf>`_ 5.29.2
+
+   * - ``pytest``
+     - FLUX
+     - `PyTest <https://docs.pytest.org/en/stable/>`_ 8.3.4
+
+   * - ``python-dotenv``
+     - FLUX
+     - `python-dotenv <https://pypi.org/project/python-dotenv/>`_ 1.0.1
+
+   * - ``seaborn``
+     - FLUX
+     - `Seaborn <https://seaborn.pydata.org/>`_ 0.13.2
+
+   * - ``transformers``
+     - FLUX
+     - `Transformers <https://huggingface.co/docs/transformers/en/index>`_ 4.47.0
+
+``pytorch_benchmark_setup.sh`` downloads the following models from Hugging Face:
+
+* `meta-llama/Llama-3.1-70B-Instruct <https://huggingface.co/meta-llama/Llama-3.1-70B-Instruct>`_
+
+* `black-forest-labs/FLUX.1-dev <https://huggingface.co/black-forest-labs/FLUX.1-dev>`_
+
+Along with the following datasets:
+
+* `WikiText <https://huggingface.co/datasets/Salesforce/wikitext>`_
+
+* `bghira/pseudo-camera-10k <https://huggingface.co/datasets/bghira/pseudo-camera-10k>`_
+
+Start training on AMD Instinct accelerators
+===========================================
+
+The prebuilt PyTorch with ROCm training environment allows users to quickly validate
+system performance, conduct training benchmarks, and achieve superior
+performance for models like Llama 3.1 and Llama 2. This container should not be
+expected to provide generalized performance across all training workloads. You
+can expect the container to perform in the model configurations described in
+the following section, but other configurations are not validated by AMD.
+
+Use the following instructions to set up the environment, configure the script
+to train models, and reproduce the benchmark results on MI300X series
+accelerators with the AMD PyTorch training Docker image.
+
+Once your environment is set up, use the following commands and examples to start benchmarking.
+
+Pretraining
+-----------
+
+To start the pretraining benchmark, use the following command with the
+appropriate options. See the following list of options and their descriptions.
+
+.. code-block:: shell
+
+   ./pytorch_benchmark_report.sh -t $training_mode -m $model_repo -p $datatype -s $sequence_length
+
+Options and available models
+^^^^^^^^^^^^^^^^^^^^^^^^^^^^
+
+.. list-table::
+   :header-rows: 1
+
+   * - Name
+     - Options
+     - Description
+
+   * - ``$training_mode``
+     - ``pretrain``
+     - Benchmark pretraining
+
+   * -
+     - ``finetune_fw``
+     - Benchmark full weight fine-tuning (Llama 3.1 70B with BF16)
+
+   * -
+     - ``finetune_lora``
+     - Benchmark LoRA fine-tuning (Llama 3.1 70B with BF16)
+
+   * - ``$datatype``
+     - FP8 or BF16
+     - Only Llama 3.1 8B supports FP8 precision.
+
+   * - ``$model_repo``
+     - Llama-3.1-8B
+     - `Llama 3.1 8B <https://huggingface.co/meta-llama/Llama-3.1-8B-Instruct>`_
+
+   * - 
+     - Llama-3.1-70B
+     - `Llama 3.1 70B <https://huggingface.co/meta-llama/Llama-3.1-70B-Instruct>`_
+
+   * - 
+     - Flux
+     - `FLUX.1 [dev] <https://huggingface.co/black-forest-labs/FLUX.1-dev>`_
+
+Fine-tuning
+-----------
+
+To start the fine-tuning benchmark, use the following command. It will run the benchmarking example of Llama 2 70B
+with the WikiText dataset using the AMD fork of `torchtune <https://github.com/AMD-AIG-AIMA/torchtune>`_.
+
+.. code-block:: shell
+
+   ./pytorch_benchmark_report.sh -t {finetune_fw, finetune_lora} -p BF16 -m Llama-3.1-70B
+
+Benchmarking examples
+---------------------
+
+Here are some examples of how to use the command.
+
+* Example 1: Llama 3.1 70B with BF16 precision with `torchtitan <https://github.com/ROCm/torchtitan>`_.
+
+  .. code-block:: shell
+
+     ./pytorch_benchmark_report.sh -t pretrain -p BF16 -m Llama-3.1-70B -s 8192
+
+* Example 2: Llama 3.1 8B with FP8 precision using Transformer Engine (TE) and Hugging Face Accelerator.
+
+  .. code-block:: shell
+
+     ./pytorch_benchmark_report.sh -t pretrain -p FP8 -m Llama-3.1-70B -s 8192
+
+* Example 3: FLUX.1-dev with BF16 precision with FluxBenchmark.
+
+  .. code-block:: shell
+
+     ./pytorch_benchmark_report.sh -t pretrain -p BF16 -m Flux
+
+* Example 4: Torchtune full weight fine-tuning with Llama 3.1 70B
+
+  .. code-block:: shell
+
+     ./pytorch_benchmark_report.sh -t finetune_fw -p BF16 -m Llama-3.1-70B
+
+* Example 5: Torchtune LoRA fine-tuning with Llama 3.1 70B
+
+  .. code-block:: shell
+
+     ./pytorch_benchmark_report.sh -t finetune_lora -p BF16 -m Llama-3.1-70B

docs/how-to/rocm-for-ai/training/index.rst

@@ -19,6 +19,10 @@ training, fine-tuning, and inference. It leverages popular machine learning fram
 
 In this guide, you'll learn about:
 
-- :doc:`Training a model <train-a-model>`
+- Training a model
 
-- :doc:`Scale model training <scale-model-training>`
+  - :doc:`Train a model with Megatron-LM <benchmark-docker/megatron-lm>`
+
+  - :doc:`Train a model with PyTorch <benchmark-docker/pytorch-training>`
+
+- :doc:`Scaling model training <scale-model-training>`

docs/how-to/rocm-for-ai/training/prerequisite-system-validation.rst

@@ -0,0 +1,130 @@
+:orphan:
+
+.. meta::
+   :description: Prerequisite system validation before using ROCm for AI.
+   :keywords: ROCm, AI, LLM, train, megatron, Llama, tutorial, docker, torch, pytorch, jax
+
+.. _train-a-model-system-validation:
+
+**********************************************
+Prerequisite system validation before training
+**********************************************
+
+Complete the following system validation and optimization steps to set up your system before starting training.
+
+Disable NUMA auto-balancing
+---------------------------
+
+Generally, application performance can benefit from disabling NUMA auto-balancing. However,
+it might be detrimental to performance with certain types of workloads.
+
+Run the command ``cat /proc/sys/kernel/numa_balancing`` to check your current NUMA (Non-Uniform
+Memory Access) settings. Output ``0`` indicates this setting is disabled. If there is no output or
+the output is ``1``, run the following command to disable NUMA auto-balancing.
+
+.. code-block:: shell
+
+   sudo sh -c 'echo 0 > /proc/sys/kernel/numa_balancing'
+
+See :ref:`mi300x-disable-numa` for more information.
+
+Hardware verification with ROCm
+-------------------------------
+
+Use the command ``rocm-smi --setperfdeterminism 1900`` to set the max clock speed up to 1900 MHz
+instead of the default 2100 MHz. This can reduce the chance of a PCC event lowering the attainable
+GPU clocks. This setting will not be required for new IFWI releases with the production PRC feature.
+You can restore this setting to its default value with the ``rocm-smi -r`` command.
+
+Run the command:
+
+.. code-block:: shell
+
+   rocm-smi --setperfdeterminism 1900
+
+See :ref:`mi300x-hardware-verification-with-rocm` for more information.
+
+RCCL Bandwidth Test for multi-node setups
+-----------------------------------------
+
+ROCm Collective Communications Library (RCCL) is a standalone library of standard collective communication
+routines for GPUs. See the :doc:`RCCL documentation <rccl:index>` for more information. Before starting
+pretraining, running a RCCL bandwidth test helps ensure that the multi-GPU or multi-node setup is optimized
+for efficient distributed training.
+
+Running the RCCL bandwidth test helps verify that:
+
+- The GPUs can communicate across nodes or within a single node.
+
+- The interconnect (such as InfiniBand, Ethernet, or Infinite fabric) is functioning as expected and
+  provides adequate bandwidth for communication.
+
+- No hardware setup or cabling issues could affect the communication between GPUs
+
+Tuning and optimizing hyperparameters
+^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
+
+In distributed training, specific hyperparameters related to distributed communication can be tuned based on
+the results of the RCCL bandwidth test. These variables are already set in the Docker image:
+
+.. code-block:: shell
+
+   # force all RCCL streams to be high priority
+   export TORCH_NCCL_HIGH_PRIORITY=1
+
+   # specify which RDMA interfaces to use for communication
+   export NCCL_IB_HCA=rdma0,rdma1,rdma2,rdma3,rdma4,rdma5,rdma6,rdma7
+
+   # define the Global ID index used in RoCE mode
+   export NCCL_IB_GID_INDEX=3
+
+   # avoid data corruption/mismatch issue that existed in past releases
+   export RCCL_MSCCL_ENABLE=0
+
+Running the RCCL Bandwidth Test
+^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
+
+It's recommended you run the RCCL bandwidth test before launching training. It ensures system
+performance is sufficient to launch training. RCCL is not included in the AMD Megatron-LM Docker
+image; follow the instructions in `<https://github.com/ROCm/rccl-tests>`__ to get started.
+See :ref:`mi300x-rccl` for more information.
+
+Run on 8 GPUs (``-g 8``), scanning from 8 bytes to 10 GB:
+
+.. code-block:: shell
+
+   ./build/all_reduce_perf -b 8 -e 10G -f 2 -g 8
+
+.. image:: ../../../data/how-to/rocm-for-ai/rccl-tests-8-gpu.png
+   :width: 800
+
+Using one MPI process per GPU and ``-g 1`` for performance-oriented runs on both single-node and multi-node is
+recommended. So, a run on 8 GPUs looks something like:
+
+.. code-block:: shell
+
+   mpirun -np 8 --bind-to numa ./build/all_reduce_perf -b 8 -e 10G -f 2 -g 1
+
+.. image:: ../../../data/how-to/rocm-for-ai/rccl-tests-1-mpi-process-per-gpu.png
+   :width: 800
+
+Running with one MPI process per GPU ensures a one-to-one mapping for CPUs and GPUs, which can be beneficial
+for smaller message sizes. This better represents the real-world use of RCCL in deep learning frameworks like
+PyTorch and TensorFlow.
+
+Use the following script to run the RCCL test for four MI300X GPU nodes. Modify paths and node addresses as needed.
+
+.. code-block::
+
+   /home/$USER/ompi_for_gpu/ompi/bin/mpirun -np 32 -H tw022:8,tw024:8,tw010:8, tw015:8 \
+   --mca pml ucx \
+   --mca btl ^openib \
+   -x NCCL_SOCKET_IFNAME=ens50f0np0 \
+   -x NCCL_IB_HCA=rdma0:1,rdma1:1,rdma2:1,rdma3:1,rdma4:1,rdma5:1,rdma6:1,rdma7:1 \
+   -x NCCL_IB_GID_INDEX=3 \
+   -x NCCL_MIN_NCHANNELS=40 \
+   -x NCCL_DEBUG=version \
+   $HOME/rccl-tests/build/all_reduce_perf -b 8 -e 8g -f 2 -g 1
+
+.. image:: ../../../data/how-to/rocm-for-ai/rccl-tests-4-mi300x-gpu-nodes.png
+   :width: 800

docs/how-to/rocm-for-ai/training/train-a-model.rst

@@ -1,503 +0,0 @@
-.. meta::
-   :description: How to train a model using ROCm Megatron-LM
-   :keywords: ROCm, AI, LLM, train, Megatron-LM, megatron, Llama, tutorial, docker, torch
-
-**************************************
-Training a model with ROCm Megatron-LM
-**************************************
-
-.. _amd-megatron-lm:
-
-The ROCm Megatron-LM framework is a specialized fork of the robust Megatron-LM, designed to
-enable efficient training of large-scale language models on AMD GPUs. By leveraging AMD Instinct™ MI300X
-accelerators, AMD Megatron-LM delivers enhanced scalability, performance, and resource utilization for AI
-workloads. It is purpose-built to :ref:`support models <amd-megatron-lm-model-support>`
-like Meta's Llama 2, Llama 3, and Llama 3.1, enabling developers to train next-generation AI models with greater
-efficiency. See the GitHub repository at `<https://github.com/ROCm/Megatron-LM>`__.
-
-For ease of use, AMD provides a ready-to-use Docker image for MI300X accelerators containing essential
-components, including PyTorch, PyTorch Lightning, ROCm libraries, and Megatron-LM utilities. It contains the
-following software to accelerate training workloads:
-
-+--------------------------+--------------------------------+
-| Software component       | Version                        |
-+==========================+================================+
-| ROCm                     | 6.1                            |
-+--------------------------+--------------------------------+
-| PyTorch                  | 2.4.0                          |
-+--------------------------+--------------------------------+
-| PyTorch Lightning        | 2.4.0                          |
-+--------------------------+--------------------------------+
-| Megatron Core            | 0.9.0                          |
-+--------------------------+--------------------------------+
-| Transformer Engine       | 1.5.0                          |
-+--------------------------+--------------------------------+
-| Flash Attention          | v2.6                           |
-+--------------------------+--------------------------------+
-| Transformers             | 4.44.0                         |
-+--------------------------+--------------------------------+
-
-Supported features and models
-=============================
-
-Megatron-LM provides the following key features to train large language models efficiently:
-
-- Transformer Engine (TE)
-
-- APEX
-
-- GEMM tuning
-
-- Torch.compile
-
-- 3D parallelism: TP + SP + CP
-
-- Distributed optimizer
-
-- Flash Attention (FA) 2
-
-- Fused kernels
-
-- Pre-training
-
-.. _amd-megatron-lm-model-support:
-
-The following models are pre-optimized for performance on the AMD Instinct MI300X accelerator.
-
-* Llama 2 7B
-
-* Llama 2 70B
-
-* Llama 3 8B
-
-* Llama 3 70B
-
-* Llama 3.1 8B
-
-* Llama 3.1 70B
-
-Prerequisite system validation steps
-====================================
-
-Complete the following system validation and optimization steps to set up your system before starting training.
-
-Disable NUMA auto-balancing
----------------------------
-
-Generally, application performance can benefit from disabling NUMA auto-balancing. However,
-it might be detrimental to performance with certain types of workloads.
-
-Run the command ``cat /proc/sys/kernel/numa_balancing`` to check your current NUMA (Non-Uniform
-Memory Access) settings. Output ``0`` indicates this setting is disabled. If there is no output or
-the output is ``1``, run the following command to disable NUMA auto-balancing.
-
-.. code-block:: shell
-
-   sudo sh -c 'echo 0 > /proc/sys/kernel/numa_balancing'
-
-See :ref:`mi300x-disable-numa` for more information.
-
-Hardware verification with ROCm
--------------------------------
-
-Use the command ``rocm-smi --setperfdeterminism 1900`` to set the max clock speed up to 1900 MHz
-instead of the default 2100 MHz. This can reduce the chance of a PCC event lowering the attainable
-GPU clocks. This setting will not be required for new IFWI releases with the production PRC feature.
-You can restore this setting to its default value with the ``rocm-smi -r`` command.
-
-Run the command:
-
-.. code-block:: shell
-
-   rocm-smi --setperfdeterminism 1900
-
-See :ref:`mi300x-hardware-verification-with-rocm` for more information.
-
-RCCL Bandwidth Test
--------------------
-
-ROCm Collective Communications Library (RCCL) is a standalone library of standard collective communication
-routines for GPUs. See the :doc:`RCCL documentation <rccl:index>` for more information. Before starting
-pre-training, running a RCCL bandwidth test helps ensure that the multi-GPU or multi-node setup is optimized
-for efficient distributed training.
-
-Running the RCCL bandwidth test helps verify that:
-
-- The GPUs can communicate across nodes or within a single node.
-
-- The interconnect (such as InfiniBand, Ethernet, or Infinite fabric) is functioning as expected and
-  provides adequate bandwidth for communication.
-
-- No hardware setup or cabling issues could affect the communication between GPUs
-
-Tuning and optimizing hyperparameters
-^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
-
-In distributed training, specific hyperparameters related to distributed communication can be tuned based on
-the results of the RCCL bandwidth test. These variables are already set in the Docker image:
-
-.. code-block:: shell
-
-   # force all RCCL streams to be high priority
-   export TORCH_NCCL_HIGH_PRIORITY=1
-
-   # specify which RDMA interfaces to use for communication
-   export NCCL_IB_HCA=rdma0,rdma1,rdma2,rdma3,rdma4,rdma5,rdma6,rdma7
-
-   # define the Global ID index used in RoCE mode
-   export NCCL_IB_GID_INDEX=3
-
-   # avoid data corruption/mismatch issue that existed in past releases
-   export RCCL_MSCCL_ENABLE=0
-
-Running the RCCL Bandwidth Test
-^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
-
-It's recommended you run the RCCL bandwidth test before launching training. It ensures system
-performance is sufficient to launch training. RCCL is not included in the AMD Megatron-LM Docker
-image; follow the instructions in `<https://github.com/ROCm/rccl-tests>`__ to get started.
-See :ref:`mi300x-rccl` for more information.
-
-Run on 8 GPUs (``-g 8``), scanning from 8 bytes to 10 GB:
-
-.. code-block:: shell
-
-   ./build/all_reduce_perf -b 8 -e 10G -f 2 -g 8
-
-.. image:: ../../../data/how-to/rocm-for-ai/rccl-tests-8-gpu.png
-   :width: 800
-
-Using one MPI process per GPU and ``-g 1`` for performance-oriented runs on both single-node and multi-node is
-recommended. So, a run on 8 GPUs looks something like:
-
-.. code-block:: shell
-
-   mpirun -np 8 --bind-to numa ./build/all_reduce_perf -b 8 -e 10G -f 2 -g 1
-
-.. image:: ../../../data/how-to/rocm-for-ai/rccl-tests-1-mpi-process-per-gpu.png
-   :width: 800
-
-Running with one MPI process per GPU ensures a one-to-one mapping for CPUs and GPUs, which can be beneficial
-for smaller message sizes. This better represents the real-world use of RCCL in deep learning frameworks like
-PyTorch and TensorFlow.
-
-Use the following script to run the RCCL test for four MI300X GPU nodes. Modify paths and node addresses as needed.
-
-.. code-block::
-
-   /home/$USER/ompi_for_gpu/ompi/bin/mpirun -np 32 -H tw022:8,tw024:8,tw010:8, tw015:8 \
-   --mca pml ucx \
-   --mca btl ^openib \
-   -x NCCL_SOCKET_IFNAME=ens50f0np0 \
-   -x NCCL_IB_HCA=rdma0:1,rdma1:1,rdma2:1,rdma3:1,rdma4:1,rdma5:1,rdma6:1,rdma7:1 \
-   -x NCCL_IB_GID_INDEX=3 \
-   -x NCCL_MIN_NCHANNELS=40 \
-   -x NCCL_DEBUG=version \
-   $HOME/rccl-tests/build/all_reduce_perf -b 8 -e 8g -f 2 -g 1
-
-.. image:: ../../../data/how-to/rocm-for-ai/rccl-tests-4-mi300x-gpu-nodes.png
-   :width: 800
-
-.. _mi300x-amd-megatron-lm-training:
-
-Start training on MI300X accelerators
-=====================================
-
-The pre-built ROCm Megatron-LM environment allows users to quickly validate system performance, conduct
-training benchmarks, and achieve superior performance for models like Llama 2 and Llama 3.1.
-
-Use the following instructions to set up the environment, configure the script to train models, and
-reproduce the benchmark results on the MI300X accelerators with the AMD Megatron-LM Docker
-image.
-
-.. _amd-megatron-lm-requirements:
-
-Download the Docker image and required packages
------------------------------------------------
-
-1. Use the following command to pull the Docker image from Docker Hub.
-
-   .. code-block:: shell
-
-      docker pull rocm/megatron-lm:24.12-dev
-
-2. Launch the Docker container.
-
-   .. code-block:: shell
-
-      docker run -it --device /dev/dri --device /dev/kfd --network host --ipc host --group-add video --cap-add SYS_PTRACE --security-opt seccomp=unconfined --privileged -v $CACHE_DIR:/root/.cache --name megatron-dev-env rocm/megatron-lm:24.12-dev /bin/bash
-
-3. Clone the ROCm Megatron-LM repository to a local directory and install the required packages on the host machine.
-
-   .. code-block:: shell
-
-      git clone https://github.com/ROCm/Megatron-LM
-      cd Megatron-LM
-
-   .. note::
-
-      This release is validated with ``ROCm/Megatron-LM`` commit `bb93ccb <https://github.com/ROCm/Megatron-LM/tree/bb93ccbfeae6363c67b361a97a27c74ab86e7e92>`_.
-      Checking out this specific commit is recommended for a stable and reproducible environment.
-
-      .. code-block:: shell
-         
-         git checkout bb93ccbfeae6363c67b361a97a27c74ab86e7e92
-
-Prepare training datasets
--------------------------
-
-If you already have the preprocessed data, you can skip this section.
-
-Use the following command to process datasets. We use GPT data as an example. You may change the merge table, use an
-end-of-document token, remove sentence splitting, and use the tokenizer type.
-
-.. code-block:: shell
-
-   python tools/preprocess_data.py \
-       --input my-corpus.json \
-       --output-prefix my-gpt2 \
-       --vocab-file gpt2-vocab.json \
-       --tokenizer-type GPT2BPETokenizer \
-       --merge-file gpt2-merges.txt \
-       --append-eod
-
-In this case, the automatically generated output files are named ``my-gpt2_text_document.bin`` and
-``my-gpt2_text_document.idx``.
-
-.. image:: ../../../data/how-to/rocm-for-ai/prep-training-datasets-my-gpt2-text-document.png
-   :width: 800
-
-.. _amd-megatron-lm-environment-setup:
-
-Environment setup
------------------
-
-In the ``examples/llama`` directory of Megatron-LM, if you're working with Llama 2 7B or Llama 2 70 B, use the
-``train_llama2.sh`` configuration script. Likewise, if you're working with Llama 3 or Llama 3.1, then use
-``train_llama3.sh`` and update the configuration script accordingly.
-
-Network interface
-^^^^^^^^^^^^^^^^^
-
-To avoid connectivity issues, ensure the correct network interface is set in your training scripts.
-
-1. Run the following command to find the active network interface on your system.
-
-   .. code-block:: shell
-
-      ip a
-
-2. Update the ``NCCL_SOCKET_IFNAME`` and ``GLOO_SOCKET_IFNAME`` variables with your system’s network interface. For
-   example:
-
-   .. code-block:: shell
-
-      export NCCL_SOCKET_IFNAME=ens50f0np0
-
-      export GLOO_SOCKET_IFNAME=ens50f0np0
-
-Dataset options
-^^^^^^^^^^^^^^^
-
-You can use either mock data or real data for training.
-
-* If you're using a real dataset, update the ``DATA_PATH`` variable to point to the location of your dataset.
-
-  .. code-block:: shell
-
-     DATA_DIR="/root/.cache/data" # Change to where your dataset is stored
-
-     DATA_PATH=${DATA_DIR}/bookcorpus_text_sentence
-
-  .. code-block:: shell
-
-     --data-path $DATA_PATH
-
-  Ensure that the files are accessible inside the Docker container.
-
-* Mock data can be useful for testing and validation. If you're using mock data, replace ``--data-path $DATA_PATH`` with the ``--mock-data`` option.
-
-  .. code-block:: shell
-
-     --mock-data
-
-Tokenizer
-^^^^^^^^^
-
-Tokenization is the process of converting raw text into tokens that can be processed by the model. For Llama
-models, this typically involves sub-word tokenization, where words are broken down into smaller units based on
-a fixed vocabulary. The tokenizer is trained along with the model on a large corpus of text, and it learns a
-fixed vocabulary that can represent a wide range of text from different domains. This allows Llama models to
-handle a variety of input sequences, including unseen words or domain-specific terms.
-
-To train any of the Llama 2 models that this Docker image supports, use the ``Llama2Tokenizer``.
-
-To train any of Llama 3 and Llama 3.1 models that this Docker image supports, use the ``HuggingFaceTokenizer``.
-Set the Hugging Face model link in the ``TOKENIZER_MODEL`` variable.
-
-For example, if you're using the Llama 3.1 8B model:
-
-.. code-block:: shell
-
-   TOKENIZER_MODEL=meta-llama/Llama-3.1-8B
-
-Run benchmark tests
--------------------
-
-.. note::
-
-   If you're running **multi node training**, update the following environment variables. They can
-   also be passed as command line arguments.
-
-   * Change ``localhost`` to the master node's hostname:
-
-     .. code-block:: shell
-
-        MASTER_ADDR="${MASTER_ADDR:-localhost}"
-
-   * Set the number of nodes you want to train on (for instance, ``2``, ``4``, ``8``):
-
-     .. code-block:: shell
-
-        NNODES="${NNODES:-1}"
-
-   * Set the rank of each node (0 for master, 1 for the first worker node, and so on):
-
-     .. code-block:: shell
-
-        NODE_RANK="${NODE_RANK:-0}"
-
-* Use this command to run a performance benchmark test of any of the Llama 2 models that this Docker image supports (see :ref:`variables <amd-megatron-lm-benchmark-test-vars>`).
-
-  .. code-block:: shell
-
-     {variables} bash examples/llama/train_llama2.sh
-
-* Use this command to run a performance benchmark test of any of the Llama 3 and Llama 3.1 models that this Docker image supports (see :ref:`variables <amd-megatron-lm-benchmark-test-vars>`).
-
-  .. code-block:: shell
-
-     {variables} bash examples/llama/train_llama3.sh
-
-.. _amd-megatron-lm-benchmark-test-vars:
-
-The benchmark tests support the same set of variables:
-
-+--------------------------+-----------------------+-----------------------+
-| Name                     | Options               | Description           |
-+==========================+=======================+=======================+
-| ``TEE_OUTPUT``           | 0 or 1                | 0: disable training   |
-|                          |                       | log                   |
-|                          |                       |                       |
-|                          |                       | 1: enable training    |
-|                          |                       | log                   |
-+--------------------------+-----------------------+-----------------------+
-| ``MBS``                  |                       | Micro batch size      |
-+--------------------------+-----------------------+-----------------------+
-| ``BS``                   |                       | Batch size            |
-+--------------------------+-----------------------+-----------------------+
-| ``TP``                   | 1, 2, 4, 8            | Tensor parallel       |
-+--------------------------+-----------------------+-----------------------+
-| ``TE_FP8``               | 0 or 1                | Datatype.             |
-|                          |                       | If it is set to 1,    |
-|                          |                       | FP8.                  |
-|                          |                       |                       |
-|                          |                       | If it is set to 0.    |
-|                          |                       | BP16                  |
-+--------------------------+-----------------------+-----------------------+
-| ``NO_TORCH_COMPILE``     | 0 or 1                | If it is set to 1,    |
-|                          |                       | enable torch.compile. |
-|                          |                       |                       |
-|                          |                       | If it is set to 0.    |
-|                          |                       | Disable torch.compile |
-|                          |                       | (default)             |
-+--------------------------+-----------------------+-----------------------+
-| ``SEQ_LENGTH``           |                       | Input sequence length |
-+--------------------------+-----------------------+-----------------------+
-| ``GEMM_TUNING``          | 0 or 1                | If it is set to 1,    |
-|                          |                       | enable gemm tuning.   |
-|                          |                       |                       |
-|                          |                       | If it is set to 0,    |
-|                          |                       | disable gemm tuning   |
-+--------------------------+-----------------------+-----------------------+
-| ``USE_FLASH_ATTN``       | 0 or 1                | 0: disable flash      |
-|                          |                       | attention             |
-|                          |                       |                       |
-|                          |                       | 1: enable flash       |
-|                          |                       | attention             |
-+--------------------------+-----------------------+-----------------------+
-| ``ENABLE_PROFILING``     | 0 or 1                | 0: disable torch      |
-|                          |                       | profiling             |
-|                          |                       |                       |
-|                          |                       | 1: enable torch       |
-|                          |                       | profiling             |
-+--------------------------+-----------------------+-----------------------+
-| ``MODEL_SIZE``           |                       | The size of the mode: |
-|                          |                       | 7B/70B, etc.          |
-+--------------------------+-----------------------+-----------------------+
-| ``TOTAL_ITERS``          |                       | Total number of       |
-|                          |                       | iterations            |
-+--------------------------+-----------------------+-----------------------+
-| ``transformer-impl``     | transformer_engine or | Enable transformer    |
-|                          | local                 | engine by default     |
-+--------------------------+-----------------------+-----------------------+
-
-Benchmarking examples
-^^^^^^^^^^^^^^^^^^^^^
-
-.. tab-set::
-
-   .. tab-item:: Single node training
-      :sync: single
-
-      Use this command to run training with Llama 2 7B model on a single node. You can specify MBS, BS, FP,
-      datatype, and so on.
-
-      .. code-block:: bash
-
-         TEE_OUTPUT=1 MBS=5 BS=120 TP=8 TE_FP8=0 NO_TORCH_COMPILE=1
-         SEQ_LENGTH=4096 bash examples/llama/train_llama2.sh
-
-      You can find the training logs at the location defined in ``$TRAIN_LOG`` in the :ref:`configuration script <amd-megatron-lm-environment-setup>`.
-
-      See the sample output:
-
-      .. image:: ../../../data/how-to/rocm-for-ai/llama2-7b-training-log-sample.png
-         :width: 800
-
-   .. tab-item:: Multi node training
-      :sync: multi
-
-      Launch the Docker container on each node.
-
-      In this example, run training with Llama 2 7B model on 2 nodes with specific MBS, BS, FP, datatype, and
-      so on.
-
-      On the master node:
-
-      .. code-block:: bash
-
-         TEE_OUTPUT=1 MBS=4 BS=64 TP=8 TE_FP8=0 NO_TORCH_COMPILE=1
-         SEQ_LENGTH=4096 bash examples/llama/train_llama2.sh
-
-      On the worker node:
-
-      .. code-block:: bash
-
-         TEE_OUTPUT=1 MBS=4 BS=64 TP=8 TE_FP8=0 NO_TORCH_COMPILE=1
-         SEQ_LENGTH=4096 bash examples/llama/train_llama2.sh
-
-      You can find the training logs at the location defined in ``$TRAIN_LOG`` in the :ref:`configuration script <amd-megatron-lm-environment-setup>`.
-
-      Sample output for 2-node training:
-
-      Master node:
-
-      .. image:: ../../../data/how-to/rocm-for-ai/2-node-training-master.png
-         :width: 800
-
-      Worker node:
-
-      .. image:: ../../../data/how-to/rocm-for-ai/2-node-training-worker.png
-         :width: 800
-

docs/reference/gpu-arch-specs.rst

@@ -21,8 +21,6 @@ For more information about ROCm hardware compatibility, see the ROCm `Compatibil
           - Model
           - Architecture
           - LLVM target name
-          - Device Major version
-          - Device Minor version
           - VRAM (GiB)
           - Compute Units
           - Wavefront Size
@@ -34,12 +32,12 @@ For more information about ROCm hardware compatibility, see the ROCm `Compatibil
           - L1 Instruction Cache (KiB)
           - VGPR File (KiB)
           - SGPR File (KiB)
+          - GFXIP Major version
+          - GFXIP Minor version
         *
           - MI325X
           - CDNA3
           - gfx942
-          - 9
-          - 4
           - 256
           - 304 (38 per XCD)
           - 64
@@ -51,12 +49,12 @@ For more information about ROCm hardware compatibility, see the ROCm `Compatibil
           - 64 per 2 CUs
           - 512
           - 12.5
+          - 9
+          - 4
         *
           - MI300X
           - CDNA3
           - gfx942
-          - 9
-          - 4
           - 192
           - 304 (38 per XCD)
           - 64
@@ -68,12 +66,12 @@ For more information about ROCm hardware compatibility, see the ROCm `Compatibil
           - 64 per 2 CUs
           - 512
           - 12.5
+          - 9
+          - 4
         *
           - MI300A
           - CDNA3
           - gfx942
-          - 9
-          - 4
           - 128
           - 228 (38 per XCD)
           - 64
@@ -85,12 +83,12 @@ For more information about ROCm hardware compatibility, see the ROCm `Compatibil
           - 64 per 2 CUs
           - 512
           - 12.5
+          - 9
+          - 4
         *
           - MI250X
           - CDNA2
           - gfx90a
-          - 9
-          - 0
           - 128
           - 220 (110 per GCD)
           - 64
@@ -102,12 +100,12 @@ For more information about ROCm hardware compatibility, see the ROCm `Compatibil
           - 32 per 2 CUs
           - 512
           - 12.5
+          - 9
+          - 0
         *
           - MI250
           - CDNA2
           - gfx90a
-          - 9
-          - 0
           - 128
           - 208 (104 per GCD)
           - 64
@@ -119,12 +117,12 @@ For more information about ROCm hardware compatibility, see the ROCm `Compatibil
           - 32 per 2 CUs
           - 512
           - 12.5
+          - 9
+          - 0
         *
           - MI210
           - CDNA2
           - gfx90a
-          - 9
-          - 0
           - 64
           - 104
           - 64
@@ -136,12 +134,12 @@ For more information about ROCm hardware compatibility, see the ROCm `Compatibil
           - 32 per 2 CUs
           - 512
           - 12.5
+          - 9
+          - 0
         *
           - MI100
           - CDNA
           - gfx908
-          - 9
-          - 0
           - 32
           - 120
           - 64
@@ -153,12 +151,12 @@ For more information about ROCm hardware compatibility, see the ROCm `Compatibil
           - 32 per 3 CUs
           - 256 VGPR and 256 AccVGPR
           - 12.5
+          - 9
+          - 0
         *
           - MI60
           - GCN5.1
           - gfx906
-          - 9
-          - 0
           - 32
           - 64
           - 64
@@ -170,12 +168,12 @@ For more information about ROCm hardware compatibility, see the ROCm `Compatibil
           - 32 per 3 CUs
           - 256
           - 12.5
+          - 9
+          - 0
         *
           - MI50 (32GB)
           - GCN5.1
           - gfx906
-          - 9
-          - 0
           - 32
           - 60
           - 64
@@ -187,12 +185,12 @@ For more information about ROCm hardware compatibility, see the ROCm `Compatibil
           - 32 per 3 CUs
           - 256
           - 12.5
+          - 9
+          - 0
         *
           - MI50 (16GB)
           - GCN5.1
           - gfx906
-          - 9
-          - 0
           - 16
           - 60
           - 64
@@ -204,12 +202,12 @@ For more information about ROCm hardware compatibility, see the ROCm `Compatibil
           - 32 per 3 CUs
           - 256
           - 12.5
+          - 9
+          - 0
         *
           - MI25
           - GCN5.0
           - gfx900
-          - 9
-          - 0
           - 16 
           - 64
           - 64
@@ -221,12 +219,12 @@ For more information about ROCm hardware compatibility, see the ROCm `Compatibil
           - 32 per 3 CUs
           - 256
           - 12.5
+          - 9
+          - 0
         *
           - MI8
           - GCN3.0
           - gfx803
-          - 8
-          - 0
           - 4
           - 64
           - 64
@@ -238,12 +236,12 @@ For more information about ROCm hardware compatibility, see the ROCm `Compatibil
           - 32 per 4 CUs
           - 256
           - 12.5
+          - 8
+          - 0
         *
           - MI6
           - GCN4.0
           - gfx803
-          - 8
-          - 0
           - 16
           - 36
           - 64
@@ -255,6 +253,8 @@ For more information about ROCm hardware compatibility, see the ROCm `Compatibil
           - 32 per 4 CUs
           - 256
           - 12.5
+          - 8
+          - 0
 
   .. tab-item:: AMD Radeon PRO GPUs
 
@@ -266,8 +266,7 @@ For more information about ROCm hardware compatibility, see the ROCm `Compatibil
           - Model
           - Architecture
           - LLVM target name
-          - Device Major version
-          - Device Minor version
+
           - VRAM (GiB)
           - Compute Units
           - Wavefront Size
@@ -280,12 +279,12 @@ For more information about ROCm hardware compatibility, see the ROCm `Compatibil
           - L0 Instruction Cache (KiB)
           - VGPR File (KiB)
           - SGPR File (KiB)
+          - GFXIP Major version
+          - GFXIP Minor version
         *
           - Radeon PRO V710
           - RDNA3
           - gfx1101
-          - 11
-          - 0
           - 28
           - 54
           - 32
@@ -298,12 +297,12 @@ For more information about ROCm hardware compatibility, see the ROCm `Compatibil
           - 32
           - 768
           - 16
+          - 11
+          - 0
         *
           - Radeon PRO W7900 Dual Slot
           - RDNA3
           - gfx1100
-          - 11
-          - 0
           - 48
           - 96
           - 32
@@ -316,12 +315,12 @@ For more information about ROCm hardware compatibility, see the ROCm `Compatibil
           - 32
           - 768
           - 16
+          - 11
+          - 0
         *
           - Radeon PRO W7900
           - RDNA3
           - gfx1100
-          - 11
-          - 0
           - 48
           - 96
           - 32
@@ -334,12 +333,12 @@ For more information about ROCm hardware compatibility, see the ROCm `Compatibil
           - 32
           - 768
           - 16
+          - 11
+          - 0
         *
           - Radeon PRO W7800
           - RDNA3
           - gfx1100
-          - 11
-          - 0
           - 32
           - 70
           - 32
@@ -352,12 +351,12 @@ For more information about ROCm hardware compatibility, see the ROCm `Compatibil
           - 32
           - 768
           - 16
+          - 11
+          - 0
         *
           - Radeon PRO W7700
           - RDNA3
           - gfx1101
-          - 11
-          - 0
           - 16
           - 48
           - 32
@@ -370,12 +369,12 @@ For more information about ROCm hardware compatibility, see the ROCm `Compatibil
           - 32
           - 768
           - 16
+          - 11
+          - 0
         *
           - Radeon PRO W6800
           - RDNA2
           - gfx1030
-          - 10
-          - 3
           - 32
           - 60
           - 32
@@ -388,12 +387,12 @@ For more information about ROCm hardware compatibility, see the ROCm `Compatibil
           - 32
           - 512
           - 16
+          - 10
+          - 3
         *
           - Radeon PRO W6600
           - RDNA2
           - gfx1032
-          - 10
-          - 3
           - 8
           - 28
           - 32
@@ -406,12 +405,12 @@ For more information about ROCm hardware compatibility, see the ROCm `Compatibil
           - 32
           - 512
           - 16
+          - 10
+          - 3
         *
           - Radeon PRO V620
           - RDNA2
           - gfx1030
-          - 10
-          - 3
           - 32
           - 72
           - 32
@@ -424,12 +423,12 @@ For more information about ROCm hardware compatibility, see the ROCm `Compatibil
           - 32
           - 512
           - 16
+          - 10
+          - 3
         *
           - Radeon Pro W5500
           - RDNA
           - gfx1012
-          - 10
-          - 1
           - 8
           - 22
           - 32
@@ -442,12 +441,12 @@ For more information about ROCm hardware compatibility, see the ROCm `Compatibil
           - 32
           - 512
           - 20
+          - 10
+          - 1
         *
           - Radeon Pro VII
           - GCN5.1
           - gfx906
-          - 9
-          - 0
           - 16
           - 60
           - 64
@@ -460,6 +459,8 @@ For more information about ROCm hardware compatibility, see the ROCm `Compatibil
           - 32 per 3 CUs
           - 256
           - 12.5
+          - 9
+          - 0
 
   .. tab-item:: AMD Radeon GPUs
 
@@ -471,8 +472,6 @@ For more information about ROCm hardware compatibility, see the ROCm `Compatibil
           - Model
           - Architecture
           - LLVM target name
-          - Device Major version
-          - Device Minor version
           - VRAM (GiB)
           - Compute Units
           - Wavefront Size
@@ -485,12 +484,12 @@ For more information about ROCm hardware compatibility, see the ROCm `Compatibil
           - L0 Instruction Cache (KiB)
           - VGPR File (KiB)
           - SGPR File (KiB)
+          - GFXIP Major version
+          - GFXIP Minor version
         *
           - Radeon RX 7900 XTX
           - RDNA3
           - gfx1100
-          - 11
-          - 0
           - 24
           - 96
           - 32
@@ -503,12 +502,12 @@ For more information about ROCm hardware compatibility, see the ROCm `Compatibil
           - 32
           - 768
           - 16
+          - 11
+          - 0
         *
           - Radeon RX 7900 XT
           - RDNA3
           - gfx1100
-          - 11
-          - 0
           - 20
           - 84
           - 32
@@ -521,12 +520,12 @@ For more information about ROCm hardware compatibility, see the ROCm `Compatibil
           - 32
           - 768
           - 16
+          - 11
+          - 0
         *
           - Radeon RX 7900 GRE
           - RDNA3
           - gfx1100
-          - 11
-          - 0
           - 16
           - 80
           - 32
@@ -539,12 +538,12 @@ For more information about ROCm hardware compatibility, see the ROCm `Compatibil
           - 32
           - 768
           - 16
+          - 11
+          - 0
         *
           - Radeon RX 7800 XT
           - RDNA3
           - gfx1101
-          - 11
-          - 0
           - 16
           - 60
           - 32
@@ -557,12 +556,12 @@ For more information about ROCm hardware compatibility, see the ROCm `Compatibil
           - 32
           - 768
           - 16
+          - 11
+          - 0
         *
           - Radeon RX 7700 XT
           - RDNA3
           - gfx1101
-          - 11
-          - 0
           - 12
           - 54
           - 32
@@ -575,12 +574,12 @@ For more information about ROCm hardware compatibility, see the ROCm `Compatibil
           - 32
           - 768
           - 16
+          - 11
+          - 0
         *
           - Radeon RX 7600
           - RDNA3
           - gfx1102
-          - 11
-          - 0
           - 8
           - 32
           - 32
@@ -593,12 +592,12 @@ For more information about ROCm hardware compatibility, see the ROCm `Compatibil
           - 32
           - 512
           - 16
+          - 11
+          - 0
         *
           - Radeon RX 6950 XT
           - RDNA2
           - gfx1030
-          - 10
-          - 3
           - 16
           - 80
           - 32
@@ -611,12 +610,12 @@ For more information about ROCm hardware compatibility, see the ROCm `Compatibil
           - 32
           - 512
           - 16
+          - 10
+          - 3
         *
           - Radeon RX 6900 XT
           - RDNA2
           - gfx1030
-          - 10
-          - 3
           - 16
           - 80
           - 32
@@ -629,12 +628,12 @@ For more information about ROCm hardware compatibility, see the ROCm `Compatibil
           - 32
           - 512
           - 16
+          - 10
+          - 3
         *
           - Radeon RX 6800 XT
           - RDNA2
           - gfx1030
-          - 10
-          - 3
           - 16
           - 72
           - 32
@@ -647,12 +646,12 @@ For more information about ROCm hardware compatibility, see the ROCm `Compatibil
           - 32
           - 512
           - 16
+          - 10
+          - 3
         *
           - Radeon RX 6800
           - RDNA2
           - gfx1030
-          - 10
-          - 3
           - 16
           - 60
           - 32
@@ -665,12 +664,12 @@ For more information about ROCm hardware compatibility, see the ROCm `Compatibil
           - 32
           - 512
           - 16
+          - 10
+          - 3
         *
           - Radeon RX 6750 XT
           - RDNA2
           - gfx1031
-          - 10
-          - 3
           - 12
           - 40
           - 32
@@ -683,12 +682,12 @@ For more information about ROCm hardware compatibility, see the ROCm `Compatibil
           - 32
           - 512
           - 16
+          - 10
+          - 3
         *
           - Radeon RX 6700 XT
           - RDNA2
           - gfx1031
-          - 10
-          - 3
           - 12
           - 40
           - 32
@@ -701,13 +700,13 @@ For more information about ROCm hardware compatibility, see the ROCm `Compatibil
           - 32
           - 512
           - 16
+          - 10
+          - 3
         *
           - Radeon RX 6700
           - RDNA2
           - gfx1031
           - 10
-          - 3
-          - 10
           - 36
           - 32
           - 128
@@ -719,12 +718,12 @@ For more information about ROCm hardware compatibility, see the ROCm `Compatibil
           - 32
           - 512
           - 16
+          - 10
+          - 3
         *
           - Radeon RX 6650 XT
           - RDNA2
           - gfx1032
-          - 10
-          - 3
           - 8
           - 32
           - 32
@@ -737,12 +736,12 @@ For more information about ROCm hardware compatibility, see the ROCm `Compatibil
           - 32
           - 512
           - 16
+          - 10
+          - 3
         *
           - Radeon RX 6600 XT
           - RDNA2
           - gfx1032
-          - 10
-          - 3
           - 8
           - 32
           - 32
@@ -755,12 +754,12 @@ For more information about ROCm hardware compatibility, see the ROCm `Compatibil
           - 32
           - 512
           - 16
+          - 10
+          - 3
         *
           - Radeon RX 6600
           - RDNA2
           - gfx1032
-          - 10
-          - 3
           - 8
           - 28
           - 32
@@ -773,12 +772,12 @@ For more information about ROCm hardware compatibility, see the ROCm `Compatibil
           - 32
           - 512
           - 16
+          - 10
+          - 3
         *
           - Radeon VII
           - GCN5.1
           - gfx906
-          - 9
-          - 0
           - 16
           - 60
           - 64
@@ -791,6 +790,8 @@ For more information about ROCm hardware compatibility, see the ROCm `Compatibil
           - 32 per 3 CUs
           - 256
           - 12.5
+          - 9
+          - 0
 
 Glossary
 ========
@@ -804,18 +805,6 @@ For more information about the terms used, see the
 Argument to pass to clang in ``--offload-arch`` to compile code for the given
 architecture.
 
-**Device major version**
-
-Indicates the core instruction set of the GPU architecture. For example, a value
-of 11 would correspond to Navi III (RDNA3).
-
-**Device minor version**
-
-Indicates a particular configuration, feature set, or variation within the group
-represented by the device compute version. For example, different models within
-the same major version might have varying levels of support for certain features
-or optimizations.
-
 **VRAM**
 
 Amount of memory available on the GPU.
@@ -898,6 +887,26 @@ Purpose Vector Registers, used specifically in matrix instructions.
 Size of the Scalar General Purpose Register (SGPR) file. Holds data used in
 scalar instructions.
 
+**GFXIP**
+
+GFXIP (Graphics IP) is a versioning system used by AMD to identify the GPU
+architecture and its instruction set. It helps categorize different generations
+of GPUs and their feature sets.
+
+**GFXIP major version**
+
+Defines the GPU's core instruction set and architecture, which determines
+compatibility with software stacks such as HIP and OpenCL. For example, a GFXIP
+11 major version corresponds to the RDNA 3 (Navi 3x) architecture, influencing
+driver support and available compute features.
+
+**GFXIP minor version**
+
+Represents specific variations within a GFXIP major version and affects feature sets,
+optimizations, and driver behavior in software stacks such as HIP and OpenCL. Different
+GPU models within the same major version can have unique capabilities, impacting
+performance and supported instructions.
+
 **GCD**
 
 Graphics Compute Die.

docs/sphinx/_toc.yml.in

@@ -40,11 +40,13 @@ subtrees:
         title: Training
         subtrees:
         - entries:
-          - file: how-to/rocm-for-ai/training/train-a-model.rst
-            title: Train a model
+          - file: how-to/rocm-for-ai/training/benchmark-docker/megatron-lm
+            title: Train a model with Megatron-LM
+          - file: how-to/rocm-for-ai/training/benchmark-docker/pytorch-training
+            title: Train a model with PyTorch
           - file: how-to/rocm-for-ai/training/scale-model-training.rst
             title: Scale model training
-      
+
       - file: how-to/rocm-for-ai/fine-tuning/index.rst
         title: Fine-tuning LLMs
         subtrees:
@@ -152,7 +154,7 @@ subtrees:
         - entries:
           - url: https://www.amd.com/system/files/TechDocs/instinct-mi200-cdna2-instruction-set-architecture.pdf
             title: AMD Instinct MI200/CDNA2 ISA
-          - url: https://www.amd.com/system/files/documents/amd-cdna2-white-paper.pdf
+          - url: https://www.amd.com/content/dam/amd/en/documents/instinct-business-docs/white-papers/amd-cdna2-white-paper.pdf
             title: White paper
       - file: conceptual/gpu-arch/mi100.md
         title: MI100 microarchitecture
@@ -160,7 +162,7 @@ subtrees:
         - entries:
           - url: https://www.amd.com/system/files/TechDocs/instinct-mi100-cdna1-shader-instruction-set-architecture%C2%A0.pdf
             title: AMD Instinct MI100/CDNA1 ISA
-          - url: https://www.amd.com/system/files/documents/amd-cdna-whitepaper.pdf
+          - url: https://www.amd.com/content/dam/amd/en/documents/instinct-business-docs/white-papers/amd-cdna-white-paper.pdf
             title: White paper
   - file: conceptual/iommu.rst
     title: Input-Output Memory Management Unit (IOMMU)

docs/sphinx/requirements.in

@@ -1,3 +1,4 @@
-rocm-docs-core==1.15.0
+rocm-docs-core==1.17.0
 sphinx-reredirects
 sphinx-sitemap
+sphinxcontrib.datatemplates==0.11.0

docs/sphinx/requirements.txt

@@ -2,7 +2,7 @@
 # This file is autogenerated by pip-compile with Python 3.10
 # by the following command:
 #
-#    pip-compile requirements.in
+#    pip-compile docs/sphinx/requirements.in
 #
 accessible-pygments==0.0.5
     # via pydata-sphinx-theme
@@ -43,6 +43,8 @@ debugpy==1.8.12
     # via ipykernel
 decorator==5.1.1
     # via ipython
+defusedxml==0.7.1
+    # via sphinxcontrib-datatemplates
 deprecated==1.2.15
     # via pygithub
 docutils==0.21.2
@@ -175,6 +177,7 @@ pyyaml==6.0.2
     #   myst-parser
     #   rocm-docs-core
     #   sphinx-external-toc
+    #   sphinxcontrib-datatemplates
 pyzmq==26.2.0
     # via
     #   ipykernel
@@ -187,7 +190,7 @@ requests==2.32.3
     # via
     #   pygithub
     #   sphinx
-rocm-docs-core==1.15.0
+rocm-docs-core==1.17.0
     # via -r requirements.in
 rpds-py==0.22.3
     # via
@@ -215,6 +218,8 @@ sphinx==8.1.3
     #   sphinx-notfound-page
     #   sphinx-reredirects
     #   sphinx-sitemap
+    #   sphinxcontrib-datatemplates
+    #   sphinxcontrib-runcmd
 sphinx-book-theme==1.1.3
     # via rocm-docs-core
 sphinx-copybutton==0.5.2
@@ -226,11 +231,13 @@ sphinx-external-toc==1.0.1
 sphinx-notfound-page==1.0.4
     # via rocm-docs-core
 sphinx-reredirects==0.1.5
-    # via -r requirements.in
+    # via -r docs/sphinx/requirements.in
 sphinx-sitemap==2.6.0
-    # via -r requirements.in
+    # via -r docs/sphinx/requirements.in
 sphinxcontrib-applehelp==2.0.0
     # via sphinx
+sphinxcontrib-datatemplates==0.11.0
+    # via -r docs/sphinx/requirements.in
 sphinxcontrib-devhelp==2.0.0
     # via sphinx
 sphinxcontrib-htmlhelp==2.1.0
@@ -239,6 +246,8 @@ sphinxcontrib-jsmath==1.0.1
     # via sphinx
 sphinxcontrib-qthelp==2.0.0
     # via sphinx
+sphinxcontrib-runcmd==0.2.0
+    # via sphinxcontrib-datatemplates
 sphinxcontrib-serializinghtml==2.0.0
     # via sphinx
 sqlalchemy==2.0.37

docs/sphinx/static/css/vllm-benchmark.css

@@ -0,0 +1,102 @@
+/* ------------------ Compatibility options grid ------------------ */
+html {
+  --compat-border-radius: 2px;
+  --compat-accent-color: var(--pst-color-primary);
+  --compat-bg-color: var(--pst-color-on-background);
+  --compat-fg-color: var(--pst-color-primary-text);
+  --compat-head-color: var(--pst-color-surface);
+  --compat-param-hover-color: var(--pst-color-link-hover);
+  --compat-param-selected-color: var(--pst-color-primary);
+}
+
+html[data-theme="light"] {
+  --compat-border-color: var(--pst-gray-500);
+  --compat-param-disabled-color: var(--pst-gray-300);
+}
+
+html[data-theme="dark"] {
+  --compat-border-color: var(--pst-gray-600);
+  --compat-param-disabled-color: var(--pst-gray-600);
+}
+
+div#vllm-benchmark-ud-params-picker.container-fluid {
+  padding: 0 0 1rem 0;
+}
+
+div[data-param-k="model"] {
+  background-color: var(--compat-bg-color);
+  padding: 2px;
+  border: solid 1px var(--compat-border-color);
+  font-weight: 500;
+  cursor: pointer;
+}
+
+div[data-param-k="model"][data-param-state="selected"] {
+  background-color: var(--compat-param-selected-color);
+  color: var(--compat-fg-color);
+}
+
+div[data-param-k="model"][data-param-state="latest-version"] {
+  background-color: var(--compat-param-selected-color);
+  color: var(--compat-fg-color);
+}
+
+div[data-param-k="model"][data-param-state="disabled"] {
+  background-color: var(--compat-param-disabled-color);
+  text-decoration: line-through;
+  /* text-decoration-color: var(--pst-color-danger); */
+  cursor: auto;
+}
+
+div[data-param-k="model"]:not([data-param-state]):hover {
+  background-color: var(--compat-param-hover-color);
+}
+
+div[data-param-k="model-group"] {
+  background-color: var(--compat-bg-color);
+  padding: 2px;
+  border: solid 1px var(--compat-border-color);
+  font-weight: 500;
+  cursor: pointer;
+}
+
+div[data-param-k="model-group"][data-param-state="selected"] {
+  background-color: var(--compat-param-selected-color);
+  color: var(--compat-fg-color);
+}
+
+div[data-param-k="model-group"][data-param-state="latest-version"] {
+  background-color: var(--compat-param-selected-color);
+  color: var(--compat-fg-color);
+}
+
+div[data-param-k="model-group"][data-param-state="disabled"] {
+  background-color: var(--compat-param-disabled-color);
+  text-decoration: line-through;
+  /* text-decoration-color: var(--pst-color-danger); */
+  cursor: auto;
+}
+
+div[data-param-k="model-group"]:not([data-param-state]):hover {
+  background-color: var(--compat-param-hover-color);
+}
+
+.model-param-head {
+  background-color: var(--compat-head-color);
+  padding: 0.15rem 0.15rem 0.15rem 0.67rem;
+  /* margin: 2px; */
+  border-right: solid 2px var(--compat-accent-color);
+  font-weight: 600;
+}
+
+.model-param {
+  /* padding: 2px; */
+  /* margin: 0 2px 0 2px; */
+  /* margin: 2px; */
+  border: solid 1px var(--compat-border-color);
+  font-weight: 500;
+}
+
+.hidden {
+  display: none !important;
+}