[docs/7.10] post-install - improve exports (#5783)

Fix element overflow on narrow/mobile screens.

Fix incorrect warning in matrix.py extension

.azuredevops/components/ROCR-Runtime.yml

@@ -37,7 +37,6 @@ parameters:
     - libdrm-dev
     - libelf-dev
     - libnuma-dev
-    - libsimde-dev
     - ninja-build
     - pkg-config
 - name: rocmDependencies

.azuredevops/components/rccl.yml

@@ -70,7 +70,7 @@ parameters:
 jobs:
 - ${{ each job in parameters.jobMatrix.buildJobs }}:
   - job: rccl_build_${{ job.target }}
-    timeoutInMinutes: 120
+    timeoutInMinutes: 90
     variables:
     - group: common
     - template: /.azuredevops/variables-global.yml

.azuredevops/components/rocm-examples.yml

@@ -17,14 +17,8 @@ parameters:
     - libdw-dev
     - libglfw3-dev
     - libmsgpack-dev
-    - libomp-dev
-    - libopencv-dev
     - libtbb-dev
-    - libtiff-dev
     - libva-amdgpu-dev
-    - libavcodec-dev
-    - libavformat-dev
-    - libavutil-dev
     - ninja-build
     - python3-pip
 - name: rocmDependencies
@@ -44,9 +38,7 @@ parameters:
     - hipTensor
     - llvm-project
     - MIOpen
-    - MIVisionX
     - rocBLAS
-    - rocDecode
     - rocFFT
     - rocJPEG
     - rocPRIM
@@ -58,7 +50,6 @@ parameters:
     - rocSPARSE
     - rocThrust
     - rocWMMA
-    - rpp
 - name: rocmTestDependencies
   type: object
   default:
@@ -75,10 +66,7 @@ parameters:
     - hipSPARSE
     - hipTensor
     - llvm-project
-    - MIOpen
-    - MIVisionX
     - rocBLAS
-    - rocDecode
     - rocFFT
     - rocminfo
     - rocPRIM
@@ -92,7 +80,6 @@ parameters:
     - rocThrust
     - roctracer
     - rocWMMA
-    - rpp
 
 - name: jobMatrix
   type: object

.azuredevops/components/rocm-libraries.yml

@@ -43,14 +43,9 @@ parameters:
     - ninja-build
     - python3-pip
     - python3-venv
-    - googletest
-    - libgtest-dev
-    - libgmock-dev
-    - libboost-filesystem-dev
 - name: pipModules
   type: object
   default:
-    - msgpack
     - joblib
     - "packaging>=22.0"
     - pytest
@@ -152,13 +147,6 @@ jobs:
           echo "##vso[task.prependpath]$USER_BASE/bin"
           echo "##vso[task.setvariable variable=PytestCmakePath]$USER_BASE/share/Pytest/cmake"
         displayName: Set cmake configure paths
-    - task: Bash@3
-      displayName: Add ROCm binaries to PATH
-      inputs:
-        targetType: inline
-        script: |
-          echo "##vso[task.prependpath]$(Agent.BuildDirectory)/rocm/bin"
-          echo "##vso[task.prependpath]$(Agent.BuildDirectory)/rocm/llvm/bin"
     - template: ${{ variables.CI_TEMPLATE_PATH }}/steps/build-cmake.yml
       parameters:
         os: ${{ job.os }}

.azuredevops/dependencies/cli11.yml

@@ -1,63 +0,0 @@
-parameters:
-- name: checkoutRepo
-  type: string
-  default: 'self'
-- name: checkoutRef
-  type: string
-  default: ''
-- name: cli11Version
-  type: string
-  default: ''
-- name: aptPackages
-  type: object
-  default:
-    - cmake
-    - git
-    - ninja-build
-
-- name: jobMatrix
-  type: object
-  default:
-    buildJobs:
-      - { os: ubuntu2204, packageManager: apt}
-      - { os: almalinux8, packageManager: dnf}
-
-jobs:
-- ${{ each job in parameters.jobMatrix.buildJobs }}:
-  - job: cli11_${{ job.os }}
-    variables:
-    - group: common
-    - template: /.azuredevops/variables-global.yml
-    pool:
-      vmImage: 'ubuntu-22.04'
-    ${{ if eq(job.os, 'almalinux8') }}:
-      container:
-        image: rocmexternalcicd.azurecr.io/manylinux228:latest
-        endpoint: ContainerService3
-    workspace:
-      clean: all
-    steps:
-    - checkout: none
-    - template: ${{ variables.CI_TEMPLATE_PATH }}/steps/dependencies-other.yml
-      parameters:
-        aptPackages: ${{ parameters.aptPackages }}
-        packageManager: ${{ job.packageManager }}
-    - template: ${{ variables.CI_TEMPLATE_PATH }}/steps/preamble.yml
-    - task: Bash@3
-      displayName: Clone cli11 ${{ parameters.cli11Version }}
-      inputs:
-        targetType: inline
-        script: git clone https://github.com/CLIUtils/CLI11.git -b ${{ parameters.cli11Version }}
-        workingDirectory: $(Agent.BuildDirectory)
-    - template: ${{ variables.CI_TEMPLATE_PATH }}/steps/build-cmake.yml
-      parameters:
-        os: ${{ job.os }}
-        cmakeBuildDir: $(Agent.BuildDirectory)/CLI11/build
-        cmakeSourceDir: $(Agent.BuildDirectory)/CLI11
-        useAmdclang: false
-        extraBuildFlags: >-
-          -DCMAKE_BUILD_TYPE=Release
-          -GNinja
-    - template: ${{ variables.CI_TEMPLATE_PATH }}/steps/artifact-upload.yml
-      parameters:
-        os: ${{ job.os }}

.azuredevops/dependencies/yamlcpp.yml

@@ -1,66 +0,0 @@
-parameters:
-- name: checkoutRepo
-  type: string
-  default: 'self'
-- name: checkoutRef
-  type: string
-  default: ''
-- name: yamlcppVersion
-  type: string
-  default: ''
-- name: aptPackages
-  type: object
-  default:
-    - cmake
-    - git
-    - ninja-build
-
-- name: jobMatrix
-  type: object
-  default:
-    buildJobs:
-      - { os: ubuntu2204, packageManager: apt}
-      - { os: almalinux8, packageManager: dnf}
-
-jobs:
-- ${{ each job in parameters.jobMatrix.buildJobs }}:
-  - job: yamlcpp_${{ job.os }}
-    variables:
-    - group: common
-    - template: /.azuredevops/variables-global.yml
-    pool:
-      vmImage: 'ubuntu-22.04'
-    ${{ if eq(job.os, 'almalinux8') }}:
-      container:
-        image: rocmexternalcicd.azurecr.io/manylinux228:latest
-        endpoint: ContainerService3
-    workspace:
-      clean: all
-    steps:
-    - checkout: none
-    - template: ${{ variables.CI_TEMPLATE_PATH }}/steps/dependencies-other.yml
-      parameters:
-        aptPackages: ${{ parameters.aptPackages }}
-        packageManager: ${{ job.packageManager }}
-    - template: ${{ variables.CI_TEMPLATE_PATH }}/steps/preamble.yml
-    - task: Bash@3
-      displayName: Clone yaml-cpp ${{ parameters.yamlcppVersion }}
-      inputs:
-        targetType: inline
-        script: git clone  https://github.com/jbeder/yaml-cpp.git -b ${{ parameters.yamlcppVersion }}
-        workingDirectory: $(Agent.BuildDirectory)
-    - template: ${{ variables.CI_TEMPLATE_PATH }}/steps/build-cmake.yml
-      parameters:
-        os: ${{ job.os }}
-        cmakeBuildDir: $(Agent.BuildDirectory)/yaml-cpp/build
-        cmakeSourceDir: $(Agent.BuildDirectory)/yaml-cpp
-        useAmdclang: false
-        extraBuildFlags: >-
-          -DCMAKE_BUILD_TYPE=Release
-          -DYAML_CPP_BUILD_TOOLS=OFF
-          -DYAML_BUILD_SHARED_LIBS=OFF
-          -DYAML_CPP_INSTALL=ON
-          -GNinja
-    - template: ${{ variables.CI_TEMPLATE_PATH }}/steps/artifact-upload.yml
-      parameters:
-        os: ${{ job.os }}

.azuredevops/tag-builds/cli11.yml

@@ -1,23 +0,0 @@
-variables:
-- group: common
-- template: /.azuredevops/variables-global.yml
-
-parameters:
-- name: cli11Version
-  type: string
-  default: "main"
-
-resources:
-  repositories:
-  - repository: pipelines_repo
-    type: github
-    endpoint: ROCm
-    name: ROCm/ROCm
-
-trigger: none
-pr: none
-
-jobs:
-  - template: ${{ variables.CI_DEPENDENCIES_PATH }}/cli11.yml
-    parameters:
-      cli11Version: ${{ parameters.cli11Version }}

.azuredevops/tag-builds/yaml-cpp.yml

@@ -1,24 +0,0 @@
-variables:
-- group: common
-- template: /.azuredevops/variables-global.yml
-
-parameters:
-- name: yamlcppVersion
-  type: string
-  default: "0.8.0"
-
-resources:
-  repositories:
-  - repository: pipelines_repo
-    type: github
-    endpoint: ROCm
-    name: ROCm/ROCm
-
-trigger: none
-pr: none
-
-jobs:
-  - template: ${{ variables.CI_DEPENDENCIES_PATH }}/yamlcpp.yml
-    parameters:
-      yamlcppVersion: ${{ parameters.yamlcppVersion }}
-      

.wordlist.txt

@@ -3,11 +3,8 @@ ABI
 ACE
 ACEs
 ACS
-AccVGPR
-AccVGPRs
 AITER
 ALU
-AllReduce
 AMD
 AMDGPU
 AMDGPUs
@@ -22,28 +19,32 @@ APIC
 APIs
 APU
 APUs
+ASAN
 ASIC
 ASICs
 ASan
-ASAN
 ASm
 ATI
-atomicRMW
+AccVGPR
+AccVGPRs
 AddressSanitizer
 AlexNet
+AllReduce
 Andrej
 Arb
 Autocast
 BARs
-BatchNorm
 BKC
 BLAS
 BMC
 BabelStream
+BatchNorm
+Bitwise
 Blit
 Blockwise
 Bluefield
 Bootloader
+BrainFloat
 Broadcom
 CAS
 CCD
@@ -62,7 +63,6 @@ CPF
 CPP
 CPU
 CPUs
-Cron
 CSC
 CSDATA
 CSE
@@ -73,8 +73,8 @@ CTests
 CU
 CUDA
 CUs
-CXX
 CX
+CXX
 Cavium
 CentOS
 ChatGPT
@@ -87,35 +87,32 @@ Concretized
 Conda
 ConnectX
 CountOnes
+Cron
 CuPy
-da
-Dashboarding
-Dataloading
 DBRX
 DDR
 DF
 DGEMM
 DGL
 DGLGraph
-dGPU
-dGPUs
 DIMM
 DKMS
 DL
 DMA
-DOMContentLoaded
 DNN
 DNNL
+DOMContentLoaded
 DPM
 DRI
 DW
 DWORD
+Dashboarding
 Dask
 DataFrame
 DataLoader
 DataParallel
+Dataloading
 Debian
-decompositions
 DeepSeek
 DeepSpeed
 Dependabot
@@ -123,24 +120,20 @@ Deprecations
 DevCap
 DirectX
 Disaggregated
-disaggregated
 Dockerfile
 Dockerized
 Doxygen
-dropless
 ELMo
 ENDPGM
 EPYC
 ESXi
 EoS
-etcd
-fas
 FBGEMM
-FIFOs
 FFT
 FFTs
 FFmpeg
 FHS
+FIFOs
 FIXME
 FMA
 FP
@@ -149,8 +142,8 @@ Filesystem
 FindDb
 Flang
 FlashAttention
-FlashInfer’s
 FlashInfer
+FlashInfer’s
 FluxBenchmark
 Fortran
 Fuyu
@@ -167,16 +160,12 @@ GDS
 GEMM
 GEMMs
 GFLOPS
-GFortran
 GFXIP
+GFortran
 GGUF
-Gemma
-GiB
 GIM
 GL
-Glibc
 GLXT
-Gloo
 GMI
 GPG
 GPGPU
@@ -185,23 +174,25 @@ GPT
 GPU
 GPU's
 GPUs
-Graphbolt
-GraphSage
 GRBM
 GRE
+Gemma
 GenAI
 GenZ
+GiB
 GitHub
 Gitpod
+Glibc
+Gloo
+GraphSage
+Graphbolt
 HBM
 HCA
 HGX
 HIPCC
-hipDataType
 HIPExtension
 HIPIFY
 HIPification
-hipification
 HIPify
 HPC
 HPCG
@@ -211,11 +202,12 @@ HSA
 HW
 HWE
 HWS
+HX
 Haswell
 Higgs
-href
-Hyperparameters
 Huggingface
+HunyuanVideo
+Hyperparameters
 IB
 ICD
 ICT
@@ -224,13 +216,10 @@ IDE
 IDEs
 IFWI
 IMDb
-IncDec
-instrSize
-interpolators
 IOMMU
 IOP
-IOPS
 IOPM
+IOPS
 IOV
 IRQ
 ISA
@@ -238,6 +227,7 @@ ISV
 ISVs
 ITL
 ImageNet
+IncDec
 InfiniBand
 Inlines
 IntelliSense
@@ -246,8 +236,8 @@ Intersphinx
 Intra
 Ioffe
 JAX's
-Jinja
 JSON
+Jinja
 Jupyter
 KFD
 KFDTest
@@ -255,10 +245,10 @@ KMD
 KV
 KVM
 Karpathy's
-KiB
-Kineto
 Keras
 Khronos
+KiB
+Kineto
 LAPACK
 LCLK
 LDS
@@ -268,21 +258,20 @@ LLVM
 LM
 LRU
 LSAN
+LSTMs
 LSan
 LTS
-LSTMs
-LteAll
 LanguageCrossEntropy
 LoRA
+LteAll
 MECO
 MEM
 MERCHANTABILITY
 MFMA
-MiB
 MIGraphX
 MIOpen
-MIOpenGEMM
 MIOpen's
+MIOpenGEMM
 MIVisionX
 MLM
 MMA
@@ -293,9 +282,9 @@ MNIST
 MPI
 MPT
 MSVC
-mul
 MVAPICH
 MVFFR
+MXFP
 Makefile
 Makefiles
 ManyLinux
@@ -308,16 +297,16 @@ Megatron
 Mellanox
 Mellanox's
 Meta's
+MiB
 Miniconda
 MirroredStrategy
 Mixtral
-MosaicML
 MoEs
 Mooncake
+MosaicML
 Mpops
 Multicore
 Multithreaded
-MXFP
 MyEnvironment
 MyST
 NANOO
@@ -325,24 +314,27 @@ NBIO
 NBIOs
 NCCL
 NCF
+NCS
 NFS
 NIC
 NICs
 NLI
 NLP
 NN
+NOP
 NPKit
 NPS
 NSP
 NUMA
 NVCC
 NVIDIA
+NVLink
 NVPTX
 NaN
 Nano
 Navi
-Noncoherently
 NoReturn
+Noncoherently
 NousResearch's
 NumPy
 OAM
@@ -369,13 +361,10 @@ OpenVX
 OpenXLA
 Optim
 Oversubscription
-PagedAttention
-Pallas
 PCC
 PCI
 PCIe
 PEFT
-perf
 PEQT
 PIL
 PILImage
@@ -385,6 +374,8 @@ PRNG
 PRs
 PaLM
 Pageable
+PagedAttention
+Pallas
 PeerDirect
 PerfDb
 Perfetto
@@ -397,8 +388,8 @@ Pretraining
 Primus
 Profiler's
 PyPi
-Pytest
 PyTorch
+Pytest
 Qcycles
 Qwen
 RAII
@@ -409,16 +400,16 @@ RDC's
 RDMA
 RDNA
 README
-Recomputation
 RHEL
+RLHF
 RMW
 RNN
 RNNs
 ROC
 ROCProfiler
 ROCT
-ROCTx
 ROCTracer
+ROCTx
 ROCclr
 ROCdbgapi
 ROCgdb
@@ -433,8 +424,10 @@ RPP
 RST
 RW
 Radeon
+Recomputation
 RelWithDebInfo
 Req
+ResNet
 Rickle
 RoCE
 Runfile
@@ -442,7 +435,6 @@ Ryzen
 SALU
 SBIOS
 SCA
-ScaledGEMM
 SDK
 SDKs
 SDMA
@@ -461,7 +453,6 @@ SKU
 SKUs
 SLES
 SLURM
-Slurm
 SMEM
 SMFMA
 SMI
@@ -472,17 +463,17 @@ SRAM
 SRAMECC
 SVD
 SWE
+ScaledGEMM
 SerDes
 ShareGPT
 Shlens
-simd
 Skylake
+Slurm
 Softmax
 Spack
 SplitK
 Supermicro
 Szegedy
-TagRAM
 TCA
 TCC
 TCCs
@@ -490,36 +481,33 @@ TCI
 TCIU
 TCP
 TCR
-TVM
-TheRock
-THREADGROUPS
-threadgroups
-TensorRT
-TensorFloat
 TF
 TFLOPS
+THREADGROUPS
 TP
 TPS
 TPU
 TPUs
 TSME
+TVM
+TagRAM
+Tagram
 Taichi
 Taichi's
-Tagram
 TensileLite
 TensorBoard
+TensorFloat
 TensorFlow
 TensorParallel
+TensorRT
+TheRock
 ToC
+TopK
 TorchAudio
-torchaudio
 TorchElastic
 TorchMIGraphX
-torchrec
 TorchScript
 TorchServe
-torchserve
-torchtext
 TorchVision
 TransferBench
 TrapStatus
@@ -531,17 +519,18 @@ UE
 UIF
 UMC
 USM
+USM
 UTCL
+UTCL
+UTIL
 UTIL
 UltraChat
 Uncached
-Unittests
 Unhandled
-unwindowed
+Unittests
 VALU
 VBIOS
 VCN
-verl's
 VGPR
 VGPRs
 VM
@@ -552,6 +541,7 @@ VSIX
 VSkipped
 Vanhoucke
 Vulkan
+WDAG
 WGP
 WGPs
 WR
@@ -560,7 +550,6 @@ WikiText
 Wojna
 Workgroups
 Writebacks
-xcc
 XCD
 XCDs
 XGBoost
@@ -580,8 +569,8 @@ ZeRO
 ZenDNN
 accuracies
 activations
-addr
 addEventListener
+addr
 ade
 ai
 alloc
@@ -592,6 +581,7 @@ amdgpu
 api
 aten
 atmi
+atomicRMW
 atomics
 autogenerated
 autotune
@@ -608,17 +598,16 @@ bilinear
 bitcode
 bitsandbytes
 bitwise
-Bitwise
 blit
 bootloader
 boson
 bosons
 br
-BrainFloat
 btn
 buildable
 bursty
 bzip
+cTDP
 cacheable
 carveout
 cd
@@ -633,6 +622,7 @@ cmd
 coalescable
 codename
 collater
+comfyui
 comgr
 compat
 completers
@@ -649,16 +639,18 @@ copyable
 cpp
 csn
 cuBLAS
-cuda
 cuDNN
-cudnn
 cuFFT
 cuLIB
 cuRAND
 cuSOLVER
 cuSPARSE
+cuda
+cudnn
 customizations
-cTDP
+dGPU
+dGPUs
+da
 dataset
 datasets
 dataspace
@@ -668,8 +660,9 @@ datatypes
 dbgapi
 de
 deallocation
-debuggability
 debian
+debuggability
+decompositions
 deepseek
 denoise
 denoised
@@ -684,10 +677,12 @@ devicelibs
 devsel
 dgl
 dimensionality
+disaggregated
 disambiguates
 distro
 distros
 dkms
+dropless
 dtype
 eb
 el
@@ -700,10 +695,14 @@ endpgm
 enqueue
 env
 epilog
+etcd
 etcetera
 ethernet
 exascale
 executables
+fam
+fam
+fas
 ffmpeg
 filesystem
 forEach
@@ -730,8 +729,8 @@ heterogenous
 hipBLAS
 hipBLASLt
 hipBLASLt's
-hipblaslt
 hipCUB
+hipDataType
 hipFFT
 hipFORT
 hipLIB
@@ -742,10 +741,12 @@ hipSPARSELt
 hipTensor
 hipamd
 hipblas
+hipblaslt
 hipcc
 hipcub
 hipfft
 hipfort
+hipification
 hipify
 hipsolver
 hipsparse
@@ -754,6 +755,7 @@ hostname
 hotspotting
 hpc
 hpp
+href
 hsa
 hsakmt
 hyperparameter
@@ -769,7 +771,9 @@ init
 initializer
 inlining
 installable
+instrSize
 interop
+interpolators
 interprocedural
 intra
 intrinsics
@@ -812,23 +816,21 @@ mjx
 mkdir
 mlirmiopen
 mtypes
+mul
 mutex
 mvffr
 namespace
 namespaces
 nanoGPT
-NCS
-NOP
-NVLink
 num
 numref
 ocl
+ol
 opencl
 opencv
 openmp
 openssl
 optimizers
-ol
 os
 oversubscription
 pageable
@@ -839,6 +841,7 @@ param
 parameterization
 passthrough
 pe
+perf
 perfcounter
 performant
 perl
@@ -871,8 +874,6 @@ pseudorandom
 px
 py
 pytorch
-recommender
-recommenders
 quantile
 quantizer
 quasirandom
@@ -884,8 +885,10 @@ radeon
 rccl
 rdc
 rdma
-redhat
 reStructuredText
+recommender
+recommenders
+redhat
 redirections
 refactorization
 reformats
@@ -899,7 +902,6 @@ rescaling
 reusability
 rhel
 rl
-RLHF
 roadmap
 roc
 rocAL
@@ -926,8 +928,8 @@ rocm
 rocminfo
 rocprim
 rocprof
-rocprofv
 rocprofiler
+rocprofv
 rocr
 rocrand
 rocsolver
@@ -938,7 +940,6 @@ rst
 runtime
 runtimes
 ryzen
-ResNet
 sL
 scalability
 scalable
@@ -953,6 +954,7 @@ sglang
 shader
 sharding
 sigmoid
+simd
 sles
 sm
 smi
@@ -973,6 +975,7 @@ submodule
 submodules
 subnet
 supercomputing
+suse
 symlink
 symlinks
 sys
@@ -982,6 +985,7 @@ td
 tensorfloat
 tf
 th
+threadgroups
 tokenization
 tokenize
 tokenized
@@ -991,17 +995,21 @@ toolchain
 toolchains
 toolset
 toolsets
+torchaudio
+torchrec
+torchserve
+torchtext
 torchtitan
 torchvision
 tp
 tqdm
 tracebacks
 txt
-TopK
 uarch
 ubuntu
-uncached
+udev
 uncacheable
+uncached
 uncorrectable
 underoptimized
 unhandled
@@ -1010,12 +1018,11 @@ unmapped
 unsqueeze
 unstacking
 unswitching
+untar
 untrusted
 untuned
+unwindowed
 upvote
-USM
-UTCL
-UTIL
 utils
 vL
 variational
@@ -1027,6 +1034,7 @@ vectorized
 vectorizer
 vectorizes
 verl
+verl's
 virtualize
 virtualized
 vjxb
@@ -1045,9 +1053,12 @@ writeback
 writebacks
 wrreq
 wzo
-xargs
+xDiT
 xGMI
 xPacked
+xargs
+xcc
+xdit
 xz
 yaml
 ysvmadyb

README.md

@@ -26,10 +26,16 @@ source software compilers, debuggers, and libraries. ROCm is fully integrated in
 
 ## Getting and Building ROCm from Source
 
-Please use [TheRock](https://github.com/ROCm/TheRock/tree/release/therock-7.9) build system to build ROCm from source.
+Please use [TheRock](https://github.com/ROCm/TheRock) build system to build ROCm from source.
 
 ## ROCm documentation
 
+This repository contains the [manifest file](https://gerrit.googlesource.com/git-repo/+/HEAD/docs/manifest-format.md)
+for ROCm releases, changelogs, and release information.
+
+The `default.xml` file contains information for all repositories and the associated commit used to build
+the current ROCm release; `default.xml` uses the [Manifest Format repository](https://gerrit.googlesource.com/git-repo/).
+
 Source code for our documentation is located in the `/docs` folder of most ROCm repositories. The
 `develop` branch of our repositories contains content for the next ROCm release.
 

default.xml

@@ -0,0 +1,67 @@
+<?xml version="1.0" encoding="UTF-8"?>
+<manifest>
+    <remote name="rocm-org" fetch="https://github.com/ROCm/" />
+    <default revision="refs/tags/rocm-7.0.2"
+     remote="rocm-org"
+     sync-c="true"
+     sync-j="4" />
+<!--list of projects for ROCm-->
+    <project name="ROCK-Kernel-Driver" />
+    <project name="ROCR-Runtime" />
+    <project name="amdsmi" />
+    <project name="aqlprofile" />
+    <project name="rdc" />
+    <project name="rocm_bandwidth_test" />
+    <project name="rocm_smi_lib" />
+    <project name="rocm-core" />
+    <project name="rocm-examples" />
+    <project name="rocminfo" />
+    <project name="rocprofiler" />
+    <project name="rocprofiler-register" />
+    <project name="rocprofiler-sdk" />
+    <project name="rocprofiler-compute" />
+    <project name="rocprofiler-systems" />
+    <project name="roctracer" />
+<!--HIP Projects-->
+    <project name="hip" />
+    <project name="hip-tests" />
+    <project name="HIPIFY" />
+    <project name="clr" />
+    <project name="hipother" />
+<!-- The following projects are all associated with the AMDGPU LLVM compiler -->
+    <project name="half" />
+    <project name="llvm-project" />
+    <project name="spirv-llvm-translator" />
+<!-- gdb projects -->
+    <project name="ROCdbgapi" />
+    <project name="ROCgdb" />
+    <project name="rocr_debug_agent" />
+<!-- ROCm Libraries -->
+    <project groups="mathlibs" name="AMDMIGraphX" />
+    <project groups="mathlibs" name="MIVisionX" />
+    <project groups="mathlibs" name="ROCmValidationSuite" />
+    <project groups="mathlibs" name="composable_kernel" />
+    <project groups="mathlibs" name="hipTensor" />
+    <project groups="mathlibs" name="hipfort" />
+    <project groups="mathlibs" name="rccl" />
+    <project groups="mathlibs" name="rocAL" />
+    <project groups="mathlibs" name="rocALUTION" />
+    <project groups="mathlibs" name="rocDecode" />
+    <project groups="mathlibs" name="rocJPEG" />
+    <!-- The following components have been migrated to rocm-libraries:
+        hipBLAS-common hipBLAS hipBLASLt hipCUB
+        hipFFT hipRAND hipSPARSE hipSPARSELt
+        MIOpen rocBLAS rocFFT rocPRIM rocRAND
+        rocSPARSE rocThrust Tensile -->
+    <project groups="mathlibs" name="rocm-libraries" />
+    <project groups="mathlibs" name="rocPyDecode" />
+    <project groups="mathlibs" name="rocSHMEM" />
+    <project groups="mathlibs" name="rocWMMA" />
+    <project groups="mathlibs" name="rocm-cmake" />
+    <project groups="mathlibs" name="rpp" />
+    <project groups="mathlibs" name="TransferBench" />
+<!-- Projects for OpenMP-Extras -->
+    <project name="aomp" path="openmp-extras/aomp" />
+    <project name="aomp-extras" path="openmp-extras/aomp-extras" />
+    <project name="flang" path="openmp-extras/flang" />
+</manifest>

docs/compatibility/compatibility-matrix.rst

@@ -0,0 +1,111 @@
+****************************************
+ROCm |ROCM_VERSION| compatibility matrix
+****************************************
+
+To plan your ROCm |ROCM_VERSION| installation, use the following selector to
+view ROCm compatibility and system requirements information for your AMD
+hardware configuration. For installation instructions, see
+:doc:`/install/rocm`.
+
+.. include:: ./includes/selector.rst
+
+----
+
+Hardware, software, and firmware requirements
+=============================================
+
+ROCm depends on a coordinated stack of compatible firmware, driver, and user
+space components. Maintaining version alignment between these layers ensures
+expected GPU operation and performance, especially for AMD data center products.
+Future preview releases will expand hardware and operating system coverage.
+
+ROCm 7.10.0 enables support for primarily compute workloads. Future releases
+will support mixed workloads (compute and graphics).
+
+.. selected:: os=ubuntu os=rhel os=sles
+
+   .. selected:: fam=radeon-pro fam=radeon
+
+      If you’re interested in testing AMD Radeon™ GPUs with preview support for
+      graphics use cases with AMD ROCm 7.10.0, install Radeon Software for Linux
+      version 25.30.1 from `Linux Drivers for AMD Radeon and Radeon PRO
+      Graphics <https://www.amd.com/en/support/download/linux-drivers.html>`__.
+
+   .. selected:: fam=ryzen
+
+      If you're interested in testing AMD Ryzen™ APUs with preview support for
+      graphics use cases with AMD ROCm 7.10.0, use the inbox graphics drivers of
+      Ubuntu 24.04.3.
+
+.. include:: ./includes/system-instinct.rst
+
+.. include:: ./includes/system-radeon-pro.rst
+
+.. include:: ./includes/system-radeon.rst
+
+.. include:: ./includes/system-ryzen.rst
+
+----
+
+.. _rocm-compat-frameworks:
+
+Deep learning frameworks
+========================
+
+ROCm |ROCM_VERSION| provides optimized support for popular deep learning
+frameworks. The following table lists supported frameworks and their supported
+versions.
+
+.. _rocm-compat-pytorch:
+
+.. matrix::
+
+   .. matrix-head::
+
+      .. matrix-row::
+
+         .. matrix-cell:: Framework
+            :header:
+
+         .. matrix-cell:: Supported versions
+            :header:
+
+   .. matrix-row::
+      :show-when: fam=instinct fam=ryzen
+
+      .. matrix-cell:: PyTorch
+
+      .. matrix-cell:: 2.9.1, 2.8.0, 2.7.1
+         :show-when: os=ubuntu os=rhel os=sles
+
+      .. matrix-cell:: 2.9.1
+         :show-when: os=windows
+
+   .. matrix-row::
+      :show-when: fam=radeon fam=radeon-pro
+
+      .. matrix-cell:: PyTorch
+
+      .. matrix-cell:: 2.9.1
+
+For installation instructions, see :ref:`pip-install-pytorch`.
+
+.. _rocm-compat-python:
+
+.. note::
+
+   ROCm |ROCM_VERSION| is compatible with Python versions **3.11**, **3.12**,
+   and **3.13**.
+
+----
+
+ROCm Core SDK components
+========================
+
+The following table lists core components included in the ROCm |ROCM_VERSION|
+release. Expect future releases in this stream to expand the list of
+components.
+
+.. include:: ./includes/core-sdk-components-linux.rst
+
+.. include:: ./includes/core-sdk-components-windows.rst

docs/compatibility/includes/core-sdk-components-linux.rst

@@ -0,0 +1,214 @@
+.. matrix::
+   :show-when: os=ubuntu os=rhel os=sles
+
+   .. matrix-head::
+
+      .. matrix-row::
+         :header:
+
+         .. matrix-cell:: Component group
+
+         .. matrix-cell:: Component name
+
+   .. matrix-row::
+
+      .. matrix-cell:: Runtimes and compilers
+         :rowspan: 5
+
+      .. matrix-cell::
+
+         `HIP <https://github.com/ROCm/rocm-systems/tree/therock-7.10/projects/hip>`__
+
+   .. matrix-row::
+
+      .. matrix-cell::
+
+         `HIPIFY <https://github.com/ROCm/HIPIFY/tree/therock-7.10>`__
+
+   .. matrix-row::
+
+      .. matrix-cell::
+
+         `LLVM <https://github.com/ROCm/llvm-project/tree/therock-7.10>`__
+
+   .. matrix-row::
+
+      .. matrix-cell::
+
+         `ROCr Runtime <https://github.com/ROCm/rocm-systems/tree/therock-7.10/projects/rocr-runtime>`__
+
+   .. matrix-row::
+
+      .. matrix-cell::
+
+         `SPIRV-LLVM-Translator <https://github.com/ROCm/SPIRV-LLVM-Translator/tree/therock-7.10>`__
+
+   .. matrix-row::
+
+      .. matrix-cell:: Control and monitoring tools
+         :rowspan: 2
+
+      .. matrix-cell::
+
+         `AMD SMI <https://github.com/ROCm/amdsmi/tree/release/therock-7.10>`__
+
+   .. matrix-row::
+
+      .. matrix-cell::
+
+         `rocminfo <https://github.com/ROCm/rocm-systems/tree/therock-7.10/projects/rocminfo>`__
+
+   .. matrix-row::
+
+      .. matrix-cell:: Profiling and debugging tools
+         :rowspan: 3
+         :show-when: fam=instinct
+
+      .. matrix-cell:: Profiling and debugging tools
+         :rowspan: 2
+         :show-when: fam=radeon-pro fam=radeon fam=ryzen
+
+   .. matrix-row::
+
+      .. matrix-cell::
+
+         `ROCm Compute Profiler (rocprofiler-compute) <https://github.com/ROCm/rocm-systems/tree/therock-7.10/projects/rocprofiler-compute>`__
+
+   .. matrix-row::
+      :show-when: fam=instinct
+
+      .. matrix-cell::
+
+         `ROCprofiler-SDK <https://github.com/ROCm/rocm-systems/tree/therock-7.10/projects/rocprofiler-sdk>`__
+
+   .. matrix-row::
+
+      .. matrix-cell:: Math and compute libraries
+         :rowspan: 18
+         :show-when: fam=instinct
+
+      .. matrix-cell:: Math and compute libraries
+         :rowspan: 17
+         :show-when: fam=radeon-pro fam=radeon fam=ryzen
+
+      .. matrix-cell::
+
+         `rocBLAS <https://github.com/ROCm/rocm-libraries/tree/therock-7.10/projects/rocblas>`__
+
+   .. matrix-row::
+
+      .. matrix-cell::
+
+         `hipBLAS <https://github.com/ROCm/rocm-libraries/tree/therock-7.10/projects/hipblas>`__
+
+   .. matrix-row::
+
+      .. matrix-cell::
+
+         `hipBLASLt <https://github.com/ROCm/rocm-libraries/tree/therock-7.10/projects/hipblaslt>`__
+
+   .. matrix-row::
+
+      .. matrix-cell::
+
+         `rocFFT <https://github.com/ROCm/rocm-libraries/tree/therock-7.10/projects/rocfft>`__
+
+   .. matrix-row::
+
+      .. matrix-cell::
+
+         `hipFFT <https://github.com/ROCm/rocm-libraries/tree/therock-7.10/projects/hipfft>`__
+
+   .. matrix-row::
+
+      .. matrix-cell::
+
+         `rocRAND <https://github.com/ROCm/rocm-libraries/tree/therock-7.10/projects/rocrand>`__
+
+   .. matrix-row::
+
+      .. matrix-cell::
+
+         `hipRAND <https://github.com/ROCm/rocm-libraries/tree/therock-7.10/projects/hiprand>`__
+
+   .. matrix-row::
+
+      .. matrix-cell::
+
+         `rocSOLVER <https://github.com/ROCm/rocm-libraries/tree/therock-7.10/projects/rocsolver>`__
+
+   .. matrix-row::
+
+      .. matrix-cell::
+
+         `hipSOLVER <https://github.com/ROCm/rocm-libraries/tree/therock-7.10/projects/hipsolver>`__
+
+   .. matrix-row::
+
+      .. matrix-cell::
+
+         `rocSPARSE <https://github.com/ROCm/rocm-libraries/tree/therock-7.10/projects/rocsparse>`__
+
+   .. matrix-row::
+
+      .. matrix-cell::
+
+         `hipSPARSE <https://github.com/ROCm/rocm-libraries/tree/therock-7.10/projects/hipsparse>`__
+
+   .. matrix-row::
+
+      .. matrix-cell::
+
+         `hipSPARSELt <https://github.com/ROCm/rocm-libraries/tree/therock-7.10/projects/hipsparselt>`__
+
+   .. matrix-row::
+
+      .. matrix-cell::
+
+         `rocPRIM <https://github.com/ROCm/rocm-libraries/tree/therock-7.10/projects/rocprim>`__
+
+   .. matrix-row::
+
+      .. matrix-cell::
+
+         `rocThrust <https://github.com/ROCm/rocm-libraries/tree/therock-7.10/projects/rocthrust>`__
+
+   .. matrix-row::
+
+      .. matrix-cell::
+
+         `hipCUB <https://github.com/ROCm/rocm-libraries/tree/therock-7.10/projects/hipcub>`__
+
+   .. matrix-row::
+
+      .. matrix-cell::
+
+         `rocWMMA <https://github.com/ROCm/rocm-libraries/tree/therock-7.10/projects/rocwmma>`__
+
+   .. matrix-row::
+
+      .. matrix-cell::
+
+         `Composable Kernel <https://github.com/ROCm/rocm-libraries/tree/therock-7.10/projects/composablekernel>`__ (partial, limited support)
+
+   .. matrix-row::
+      :show-when: fam=instinct
+
+      .. matrix-cell::
+
+         `MIOpen <https://github.com/ROCm/rocm-libraries/tree/therock-7.10/projects/miopen>`__
+
+   .. matrix-row::
+
+      .. matrix-cell:: Communication librarires
+
+      .. matrix-cell::
+
+         `RCCL <https://github.com/ROCm/rccl/tree/release/therock-7.10>`__
+   .. matrix-row::
+
+      .. matrix-cell:: Support libraries
+
+      .. matrix-cell::
+
+         `ROCm CMake <https://github.com/ROCm/rocm-cmake/tree/therock-7.10>`__

docs/compatibility/includes/core-sdk-components-windows.rst

@@ -0,0 +1,139 @@
+.. matrix::
+   :show-when: os=windows
+
+   .. matrix-head::
+
+      .. matrix-row::
+         :header:
+
+         .. matrix-cell:: Component group
+
+         .. matrix-cell:: Component name
+
+   .. matrix-row::
+
+      .. matrix-cell:: Runtimes and compilers
+         :rowspan: 3
+
+      .. matrix-cell::
+
+         `HIP <https://github.com/ROCm/rocm-systems/tree/therock-7.10/projects/hip>`__
+
+   .. matrix-row::
+
+      .. matrix-cell::
+
+         `HIPIFY <https://github.com/ROCm/HIPIFY/tree/therock-7.10>`__
+
+   .. matrix-row::
+
+      .. matrix-cell::
+
+         `LLVM <https://github.com/ROCm/llvm-project/tree/therock-7.10>`__
+
+   .. matrix-row::
+
+      .. matrix-cell:: Control and monitoring tools
+
+      .. matrix-cell:: hipinfo
+
+   .. matrix-row::
+
+      .. matrix-cell:: Math and compute libraries
+         :rowspan: 15
+
+      .. matrix-cell::
+
+         `rocBLAS <https://github.com/ROCm/rocm-libraries/tree/therock-7.10/projects/rocblas>`__
+
+   .. matrix-row::
+
+      .. matrix-cell::
+
+         `hipBLAS <https://github.com/ROCm/rocm-libraries/tree/therock-7.10/projects/hipblas>`__
+
+   .. matrix-row::
+
+      .. matrix-cell::
+
+         `hipBLASLt <https://github.com/ROCm/rocm-libraries/tree/therock-7.10/projects/hipblaslt>`__
+
+   .. matrix-row::
+
+      .. matrix-cell::
+
+         `rocFFT <https://github.com/ROCm/rocm-libraries/tree/therock-7.10/projects/rocfft>`__
+
+   .. matrix-row::
+
+      .. matrix-cell::
+
+         `hipFFT <https://github.com/ROCm/rocm-libraries/tree/therock-7.10/projects/hipfft>`__
+
+   .. matrix-row::
+
+      .. matrix-cell::
+
+         `rocRAND <https://github.com/ROCm/rocm-libraries/tree/therock-7.10/projects/rocrand>`__
+
+   .. matrix-row::
+
+      .. matrix-cell::
+
+         `hipRAND <https://github.com/ROCm/rocm-libraries/tree/therock-7.10/projects/rocsolver>`__
+
+   .. matrix-row::
+
+      .. matrix-cell::
+
+         `rocSOLVER <https://github.com/ROCm/rocm-libraries/tree/therock-7.9.0/projects/rocsolver>`__
+
+   .. matrix-row::
+
+      .. matrix-cell::
+
+         `hipSOLVER <https://github.com/ROCm/rocm-libraries/tree/therock-7.10/projects/hipsolver>`__
+
+   .. matrix-row::
+
+      .. matrix-cell::
+
+         `rocSPARSE <https://github.com/ROCm/rocm-libraries/tree/therock-7.10/projects/rocsparse>`__
+
+   .. matrix-row::
+
+      .. matrix-cell::
+
+         `hipSPARSE <https://github.com/ROCm/rocm-libraries/tree/therock-7.10/projects/hipsparse>`__
+
+   .. matrix-row::
+
+      .. matrix-cell::
+
+         `rocPRIM <https://github.com/ROCm/rocm-libraries/tree/therock-7.10/projects/rocprim>`__
+
+   .. matrix-row::
+
+      .. matrix-cell::
+
+         `rocThrust <https://github.com/ROCm/rocm-libraries/tree/therock-7.10/projects/rocthrust>`__
+
+   .. matrix-row::
+
+      .. matrix-cell::
+
+         `hipCUB <https://github.com/ROCm/rocm-libraries/tree/therock-7.10/projects/hipcub>`__
+
+   .. matrix-row::
+
+      .. matrix-cell::
+
+         `rocWMMA <https://github.com/ROCm/rocm-libraries/tree/therock-7.10/projects/rocwmma>`__
+
+   .. matrix-row::
+
+      .. matrix-cell:: Support libraries
+
+      .. matrix-cell::
+
+         `ROCm CMake <https://github.com/ROCm/rocm-cmake/tree/therock-7.10>`__

docs/compatibility/includes/selector.rst

@@ -0,0 +1,128 @@
+.. selector:: AMD device family
+   :key: fam
+
+   .. selector-option:: Instinct
+      :value: instinct
+      :width: 3
+
+   .. selector-option:: Radeon PRO
+      :value: radeon-pro
+      :width: 3
+
+   .. selector-option:: Radeon RX
+      :value: radeon
+      :width: 3
+
+   .. selector-option:: Ryzen AI
+      :value: ryzen
+      :width: 3
+
+.. selector:: Instinct GPU
+   :key: gfx
+   :show-when: fam=instinct
+
+   .. selector-info:: https://www.amd.com/en/products/accelerators/instinct.html
+
+   .. selector-option:: Instinct MI355X<br>Instinct MI350X
+      :value: 950
+      :width: 4
+
+   .. selector-option:: Instinct MI325X<br>Instinct MI300X<br>Instinct MI300A
+      :value: 942
+      :width: 4
+
+   .. selector-option:: Instinct MI250X<br>Instinct MI250<br>Instinct MI210
+      :value: 90a
+      :width: 4
+
+.. selector:: Radeon PRO GPU
+   :key: gfx
+   :show-when: fam=radeon-pro
+
+   .. selector-info:: https://www.amd.com/en/products/graphics/workstations.html
+
+   .. selector-option:: Radeon PRO W7900D<br>Radeon PRO W7900<br>Radeon PRO W7800 48GB<br>Radeon PRO W7800
+      :value: 1100
+      :width: 6
+
+   .. selector-option:: Radeon PRO W7700
+      :value: 1101
+      :width: 6
+
+.. selector:: Radeon RX GPU
+   :key: gfx
+   :show-when: fam=radeon
+
+   .. selector-info:: https://www.amd.com/en/products/graphics/desktops/radeon.html
+
+   .. selector-option:: Radeon RX 7900 XTX<br>Radeon RX 7900 XT<br>Radeon RX 7900 GRE
+      :value: 1100
+
+   .. selector-option:: Radeon RX 7800 XT<br>Radeon RX 7700 XT
+      :value: 1101
+
+.. selector:: Ryzen AI APU
+   :key: gfx
+   :show-when: fam=ryzen
+
+   .. selector-info:: https://www.amd.com/en/products/processors/workstations/mobile.html
+
+   .. selector-option:: Ryzen AI Max+ PRO 395<br>Ryzen AI Max PRO 390, 385, 380<br>Ryzen AI Max+ 395<br>Ryzen AI Max 390, 385
+      :value: 1151
+      :width: 6
+
+   .. selector-option:: Ryzen AI 9 HX 375<br>Ryzen AI 9 HX 370<br>Ryzen AI 9 365
+      :value: 1150
+      :width: 6
+
+.. selector:: Operating system
+   :key: os
+   :show-when: fam=instinct
+
+   .. selector-option:: Ubuntu
+      :value: ubuntu
+      :icon: fab fa-ubuntu fa-lg
+      :width: 4
+
+   .. selector-option:: RHEL
+      :value: rhel
+      :icon: fab fa-redhat fa-lg
+      :width: 4
+
+   .. selector-option:: SLES
+      :value: sles
+      :icon: fab fa-suse fa-lg
+      :width: 4
+
+.. selector:: Operating system
+   :key: os
+   :show-when: fam=radeon-pro fam=radeon
+
+   .. selector-option:: Ubuntu
+      :value: ubuntu
+      :icon: fab fa-ubuntu fa-lg
+      :width: 4
+
+   .. selector-option:: RHEL
+      :value: rhel
+      :icon: fab fa-redhat fa-lg
+      :width: 4
+
+   .. selector-option:: Windows
+      :value: windows
+      :icon: fab fa-windows fa-lg
+      :width: 4
+
+.. selector:: Operating system
+   :key: os
+   :show-when: fam=ryzen
+
+   .. selector-option:: Ubuntu
+      :value: ubuntu
+      :icon: fab fa-ubuntu fa-lg
+      :width: 6
+
+   .. selector-option:: Windows
+      :value: windows
+      :icon: fab fa-windows fa-lg
+      :width: 6

docs/compatibility/includes/system-instinct.rst

@@ -0,0 +1,147 @@
+.. matrix::
+   :show-when: fam=instinct
+
+   .. matrix-row::
+      :show-when: gfx=950
+
+      .. matrix-cell:: AMD Instinct MI350 Series
+         :header:
+
+      .. matrix-cell::
+
+         `Instinct MI355X <https://www.amd.com/en/products/accelerators/instinct/mi350/mi355x.html>`__
+
+         `Instinct MI350X <https://www.amd.com/en/products/accelerators/instinct/mi350/mi350x.html>`__
+
+   .. matrix-row::
+      :show-when: gfx=942
+
+      .. matrix-cell:: AMD Instinct MI300 Series
+         :header:
+
+      .. matrix-cell::
+
+         `Instinct MI325X <https://www.amd.com/en/products/accelerators/instinct/mi300/mi325x.html>`__
+
+         `Instinct MI300X <https://www.amd.com/en/products/accelerators/instinct/mi300/mi300x.html>`__
+
+         `Instinct MI300A <https://www.amd.com/en/products/accelerators/instinct/mi300/mi300a.html>`__
+
+   .. matrix-row::
+      :show-when: gfx=90a
+
+      .. matrix-cell:: AMD Instinct MI200 Series
+         :header:
+
+      .. matrix-cell::
+
+         `Instinct MI250X <https://www.amd.com/en/products/accelerators/instinct/mi200/mi250x.html>`__
+
+         `Instinct MI250 <https://www.amd.com/en/products/accelerators/instinct/mi200/mi250.html>`__
+
+         `Instinct MI210 <https://www.amd.com/en/products/accelerators/instinct/mi200/mi210.html>`__
+
+   .. matrix-row::
+
+      .. matrix-cell:: Architecture
+         :header:
+
+      .. matrix-cell:: CDNA 4
+         :show-when: gfx=950
+
+      .. matrix-cell:: CDNA 3
+         :show-when: gfx=942
+
+      .. matrix-cell:: CDNA 2
+         :show-when: gfx=90a
+
+   .. matrix-row::
+
+      .. matrix-cell:: LLVM target
+         :header:
+
+      .. matrix-cell:: gfx950
+         :show-when: gfx=950
+
+      .. matrix-cell:: gfx942
+         :show-when: gfx=942
+
+      .. matrix-cell:: gfx90a
+         :show-when: gfx=90a
+
+   .. matrix-row::
+      :show-when: os=ubuntu
+
+      .. matrix-cell:: Supported Ubuntu versions
+         :header:
+
+      .. matrix-cell::
+
+         Ubuntu 24.04.3 (GA kernel: 6.8)
+
+         Ubuntu 22.04.5 (GA kernel: 5.15)
+
+   .. matrix-row::
+      :show-when: os=rhel
+
+      .. matrix-cell:: Supported Red Hat Enterprise Linux versions
+         :header:
+
+      .. matrix-cell::
+
+	 RHEL 10.1 (kernel: 6.12.0-124)
+
+	 RHEL 10.0 (kernel: 6.12.0-55)
+
+	 RHEL 9.7 (kernel: 5.14.0-611)
+
+	 RHEL 9.6 (kernel: 5.14.0-570)
+
+	 RHEL 8.10 (kernel: 4.18.0-553)
+
+   .. matrix-row::
+      :show-when: os=sles
+
+      .. matrix-cell:: Supported SUSE Linux Enterprise Server version
+         :header:
+
+      .. matrix-cell:: SLES 15.7 (kernel: 6.4.0-150700.51)
+
+   .. matrix-row::
+
+      .. matrix-cell:: Supported AMD GPU Driver (amdgpu) versions
+         :header:
+
+      .. matrix-cell:: 
+
+         `30.20.0 <https://instinct.docs.amd.com/projects/amdgpu-docs/en/docs-30.20.0/>`__,
+         `30.10.2 <https://instinct.docs.amd.com/projects/amdgpu-docs/en/docs-30.10.2/>`__,
+         `30.10.1 <https://instinct.docs.amd.com/projects/amdgpu-docs/en/docs-30.10.1/>`__,
+         `30.10.0 <https://instinct.docs.amd.com/projects/amdgpu-docs/en/docs-30.10/>`__
+
+   .. matrix-row::
+
+      .. matrix-cell:: Supported PLDM bundle (firmware) versions
+         :header:
+
+      .. matrix-cell:: 01.25.15.04, 01.25.13.09
+         :show-when: gfx=950
+
+      .. matrix-cell::
+         :show-when: gfx=942
+
+         **MI325X** 01.25.04.02, 01.25.03.03
+
+         **MI300X** 01.25.05.00 (or later), 01.25.03.12
+
+         **MI300A** BKC 26, 25
+
+      .. matrix-cell::
+         :show-when: gfx=90a
+
+         **MI250X** IFWI 47 (or later)
+
+         **MI250** Maintenance update 5 with IFWI 75 (or later)
+
+         **MI210** Maintenance update 5 with IFWI 75 (or later)
+

docs/compatibility/includes/system-radeon-pro.rst

@@ -0,0 +1,111 @@
+.. matrix::
+   :show-when: fam=radeon-pro
+
+   .. matrix-row::
+      :show-when: gfx=1100
+
+      .. matrix-cell:: AMD Radeon PRO W7000 Series
+         :header:
+
+      .. matrix-cell::
+
+         `Radeon PRO W7900D <https://www.amd.com/en/support/downloads/drivers.html/graphics/radeon-pro/radeon-pro-w7000-series/amd-radeon-pro-w7900d.html>`__
+
+         `Radeon PRO W7900 <https://www.amd.com/en/products/graphics/workstations/radeon-pro/w7900.html>`__
+
+         `Radeon PRO W7800 48GB <https://www.amd.com/en/products/graphics/workstations/radeon-pro/w7800-48gb.html>`__
+
+         `Radeon PRO W7800 <https://www.amd.com/en/products/graphics/workstations/radeon-pro/w7800.html>`__
+
+   .. matrix-row::
+      :show-when: gfx=1101
+
+      .. matrix-cell:: AMD Radeon PRO W7000 Series
+         :header:
+
+      .. matrix-cell::
+
+         `Radeon PRO W7700 <https://www.amd.com/en/products/graphics/workstations/radeon-pro/w7700.html>`__
+
+   .. matrix-row::
+
+      .. matrix-cell:: Architecture
+         :header:
+
+      .. matrix-cell:: RDNA 3
+
+   .. matrix-row::
+
+      .. matrix-cell:: LLVM target
+         :header:
+
+      .. matrix-cell:: gfx1101
+         :show-when: gfx=1101
+
+      .. matrix-cell:: gfx1100
+         :show-when: gfx=1100
+
+   .. matrix-row::
+      :show-when: os=ubuntu
+
+      .. matrix-cell:: Supported Ubuntu versions
+         :header:
+
+      .. matrix-cell::
+
+         24.04.3 (GA kernel: 6.8)
+
+         22.04.5 (GA kernel: 5.15)
+
+   .. matrix-row::
+      :show-when: os=rhel
+
+      .. matrix-cell:: Supported RHEL versions
+         :header:
+
+      .. matrix-cell:: 10.1, 10.0
+
+   .. matrix-row::
+      :show-when: os=windows
+
+      .. matrix-cell:: Supported Windows version
+         :header:
+
+      .. matrix-cell:: Windows 11 25H2
+
+   .. matrix-row::
+
+      .. matrix-cell:: Supported AMD GPU Driver (amdgpu) versions
+         :header:
+
+      .. matrix-cell:: 
+
+         `30.20.0 <https://instinct.docs.amd.com/projects/amdgpu-docs/en/docs-30.20.0/>`__,
+         `30.10.2 <https://instinct.docs.amd.com/projects/amdgpu-docs/en/docs-30.10.2/>`__,
+         `30.10.1 <https://instinct.docs.amd.com/projects/amdgpu-docs/en/docs-30.10.1/>`__,
+         `30.10.0 <https://instinct.docs.amd.com/projects/amdgpu-docs/en/docs-30.10/>`__
+
+   .. matrix-row::
+      :show-when: os=ubuntu os=rhel os=sles
+
+      .. matrix-cell:: Supported Radeon Software for Linux version
+         :header:
+
+      .. matrix-cell::
+
+         `25.30.1 <https://www.amd.com/en/support/download/linux-drivers.html#linux-for-radeon-pro>`__
+
+   .. matrix-row::
+      :show-when: os=windows
+
+      .. matrix-cell:: Supported Adrenalin Driver version
+         :header:
+
+      .. matrix-cell::
+
+         `25.11.1 <https://www.amd.com/en/resources/support-articles/release-notes/RN-RAD-WIN-25-11-1.html>`__
+         (generally recommended)
+
+         `25.20.01.17 <https://www.amd.com/en/resources/support-articles/release-notes/RN-AMDGPU-WINDOWS-PYTORCH-7-1-1.html>`__
+         (recommended for ComfyUI)
+

docs/compatibility/includes/system-radeon.rst

@@ -0,0 +1,112 @@
+.. matrix::
+   :show-when: fam=radeon
+
+   .. matrix-row::
+      :show-when: gfx=1100
+
+      .. matrix-cell:: AMD Radeon RX 7000 Series
+         :header:
+
+      .. matrix-cell::
+
+         `Radeon RX 7900 XTX <https://www.amd.com/en/products/graphics/desktops/radeon/7000-series/amd-radeon-rx-7900xtx.html>`__
+
+         `Radeon RX 7900 XT <https://www.amd.com/en/products/graphics/desktops/radeon/7000-series/amd-radeon-rx-7900xt.html>`__
+
+         `Radeon RX 7900 GRE <https://www.amd.com/en/products/graphics/desktops/radeon/7000-series/amd-radeon-rx-7900-gre.html>`__
+
+   .. matrix-row::
+      :show-when: gfx=1101
+
+      .. matrix-cell:: AMD Radeon RX 7000 Series
+         :header:
+
+      .. matrix-cell::
+
+         `Radeon RX 7800 XT <https://www.amd.com/en/products/graphics/desktops/radeon/7000-series/amd-radeon-rx-7800-xt.html>`__
+
+         `Radeon RX 7700 XT <https://www.amd.com/en/products/graphics/desktops/radeon/7000-series/amd-radeon-rx-7700-xt.html>`__
+
+   .. matrix-row::
+
+      .. matrix-cell:: Architecture
+         :header:
+
+      .. matrix-cell:: RDNA 3
+         :show-when: gfx=1101 gfx=1100
+
+   .. matrix-row::
+
+      .. matrix-cell:: LLVM target
+         :header:
+
+      .. matrix-cell:: gfx1100
+         :show-when: gfx=1100
+
+      .. matrix-cell:: gfx1101
+         :show-when: gfx=1101
+
+   .. matrix-row::
+      :show-when: os=ubuntu
+
+      .. matrix-cell:: Supported Ubuntu versions
+         :header:
+
+      .. matrix-cell::
+
+         24.04.3 (GA kernel: 6.8)
+
+         22.04.5 (GA kernel: 5.15)
+
+   .. matrix-row::
+      :show-when: os=rhel
+
+      .. matrix-cell:: Supported RHEL versions
+         :header:
+
+      .. matrix-cell:: 10.1, 10.0
+
+   .. matrix-row::
+      :show-when: os=windows
+
+      .. matrix-cell:: Supported Windows version
+         :header:
+
+      .. matrix-cell:: Windows 11 25H2
+
+   .. matrix-row::
+
+      .. matrix-cell:: Supported AMD GPU Driver (amdgpu) versions
+         :header:
+
+      .. matrix-cell:: 
+
+         `30.20.0 <https://instinct.docs.amd.com/projects/amdgpu-docs/en/docs-30.20.0/>`__,
+         `30.10.2 <https://instinct.docs.amd.com/projects/amdgpu-docs/en/docs-30.10.2/>`__,
+         `30.10.1 <https://instinct.docs.amd.com/projects/amdgpu-docs/en/docs-30.10.1/>`__,
+         `30.10.0 <https://instinct.docs.amd.com/projects/amdgpu-docs/en/docs-30.10/>`__
+
+   .. matrix-row::
+      :show-when: os=ubuntu os=rhel os=sles
+
+      .. matrix-cell:: Supported Radeon Software for Linux version
+         :header:
+
+      .. matrix-cell::
+
+         `25.30.1 <https://www.amd.com/en/support/download/linux-drivers.html#linux-for-radeon-pro>`__
+
+   .. matrix-row::
+      :show-when: os=windows
+
+      .. matrix-cell:: Supported Adrenalin Driver version
+         :header:
+
+      .. matrix-cell::
+
+         `25.11.1 <https://www.amd.com/en/resources/support-articles/release-notes/RN-RAD-WIN-25-11-1.html>`__
+         (generally recommended)
+
+         `25.20.01.17 <https://www.amd.com/en/resources/support-articles/release-notes/RN-AMDGPU-WINDOWS-PYTORCH-7-1-1.html>`__
+         (recommended for ComfyUI)
+

docs/compatibility/includes/system-ryzen.rst

@@ -0,0 +1,103 @@
+.. matrix::
+   :show-when: fam=ryzen
+
+   .. matrix-row::
+      :show-when: gfx=1151
+
+      .. matrix-cell:: AMD Ryzen AI Max PRO 300 Series
+         :header:
+
+      .. matrix-cell::
+
+         `Ryzen AI Max+ PRO 395 <https://www.amd.com/en/products/processors/laptop/ryzen-pro/ai-max-pro-300-series/amd-ryzen-ai-max-plus-pro-395.html>`__
+
+         `Ryzen AI Max PRO 390 <https://www.amd.com/en/products/processors/laptop/ryzen-pro/ai-max-pro-300-series/amd-ryzen-ai-max-pro-390.html>`__
+
+         `Ryzen AI Max PRO 385 <https://www.amd.com/en/products/processors/laptop/ryzen-pro/ai-max-pro-300-series/amd-ryzen-ai-max-pro-385.html>`__
+
+         `Ryzen AI Max PRO 380 <https://www.amd.com/en/products/processors/laptop/ryzen-pro/ai-max-pro-300-series/amd-ryzen-ai-max-pro-380.html>`__
+
+   .. matrix-row::
+      :show-when: gfx=1151
+
+      .. matrix-cell:: AMD Ryzen AI Max 300 Series
+         :header:
+
+      .. matrix-cell::
+
+         `Ryzen AI Max+ 395 <https://www.amd.com/en/products/processors/laptop/ryzen/ai-300-series/amd-ryzen-ai-max-plus-395.html>`__
+
+         `Ryzen AI Max 390 <https://www.amd.com/en/products/processors/laptop/ryzen/ai-300-series/amd-ryzen-ai-max-390.html>`__
+
+         `Ryzen AI Max 385 <https://www.amd.com/en/products/processors/laptop/ryzen/ai-300-series/amd-ryzen-ai-max-385.html>`__
+
+   .. matrix-row::
+      :show-when: gfx=1150
+
+      .. matrix-cell:: AMD Ryzen AI 300 Series
+         :header:
+
+      .. matrix-cell::
+
+         `Ryzen AI 9 HX 375 <https://www.amd.com/en/products/processors/laptop/ryzen/ai-300-series/amd-ryzen-ai-9-hx-375.html>`__
+
+         `Ryzen AI 9 HX 370 <https://www.amd.com/en/products/processors/laptop/ryzen/ai-300-series/amd-ryzen-ai-9-hx-370.html>`__
+
+         `Ryzen AI 9 365 <https://www.amd.com/en/products/processors/laptop/ryzen/ai-300-series/amd-ryzen-ai-9-365.html>`__
+
+   .. matrix-row::
+
+      .. matrix-cell:: Architecture
+         :header:
+
+      .. matrix-cell:: RDNA 3.5
+
+   .. matrix-row::
+
+      .. matrix-cell:: LLVM target
+         :header:
+
+      .. matrix-cell:: gfx1151
+         :show-when: gfx=1151
+
+      .. matrix-cell:: gfx1150
+         :show-when: gfx=1150
+
+   .. matrix-row::
+      :show-when: os=ubuntu
+
+      .. matrix-cell:: Supported Ubuntu versions
+         :header:
+
+      .. matrix-cell:: 24.04.3 (HWE kernel: 6.14)
+
+   .. matrix-row::
+      :show-when: os=windows
+
+      .. matrix-cell:: Supported Windows version
+         :header:
+
+      .. matrix-cell:: Windows 11 25H2
+
+   .. matrix-row::
+      :show-when: os=ubuntu
+
+      .. matrix-cell:: Supported kernel driver version
+         :header:
+
+      .. matrix-cell:: Inbox kernel driver in Ubuntu 24.04.3
+
+   .. matrix-row::
+      :show-when: os=windows
+
+      .. matrix-cell:: Supported Adrenalin Driver version
+         :header:
+
+      .. matrix-cell::
+
+         `25.11.1 <https://www.amd.com/en/resources/support-articles/release-notes/RN-RAD-WIN-25-11-1.html>`__
+         (generally recommended)
+
+         `25.20.01.17 <https://www.amd.com/en/resources/support-articles/release-notes/RN-AMDGPU-WINDOWS-PYTORCH-7-1-1.html>`__
+         (recommended for ComfyUI)
+

docs/conf.py

@@ -9,8 +9,8 @@ import shutil
 import sys
 from pathlib import Path
 
-ROCM_VERSION = "7.9.0"
-GA_DATE = "2025-10-20"
+ROCM_VERSION = "7.10.0"
+GA_DATE = "2025-12-11"
 
 DOCS_DIR = Path(__file__).parent.resolve()
 ROOT_DIR = DOCS_DIR.parent
@@ -102,46 +102,40 @@ html_baseurl = os.environ.get("READTHEDOCS_CANONICAL_URL", "rocm.docs.amd.com")
 html_context = {}
 
 # configurations for PDF output by Read the Docs
-project = "ROCm Documentation"
+project = "ROCm documentation"
 project_path = str(DOCS_DIR).replace("\\", "/")
 author = "Advanced Micro Devices, Inc."
 copyright = "Copyright (c) %Y Advanced Micro Devices, Inc. All rights reserved."
 version = ROCM_VERSION
 release = ROCM_VERSION
-setting_all_article_info = True
+setting_all_article_info = False
 all_article_info_os = ["linux", "windows"]
 all_article_info_author = ""
 
 # pages with specific settings
 article_pages = [
-    {"file": "about/release-notes", "date": GA_DATE},
+    {"file": "about/release-notes", "date": GA_DATE, "os": ["linux", "windows"]},
 ]
 
 external_toc_path = "./sphinx/_toc.yml"
 
 # Register Sphinx extensions and static assets
 sys.path.append(str(DOCS_DIR / "extension"))
-
-# html_static_path = ["sphinx/static/css", "extension/how-to/rocm-for-ai/inference"]
-# html_css_files = [
-    # "rocm_custom.css",
-    # "rocm_rn.css",
-    # "dynamic_picker.css",
-    # "vllm-benchmark.css",
-# ]
-templates_path = ["extension/rocm_docs_custom/templates", "extension/templates"]
-
 extensions = [
     "rocm_docs",
     "rocm_docs_custom.selector",
-    "rocm_docs_custom.table",
+    "rocm_docs_custom.matrix",
     "rocm_docs_custom.icon",
+    # "sphinxcontrib.datatemplates",
     # "sphinx_reredirects",
     # "sphinx_sitemap",
-    # "sphinxcontrib.datatemplates",
     # "version-ref",
     # "csv-to-list-table",
 ]
+templates_path = ["extension/rocm_docs_custom/templates"]
+
+html_static_path = ["sphinx/static"]
+html_js_files = ["setup-toc-install-headings.js"]
 
 # compatibility_matrix_file = str(
 #     DOCS_DIR / "compatibility/compatibility-matrix-historical-6.0.csv"
@@ -150,17 +144,33 @@ extensions = [
 external_projects_current_project = "rocm"
 html_theme = "rocm_docs_theme"
 html_theme_options = {
-    "flavor": "rocm-docs-home",
+    "announcement": f"This is ROCm {ROCM_VERSION} technology preview release documentation. For the latest production stream release, refer to <a id='rocm-banner' href='https://rocm.docs.amd.com/en/latest/'>ROCm documentation</a>.",
+    "flavor": "generic",
+    "header_title": f"ROCm™ {ROCM_VERSION} Preview",
+    "header_link": f"https://rocm.docs.amd.com/en/{ROCM_VERSION}-preview/index.html",
+    "version_list_link": f"https://rocm.docs.amd.com/en/{ROCM_VERSION}-preview/release/versions.html",
+    "nav_secondary_items": {
+        "GitHub": "https://github.com/ROCm/ROCm",
+        "Community": "https://github.com/ROCm/ROCm/discussions",
+        "Blogs": "https://rocm.blogs.amd.com/",
+        "Instinct™ Docs": "https://instinct.docs.amd.com/",
+        "Support": "https://github.com/ROCm/ROCm/issues/new/choose",
+    },
     "link_main_doc": False,
     "secondary_sidebar_items": {
         "**": ["page-toc"],
+        "compatibility/compatibility-matrix": ["selector-toc2"],
         "install/rocm": ["selector-toc2"],
-        "install/compatibility-matrix": ["selector-toc2"],
-    }
+        "rocm-for-ai/pytorch-comfyui": ["selector-toc2"],
+    },
 }
 html_title = f"AMD ROCm {ROCM_VERSION} preview"
 
 numfig = False
+rst_prolog = f"""
+.. |ROCM_VERSION| replace:: {ROCM_VERSION}
+"""
+
 suppress_warnings = ["autosectionlabel.*"]
 
 # html_baseurl = os.environ.get("READTHEDOCS_CANONICAL_URL", "https://rocm-stg.amd.com/")

docs/data/how-to/rocm-for-ai/training/jax-maxtext-benchmark-models.yaml

@@ -1,16 +1,47 @@
 dockers:
-  - pull_tag: rocm/jax-training:maxtext-v25.9
+  - pull_tag: rocm/jax-training:maxtext-v25.7-jax060
     docker_hub_url: https://hub.docker.com/layers/rocm/jax-training/maxtext-v25.7/images/sha256-45f4c727d4019a63fc47313d3a5f5a5105569539294ddfd2d742218212ae9025
     components:
-      ROCm: 7.0.0
-      JAX: 0.6.2
-      Python: 3.10.18
-      Transformer Engine: 2.2.0.dev0+c91bac54
+      ROCm: 6.4.1
+      JAX: 0.6.0
+      Python: 3.10.12
+      Transformer Engine: 2.1.0+90d703dd
+      hipBLASLt: 1.1.0-499ece1c21
+  - pull_tag: rocm/jax-training:maxtext-v25.7
+    docker_hub_url: https://hub.docker.com/layers/rocm/jax-training/maxtext-v25.7/images/sha256-45f4c727d4019a63fc47313d3a5f5a5105569539294ddfd2d742218212ae9025
+    components:
+      ROCm: 6.4.1
+      JAX: 0.5.0
+      Python: 3.10.12
+      Transformer Engine: 2.1.0+90d703dd
       hipBLASLt: 1.x.x
 model_groups:
   - group: Meta Llama
     tag: llama
     models:
+      - model: Llama 3.3 70B
+        mad_tag: jax_maxtext_train_llama-3.3-70b
+        model_repo: Llama-3.3-70B
+        precision: bf16
+        doc_options: ["single-node"]
+      - model: Llama 3.1 8B
+        mad_tag: jax_maxtext_train_llama-3.1-8b
+        model_repo: Llama-3.1-8B
+        precision: bf16
+        doc_options: ["single-node"]
+      - model: Llama 3.1 70B
+        mad_tag: jax_maxtext_train_llama-3.1-70b
+        model_repo: Llama-3.1-70B
+        precision: bf16
+        doc_options: ["single-node"]
+      - model: Llama 3 8B
+        mad_tag: jax_maxtext_train_llama-3-8b
+        multinode_training_script: llama3_8b_multinode.sh
+        doc_options: ["multi-node"]
+      - model: Llama 3 70B
+        mad_tag: jax_maxtext_train_llama-3-70b
+        multinode_training_script: llama3_70b_multinode.sh
+        doc_options: ["multi-node"]
       - model: Llama 2 7B
         mad_tag: jax_maxtext_train_llama-2-7b
         model_repo: Llama-2-7B
@@ -23,29 +54,6 @@ model_groups:
         precision: bf16
         multinode_training_script: llama2_70b_multinode.sh
         doc_options: ["single-node", "multi-node"]
-      - model: Llama 3 8B (multi-node)
-        mad_tag: jax_maxtext_train_llama-3-8b
-        multinode_training_script: llama3_8b_multinode.sh
-        doc_options: ["multi-node"]
-      - model: Llama 3 70B (multi-node)
-        mad_tag: jax_maxtext_train_llama-3-70b
-        multinode_training_script: llama3_70b_multinode.sh
-        doc_options: ["multi-node"]
-      - model: Llama 3.1 8B
-        mad_tag: jax_maxtext_train_llama-3.1-8b
-        model_repo: Llama-3.1-8B
-        precision: bf16
-        doc_options: ["single-node"]
-      - model: Llama 3.1 70B
-        mad_tag: jax_maxtext_train_llama-3.1-70b
-        model_repo: Llama-3.1-70B
-        precision: bf16
-        doc_options: ["single-node"]
-      - model: Llama 3.3 70B
-        mad_tag: jax_maxtext_train_llama-3.3-70b
-        model_repo: Llama-3.3-70B
-        precision: bf16
-        doc_options: ["single-node"]
   - group: DeepSeek
     tag: deepseek
     models:

docs/data/how-to/rocm-for-ai/training/megatron-lm-benchmark-models.yaml

@@ -1,21 +1,14 @@
 dockers:
-  MI355X and MI350X:
-    pull_tag: rocm/megatron-lm:v25.9_gfx950
-    docker_hub_url: https://hub.docker.com/layers/rocm/megatron-lm/v25.9_gfx950/images/sha256-1a198be32f49efd66d0ff82066b44bd99b3e6b04c8e0e9b36b2c481e13bff7b6
-    components: &docker_components
-      ROCm: 7.0.0
-      Primus: aab4234
-      PyTorch: 2.9.0.dev20250821+rocm7.0.0.lw.git125803b7
+  - pull_tag: rocm/megatron-lm:v25.8_py310
+    docker_hub_url: https://hub.docker.com/layers/rocm/megatron-lm/v25.8_py310/images/sha256-50fc824361054e445e86d5d88d5f58817f61f8ec83ad4a7e43ea38bbc4a142c0
+    components:
+      ROCm: 6.4.3
+      PyTorch: 2.8.0a0+gitd06a406
       Python: "3.10"
       Transformer Engine: 2.2.0.dev0+54dd2bdc
-      Flash Attention: 2.8.3
-      hipBLASLt: 911283acd1
-      Triton: 3.4.0+rocm7.0.0.git56765e8c
-      RCCL: 2.26.6
-  MI325X and MI300X:
-    pull_tag: rocm/megatron-lm:v25.9_gfx942
-    docker_hub_url: https://hub.docker.com/layers/rocm/megatron-lm/v25.9_gfx942/images/sha256-df6ab8f45b4b9ceb100fb24e19b2019a364e351ee3b324dbe54466a1d67f8357
-    components: *docker_components
+      hipBLASLt: d1b517fc7a
+      Triton: 3.3.0
+      RCCL: 2.22.3
 model_groups:
   - group: Meta Llama
     tag: llama
@@ -26,6 +19,8 @@ model_groups:
         mad_tag: pyt_megatron_lm_train_llama-3.1-8b
       - model: Llama 3.1 70B
         mad_tag: pyt_megatron_lm_train_llama-3.1-70b
+      - model: Llama 3.1 70B (proxy)
+        mad_tag: pyt_megatron_lm_train_llama-3.1-70b-proxy
       - model: Llama 2 7B
         mad_tag: pyt_megatron_lm_train_llama-2-7b
       - model: Llama 2 70B

docs/data/how-to/rocm-for-ai/training/previous-versions/jax-maxtext-v25.7-benchmark-models.yaml

@@ -1,72 +0,0 @@
-dockers:
-  - pull_tag: rocm/jax-training:maxtext-v25.7-jax060
-    docker_hub_url: https://hub.docker.com/layers/rocm/jax-training/maxtext-v25.7/images/sha256-45f4c727d4019a63fc47313d3a5f5a5105569539294ddfd2d742218212ae9025
-    components:
-      ROCm: 6.4.1
-      JAX: 0.6.0
-      Python: 3.10.12
-      Transformer Engine: 2.1.0+90d703dd
-      hipBLASLt: 1.1.0-499ece1c21
-  - pull_tag: rocm/jax-training:maxtext-v25.7
-    docker_hub_url: https://hub.docker.com/layers/rocm/jax-training/maxtext-v25.7/images/sha256-45f4c727d4019a63fc47313d3a5f5a5105569539294ddfd2d742218212ae9025
-    components:
-      ROCm: 6.4.1
-      JAX: 0.5.0
-      Python: 3.10.12
-      Transformer Engine: 2.1.0+90d703dd
-      hipBLASLt: 1.x.x
-model_groups:
-  - group: Meta Llama
-    tag: llama
-    models:
-      - model: Llama 3.3 70B
-        mad_tag: jax_maxtext_train_llama-3.3-70b
-        model_repo: Llama-3.3-70B
-        precision: bf16
-        doc_options: ["single-node"]
-      - model: Llama 3.1 8B
-        mad_tag: jax_maxtext_train_llama-3.1-8b
-        model_repo: Llama-3.1-8B
-        precision: bf16
-        doc_options: ["single-node"]
-      - model: Llama 3.1 70B
-        mad_tag: jax_maxtext_train_llama-3.1-70b
-        model_repo: Llama-3.1-70B
-        precision: bf16
-        doc_options: ["single-node"]
-      - model: Llama 3 8B
-        mad_tag: jax_maxtext_train_llama-3-8b
-        multinode_training_script: llama3_8b_multinode.sh
-        doc_options: ["multi-node"]
-      - model: Llama 3 70B
-        mad_tag: jax_maxtext_train_llama-3-70b
-        multinode_training_script: llama3_70b_multinode.sh
-        doc_options: ["multi-node"]
-      - model: Llama 2 7B
-        mad_tag: jax_maxtext_train_llama-2-7b
-        model_repo: Llama-2-7B
-        precision: bf16
-        multinode_training_script: llama2_7b_multinode.sh
-        doc_options: ["single-node", "multi-node"]
-      - model: Llama 2 70B
-        mad_tag: jax_maxtext_train_llama-2-70b
-        model_repo: Llama-2-70B
-        precision: bf16
-        multinode_training_script: llama2_70b_multinode.sh
-        doc_options: ["single-node", "multi-node"]
-  - group: DeepSeek
-    tag: deepseek
-    models:
-      - model: DeepSeek-V2-Lite (16B)
-        mad_tag: jax_maxtext_train_deepseek-v2-lite-16b
-        model_repo: DeepSeek-V2-lite
-        precision: bf16
-        doc_options: ["single-node"]
-  - group: Mistral AI
-    tag: mistral
-    models:
-      - model: Mixtral 8x7B
-        mad_tag: jax_maxtext_train_mixtral-8x7b
-        model_repo: Mixtral-8x7B
-        precision: bf16
-        doc_options: ["single-node"]

docs/data/how-to/rocm-for-ai/training/previous-versions/megatron-lm-v25.8-benchmark-models.yaml

@@ -1,48 +0,0 @@
-dockers:
-  - pull_tag: rocm/megatron-lm:v25.8_py310
-    docker_hub_url: https://hub.docker.com/layers/rocm/megatron-lm/v25.8_py310/images/sha256-50fc824361054e445e86d5d88d5f58817f61f8ec83ad4a7e43ea38bbc4a142c0
-    components:
-      ROCm: 6.4.3
-      PyTorch: 2.8.0a0+gitd06a406
-      Python: "3.10"
-      Transformer Engine: 2.2.0.dev0+54dd2bdc
-      hipBLASLt: d1b517fc7a
-      Triton: 3.3.0
-      RCCL: 2.22.3
-model_groups:
-  - group: Meta Llama
-    tag: llama
-    models:
-      - model: Llama 3.3 70B
-        mad_tag: pyt_megatron_lm_train_llama-3.3-70b
-      - model: Llama 3.1 8B
-        mad_tag: pyt_megatron_lm_train_llama-3.1-8b
-      - model: Llama 3.1 70B
-        mad_tag: pyt_megatron_lm_train_llama-3.1-70b
-      - model: Llama 3.1 70B (proxy)
-        mad_tag: pyt_megatron_lm_train_llama-3.1-70b-proxy
-      - model: Llama 2 7B
-        mad_tag: pyt_megatron_lm_train_llama-2-7b
-      - model: Llama 2 70B
-        mad_tag: pyt_megatron_lm_train_llama-2-70b
-  - group: DeepSeek
-    tag: deepseek
-    models:
-      - model: DeepSeek-V3 (proxy)
-        mad_tag: pyt_megatron_lm_train_deepseek-v3-proxy
-      - model: DeepSeek-V2-Lite
-        mad_tag: pyt_megatron_lm_train_deepseek-v2-lite-16b
-  - group: Mistral AI
-    tag: mistral
-    models:
-      - model: Mixtral 8x7B
-        mad_tag: pyt_megatron_lm_train_mixtral-8x7b
-      - model: Mixtral 8x22B (proxy)
-        mad_tag: pyt_megatron_lm_train_mixtral-8x22b-proxy
-  - group: Qwen
-    tag: qwen
-    models:
-      - model: Qwen 2.5 7B
-        mad_tag: pyt_megatron_lm_train_qwen2.5-7b
-      - model: Qwen 2.5 72B
-        mad_tag: pyt_megatron_lm_train_qwen2.5-72b

docs/data/how-to/rocm-for-ai/training/previous-versions/primus-megatron-v25.8-benchmark-models.yaml

@@ -1,58 +0,0 @@
-dockers:
-  - pull_tag: rocm/megatron-lm:v25.8_py310
-    docker_hub_url: https://hub.docker.com/layers/rocm/megatron-lm/v25.8_py310/images/sha256-50fc824361054e445e86d5d88d5f58817f61f8ec83ad4a7e43ea38bbc4a142c0
-    components:
-      ROCm: 6.4.3
-      Primus: 927a717
-      PyTorch: 2.8.0a0+gitd06a406
-      Python: "3.10"
-      Transformer Engine: 2.2.0.dev0+54dd2bdc
-      hipBLASLt: d1b517fc7a
-      Triton: 3.3.0
-      RCCL: 2.22.3
-model_groups:
-  - group: Meta Llama
-    tag: llama
-    models:
-      - model: Llama 3.3 70B
-        mad_tag: primus_pyt_megatron_lm_train_llama-3.3-70b
-        config_name: llama3.3_70B-pretrain.yaml
-      - model: Llama 3.1 70B
-        mad_tag: primus_pyt_megatron_lm_train_llama-3.1-70b
-        config_name: llama3.1_70B-pretrain.yaml
-      - model: Llama 3.1 8B
-        mad_tag: primus_pyt_megatron_lm_train_llama-3.1-8b
-        config_name: llama3.1_8B-pretrain.yaml
-      - model: Llama 2 7B
-        mad_tag: primus_pyt_megatron_lm_train_llama-2-7b
-        config_name: llama2_7B-pretrain.yaml
-      - model: Llama 2 70B
-        mad_tag: primus_pyt_megatron_lm_train_llama-2-70b
-        config_name: llama2_70B-pretrain.yaml
-  - group: DeepSeek
-    tag: deepseek
-    models:
-      - model: DeepSeek-V3 (proxy)
-        mad_tag: primus_pyt_megatron_lm_train_deepseek-v3-proxy
-        config_name: deepseek_v3-pretrain.yaml
-      - model: DeepSeek-V2-Lite
-        mad_tag: primus_pyt_megatron_lm_train_deepseek-v2-lite-16b
-        config_name: deepseek_v2_lite-pretrain.yaml
-  - group: Mistral AI
-    tag: mistral
-    models:
-      - model: Mixtral 8x7B
-        mad_tag: primus_pyt_megatron_lm_train_mixtral-8x7b
-        config_name: mixtral_8x7B_v0.1-pretrain.yaml
-      - model: Mixtral 8x22B (proxy)
-        mad_tag: primus_pyt_megatron_lm_train_mixtral-8x22b-proxy
-        config_name: mixtral_8x22B_v0.1-pretrain.yaml
-  - group: Qwen
-    tag: qwen
-    models:
-      - model: Qwen 2.5 7B
-        mad_tag: primus_pyt_megatron_lm_train_qwen2.5-7b
-        config_name: primus_qwen2.5_7B-pretrain.yaml
-      - model: Qwen 2.5 72B
-        mad_tag: primus_pyt_megatron_lm_train_qwen2.5-72b
-        config_name: qwen2.5_72B-pretrain.yaml

docs/data/how-to/rocm-for-ai/training/previous-versions/primus-pytorch-v25.8-benchmark-models.yaml

@@ -1,24 +0,0 @@
-dockers:
-  - pull_tag: rocm/pytorch-training:v25.8
-    docker_hub_url: https://hub.docker.com/layers/rocm/pytorch-training/v25.8/images/sha256-5082ae01d73fec6972b0d84e5dad78c0926820dcf3c19f301d6c8eb892e573c5
-    components:
-      ROCm: 6.4.3
-      PyTorch: 2.8.0a0+gitd06a406
-      Python: 3.10.18
-      Transformer Engine: 2.2.0.dev0+a1e66aae
-      Flash Attention: 3.0.0.post1
-      hipBLASLt: 1.1.0-d1b517fc7a
-model_groups:
-  - group: Meta Llama
-    tag: llama
-    models:
-    - model: Llama 3.1 8B
-      mad_tag: primus_pyt_train_llama-3.1-8b
-      model_repo: Llama-3.1-8B
-      url: https://huggingface.co/meta-llama/Llama-3.1-8B
-      precision: BF16
-    - model: Llama 3.1 70B
-      mad_tag: primus_pyt_train_llama-3.1-70b
-      model_repo: Llama-3.1-70B
-      url: https://huggingface.co/meta-llama/Llama-3.1-70B
-      precision: BF16

docs/data/how-to/rocm-for-ai/training/previous-versions/pytorch-training-v25.8-benchmark-models.yaml

@@ -1,178 +0,0 @@
-dockers:
-  - pull_tag: rocm/pytorch-training:v25.8
-    docker_hub_url: https://hub.docker.com/layers/rocm/pytorch-training/v25.8/images/sha256-5082ae01d73fec6972b0d84e5dad78c0926820dcf3c19f301d6c8eb892e573c5
-    components:
-      ROCm: 6.4.3
-      PyTorch: 2.8.0a0+gitd06a406
-      Python: 3.10.18
-      Transformer Engine: 2.2.0.dev0+a1e66aae
-      Flash Attention: 3.0.0.post1
-      hipBLASLt: 1.1.0-d1b517fc7a
-model_groups:
-  - group: Meta Llama
-    tag: llama
-    models:
-    - model: Llama 4 Scout 17B-16E
-      mad_tag: pyt_train_llama-4-scout-17b-16e
-      model_repo: Llama-4-17B_16E
-      url: https://huggingface.co/meta-llama/Llama-4-Scout-17B-16E
-      precision: BF16
-      training_modes: [finetune_fw, finetune_lora]
-    - model: Llama 3.3 70B
-      mad_tag: pyt_train_llama-3.3-70b
-      model_repo: Llama-3.3-70B
-      url: https://huggingface.co/meta-llama/Llama-3.3-70B-Instruct
-      precision: BF16
-      training_modes: [finetune_fw, finetune_lora, finetune_qlora]
-    - model: Llama 3.2 1B
-      mad_tag: pyt_train_llama-3.2-1b
-      model_repo: Llama-3.2-1B
-      url: https://huggingface.co/meta-llama/Llama-3.2-1B
-      precision: BF16
-      training_modes: [finetune_fw, finetune_lora]
-    - model: Llama 3.2 3B
-      mad_tag: pyt_train_llama-3.2-3b
-      model_repo: Llama-3.2-3B
-      url: https://huggingface.co/meta-llama/Llama-3.2-3B
-      precision: BF16
-      training_modes: [finetune_fw, finetune_lora]
-    - model: Llama 3.2 Vision 11B
-      mad_tag: pyt_train_llama-3.2-vision-11b
-      model_repo: Llama-3.2-Vision-11B
-      url: https://huggingface.co/meta-llama/Llama-3.2-11B-Vision
-      precision: BF16
-      training_modes: [finetune_fw]
-    - model: Llama 3.2 Vision 90B
-      mad_tag: pyt_train_llama-3.2-vision-90b
-      model_repo: Llama-3.2-Vision-90B
-      url: https://huggingface.co/meta-llama/Llama-3.2-90B-Vision
-      precision: BF16
-      training_modes: [finetune_fw]
-    - model: Llama 3.1 8B
-      mad_tag: pyt_train_llama-3.1-8b
-      model_repo: Llama-3.1-8B
-      url: https://huggingface.co/meta-llama/Llama-3.1-8B
-      precision: BF16
-      training_modes: [pretrain, finetune_fw, finetune_lora, HF_pretrain]
-    - model: Llama 3.1 70B
-      mad_tag: pyt_train_llama-3.1-70b
-      model_repo: Llama-3.1-70B
-      url: https://huggingface.co/meta-llama/Llama-3.1-70B-Instruct
-      precision: BF16
-      training_modes: [pretrain, finetune_fw, finetune_lora]
-    - model: Llama 3.1 405B
-      mad_tag: pyt_train_llama-3.1-405b
-      model_repo: Llama-3.1-405B
-      url: https://huggingface.co/meta-llama/Llama-3.1-405B
-      precision: BF16
-      training_modes: [finetune_qlora]
-    - model: Llama 3 8B
-      mad_tag: pyt_train_llama-3-8b
-      model_repo: Llama-3-8B
-      url: https://huggingface.co/meta-llama/Meta-Llama-3-8B
-      precision: BF16
-      training_modes: [finetune_fw, finetune_lora]
-    - model: Llama 3 70B
-      mad_tag: pyt_train_llama-3-70b
-      model_repo: Llama-3-70B
-      url: https://huggingface.co/meta-llama/Meta-Llama-3-70B
-      precision: BF16
-      training_modes: [finetune_fw, finetune_lora]
-    - model: Llama 2 7B
-      mad_tag: pyt_train_llama-2-7b
-      model_repo: Llama-2-7B
-      url: https://github.com/meta-llama/llama-models/tree/main/models/llama2
-      precision: BF16
-      training_modes: [finetune_fw, finetune_lora, finetune_qlora]
-    - model: Llama 2 13B
-      mad_tag: pyt_train_llama-2-13b
-      model_repo: Llama-2-13B
-      url: https://github.com/meta-llama/llama-models/tree/main/models/llama2
-      precision: BF16
-      training_modes: [finetune_fw, finetune_lora]
-    - model: Llama 2 70B
-      mad_tag: pyt_train_llama-2-70b
-      model_repo: Llama-2-70B
-      url: https://github.com/meta-llama/llama-models/tree/main/models/llama2
-      precision: BF16
-      training_modes: [finetune_lora, finetune_qlora]
-  - group: OpenAI
-    tag: openai
-    models:
-    - model: GPT OSS 20B
-      mad_tag: pyt_train_gpt_oss_20b
-      model_repo: GPT-OSS-20B
-      url: https://huggingface.co/openai/gpt-oss-20b
-      precision: BF16
-      training_modes: [HF_finetune_lora]
-    - model: GPT OSS 120B
-      mad_tag: pyt_train_gpt_oss_120b
-      model_repo: GPT-OSS-120B
-      url: https://huggingface.co/openai/gpt-oss-120b
-      precision: BF16
-      training_modes: [HF_finetune_lora]
-  - group: Qwen
-    tag: qwen
-    models:
-    - model: Qwen 3 8B
-      mad_tag: pyt_train_qwen3-8b
-      model_repo: Qwen3-8B
-      url: https://huggingface.co/Qwen/Qwen3-8B
-      precision: BF16
-      training_modes: [finetune_fw, finetune_lora]
-    - model: Qwen 3 32B
-      mad_tag: pyt_train_qwen3-32b
-      model_repo: Qwen3-32
-      url: https://huggingface.co/Qwen/Qwen3-32B
-      precision: BF16
-      training_modes: [finetune_lora]
-    - model: Qwen 2.5 32B
-      mad_tag: pyt_train_qwen2.5-32b
-      model_repo: Qwen2.5-32B
-      url: https://huggingface.co/Qwen/Qwen2.5-32B
-      precision: BF16
-      training_modes: [finetune_lora]
-    - model: Qwen 2.5 72B
-      mad_tag: pyt_train_qwen2.5-72b
-      model_repo: Qwen2.5-72B
-      url: https://huggingface.co/Qwen/Qwen2.5-72B
-      precision: BF16
-      training_modes: [finetune_lora]
-    - model: Qwen 2 1.5B
-      mad_tag: pyt_train_qwen2-1.5b
-      model_repo: Qwen2-1.5B
-      url: https://huggingface.co/Qwen/Qwen2-1.5B
-      precision: BF16
-      training_modes: [finetune_fw, finetune_lora]
-    - model: Qwen 2 7B
-      mad_tag: pyt_train_qwen2-7b
-      model_repo: Qwen2-7B
-      url: https://huggingface.co/Qwen/Qwen2-7B
-      precision: BF16
-      training_modes: [finetune_fw, finetune_lora]
-  - group: Stable Diffusion
-    tag: sd
-    models:
-    - model: Stable Diffusion XL
-      mad_tag: pyt_huggingface_stable_diffusion_xl_2k_lora_finetuning
-      model_repo: SDXL
-      url: https://huggingface.co/stabilityai/stable-diffusion-xl-base-1.0
-      precision: BF16
-      training_modes: [finetune_lora]
-  - group: Flux
-    tag: flux
-    models:
-    - model: FLUX.1-dev
-      mad_tag: pyt_train_flux
-      model_repo: Flux
-      url: https://huggingface.co/black-forest-labs/FLUX.1-dev
-      precision: BF16
-      training_modes: [pretrain]
-  - group: NCF
-    tag: ncf
-    models:
-    - model: NCF
-      mad_tag: pyt_ncf_training
-      model_repo:
-      url: https://github.com/NVIDIA/DeepLearningExamples/tree/master/PyTorch/Recommendation/NCF
-      precision: FP32

docs/data/how-to/rocm-for-ai/training/primus-megatron-benchmark-models.yaml

@@ -1,22 +1,15 @@
 dockers:
-  MI355X and MI350X:
-    pull_tag: rocm/primus:v25.9_gfx950
-    docker_hub_url: https://hub.docker.com/layers/rocm/primus/v25.9_gfx950/images/sha256-1a198be32f49efd66d0ff82066b44bd99b3e6b04c8e0e9b36b2c481e13bff7b6
-    components: &docker_components
-      ROCm: 7.0.0
-      Primus: 0.3.0
-      Primus Turbo: 0.1.1
-      PyTorch: 2.9.0.dev20250821+rocm7.0.0.lw.git125803b7
+  - pull_tag: rocm/megatron-lm:v25.8_py310
+    docker_hub_url: https://hub.docker.com/layers/rocm/megatron-lm/v25.8_py310/images/sha256-50fc824361054e445e86d5d88d5f58817f61f8ec83ad4a7e43ea38bbc4a142c0
+    components:
+      ROCm: 6.4.3
+      Primus: 927a717
+      PyTorch: 2.8.0a0+gitd06a406
       Python: "3.10"
       Transformer Engine: 2.2.0.dev0+54dd2bdc
-      Flash Attention: 2.8.3
-      hipBLASLt: 911283acd1
-      Triton: 3.4.0+rocm7.0.0.git56765e8c
-      RCCL: 2.26.6
-  MI325X and MI300X:
-    pull_tag: rocm/primus:v25.9_gfx942
-    docker_hub_url: https://hub.docker.com/layers/rocm/primus/v25.9_gfx942/images/sha256-df6ab8f45b4b9ceb100fb24e19b2019a364e351ee3b324dbe54466a1d67f8357
-    components: *docker_components
+      hipBLASLt: d1b517fc7a
+      Triton: 3.3.0
+      RCCL: 2.22.3
 model_groups:
   - group: Meta Llama
     tag: llama

docs/data/how-to/rocm-for-ai/training/primus-pytorch-benchmark-models.yaml

@@ -1,39 +1,24 @@
 dockers:
-  MI355X and MI350X:
-    pull_tag: rocm/primus:v25.9_gfx950
-    docker_hub_url: https://hub.docker.com/layers/rocm/primus/v25.9_gfx950/images/sha256-1a198be32f49efd66d0ff82066b44bd99b3e6b04c8e0e9b36b2c481e13bff7b6
-    components: &docker_components
-      ROCm: 7.0.0
-      Primus: 0.3.0
-      Primus Turbo: 0.1.1
-      PyTorch: 2.9.0.dev20250821+rocm7.0.0.lw.git125803b7
-      Python: "3.10"
-      Transformer Engine: 2.2.0.dev0+54dd2bdc
-      Flash Attention: 2.8.3
-      hipBLASLt: 911283acd1
-      Triton: 3.4.0+rocm7.0.0.git56765e8c
-      RCCL: 2.26.6
-  MI325X and MI300X:
-    pull_tag: rocm/primus:v25.9_gfx942
-    docker_hub_url: https://hub.docker.com/layers/rocm/primus/v25.9_gfx942/images/sha256-df6ab8f45b4b9ceb100fb24e19b2019a364e351ee3b324dbe54466a1d67f8357
-    components: *docker_components
+  - pull_tag: rocm/pytorch-training:v25.8
+    docker_hub_url: https://hub.docker.com/layers/rocm/pytorch-training/v25.8/images/sha256-5082ae01d73fec6972b0d84e5dad78c0926820dcf3c19f301d6c8eb892e573c5
+    components:
+      ROCm: 6.4.3
+      PyTorch: 2.8.0a0+gitd06a406
+      Python: 3.10.18
+      Transformer Engine: 2.2.0.dev0+a1e66aae
+      Flash Attention: 3.0.0.post1
+      hipBLASLt: 1.1.0-d1b517fc7a
 model_groups:
   - group: Meta Llama
     tag: llama
     models:
     - model: Llama 3.1 8B
       mad_tag: primus_pyt_train_llama-3.1-8b
-      model_repo: meta-llama/Llama-3.1-8B
+      model_repo: Llama-3.1-8B
       url: https://huggingface.co/meta-llama/Llama-3.1-8B
       precision: BF16
-      config_file:
-        bf16: "./llama3_8b_fsdp_bf16.toml"
-        fp8: "./llama3_8b_fsdp_fp8.toml"
     - model: Llama 3.1 70B
       mad_tag: primus_pyt_train_llama-3.1-70b
-      model_repo: meta-llama/Llama-3.1-70B
+      model_repo: Llama-3.1-70B
       url: https://huggingface.co/meta-llama/Llama-3.1-70B
       precision: BF16
-      config_file:
-        bf16: "./llama3_70b_fsdp_bf16.toml"
-        fp8: "./llama3_70b_fsdp_fp8.toml"

docs/data/how-to/rocm-for-ai/training/pytorch-training-benchmark-models.yaml

@@ -1,21 +1,13 @@
 dockers:
-  MI355X and MI350X:
-    pull_tag: rocm/pytorch-training:v25.9_gfx950
-    docker_hub_url: https://hub.docker.com/layers/rocm/pytorch-training/v25.9_gfx950/images/sha256-1a198be32f49efd66d0ff82066b44bd99b3e6b04c8e0e9b36b2c481e13bff7b6
-    components: &docker_components
-      ROCm: 7.0.0
-      Primus: aab4234
-      PyTorch: 2.9.0.dev20250821+rocm7.0.0.lw.git125803b7
-      Python: "3.10"
-      Transformer Engine: 2.2.0.dev0+54dd2bdc
-      Flash Attention: 2.8.3
-      hipBLASLt: 911283acd1
-      Triton: 3.4.0+rocm7.0.0.git56765e8c
-      RCCL: 2.26.6
-  MI325X and MI300X:
-    pull_tag: rocm/pytorch-training:v25.9_gfx942
-    docker_hub_url: https://hub.docker.com/layers/rocm/pytorch-training/v25.9_gfx942/images/sha256-df6ab8f45b4b9ceb100fb24e19b2019a364e351ee3b324dbe54466a1d67f8357
-    components: *docker_components
+  - pull_tag: rocm/pytorch-training:v25.8
+    docker_hub_url: https://hub.docker.com/layers/rocm/pytorch-training/v25.8/images/sha256-5082ae01d73fec6972b0d84e5dad78c0926820dcf3c19f301d6c8eb892e573c5
+    components:
+      ROCm: 6.4.3
+      PyTorch: 2.8.0a0+gitd06a406
+      Python: 3.10.18
+      Transformer Engine: 2.2.0.dev0+a1e66aae
+      Flash Attention: 3.0.0.post1
+      hipBLASLt: 1.1.0-d1b517fc7a
 model_groups:
   - group: Meta Llama
     tag: llama
@@ -166,7 +158,7 @@ model_groups:
       model_repo: SDXL
       url: https://huggingface.co/stabilityai/stable-diffusion-xl-base-1.0
       precision: BF16
-      training_modes: [posttrain-p]
+      training_modes: [finetune_lora]
   - group: Flux
     tag: flux
     models:
@@ -175,7 +167,7 @@ model_groups:
       model_repo: Flux
       url: https://huggingface.co/black-forest-labs/FLUX.1-dev
       precision: BF16
-      training_modes: [posttrain-p]
+      training_modes: [pretrain]
   - group: NCF
     tag: ncf
     models:

docs/data/reference/gpu-atomics-operation/cas-atomics_nopcie_instinct.csv

@@ -1,4 +1,4 @@
-Atomic,MI100,MI200 PCIe,MI200 A+A,MI300X Series,MI300A,MI350X Series
+Atomic,MI100,MI200 PCIe,MI200 A+A,MI300X series,MI300A,MI350X series
 32 bit atomicAdd,✅ CAS,✅ CAS,✅ CAS,✅ CAS,✅ CAS,✅ CAS
 32 bit atomicSub,✅ CAS,✅ CAS,✅ CAS,✅ CAS,✅ CAS,✅ CAS
 32 bit atomicMin,✅ CAS,✅ CAS,✅ CAS,✅ CAS,✅ CAS,✅ CAS

docs/data/reference/gpu-atomics-operation/cas-atomics_pcie_instinct.csv

@@ -1,4 +1,4 @@
-Atomic,MI100,MI200 PCIe,MI200 A+A,MI300X Series,MI300A,MI350X Series
+Atomic,MI100,MI200 PCIe,MI200 A+A,MI300X series,MI300A,MI350X series
 32 bit atomicAdd,✅ CAS,✅ CAS,✅ CAS,✅ CAS,✅ CAS,✅ CAS
 32 bit atomicSub,✅ CAS,✅ CAS,✅ CAS,✅ CAS,✅ CAS,✅ CAS
 32 bit atomicMin,✅ CAS,✅ CAS,✅ CAS,✅ CAS,✅ CAS,✅ CAS

docs/data/reference/gpu-atomics-operation/hw-atomics_nopcie_instinct.csv

@@ -1,4 +1,4 @@
-Atomic,MI100,MI200 PCIe,MI200 A+A,MI300X Series,MI300A,MI350X Series
+Atomic,MI100,MI200 PCIe,MI200 A+A,MI300X series,MI300A,MI350X series
 32 bit atomicAdd,✅ Native,✅ Native,✅ Native,✅ Native,✅ Native,✅ Native
 32 bit atomicSub,✅ Native,✅ Native,✅ Native,✅ Native,✅ Native,✅ Native
 32 bit atomicMin,✅ Native,✅ Native,✅ Native,✅ Native,✅ Native,✅ Native

docs/data/reference/gpu-atomics-operation/hw-atomics_pcie_instinct.csv

@@ -1,4 +1,4 @@
-Atomic,MI100,MI200 PCIe,MI200 A+A,MI300X Series,MI300A,MI350X Series
+Atomic,MI100,MI200 PCIe,MI200 A+A,MI300X series,MI300A,MI350X series
 32 bit atomicAdd,✅ Native,✅ Native,✅ Native,✅ Native,✅ Native,✅ Native
 32 bit atomicSub,✅ Native,✅ Native,✅ Native,✅ Native,✅ Native,✅ Native
 32 bit atomicMin,✅ Native,✅ Native,✅ Native,✅ Native,✅ Native,✅ Native

docs/extension/rocm_docs_custom/matrix.py

@@ -0,0 +1,289 @@
+from pathlib import Path
+
+from docutils import nodes
+from docutils.parsers.rst import directives
+from sphinx.util.docutils import SphinxDirective
+
+from .utils import kv_to_data_attr, logger
+
+
+class CustomTable(nodes.General, nodes.Element):
+    """Bootstrap-flavoured table container."""
+
+    @staticmethod
+    def visit_html(translator, node):
+        show_when_attr = kv_to_data_attr("show-when", node.get("show-when", ""))
+
+        classes = ["rocm-docs-table", "table"]
+        classes.extend(node.get("classes", []))
+        class_attr = " ".join(classes)
+        table_id = node.get("id") or ""
+
+        attrs = []
+        if table_id:
+            attrs.append(f'id="{table_id}"')
+        attrs.append(f'class="{class_attr}"')
+        if show_when_attr:
+            attrs.append(show_when_attr)
+
+        attrs_str = " ".join(attrs)
+        translator.body.append(f"<!-- start custom-table --><table {attrs_str}>")
+
+        caption = node.get("caption", "")
+        if caption:
+            translator.body.append(f"<caption>{caption}</caption>")
+
+        translator._in_matrix_body = False  # internal state flag
+
+    @staticmethod
+    def depart_html(translator, node):
+        # Close an open <tbody> if present
+        if getattr(translator, "_in_matrix_body", False):
+            translator.body.append("</tbody>")
+            translator._in_matrix_body = False
+
+        translator.body.append("</table><!-- end custom-table -->")
+
+
+class CustomTableDirective(SphinxDirective):
+    """.. matrix:: Optional caption"""
+
+    required_arguments = 0
+    optional_arguments = 1
+    final_argument_whitespace = True
+    has_content = True
+    option_spec = {
+        "id": directives.unchanged,
+        "class": directives.class_option,
+        "show-when": directives.unchanged,
+    }
+
+    def run(self):
+        node = CustomTable()
+        node["caption"] = self.arguments[0] if self.arguments else ""
+        node["id"] = self.options.get("id", "")
+        node["classes"] = self.options.get("class", [])
+        node["show-when"] = self.options.get("show-when", "")
+        self.state.nested_parse(self.content, self.content_offset, node)
+        return [node]
+
+
+class CustomTableHead(nodes.General, nodes.Element):
+    """A table header section (renders <thead>).</thead>"""
+
+    @staticmethod
+    def visit_html(translator, node):
+        translator.body.append("<!-- start table head --><thead>")
+
+    @staticmethod
+    def depart_html(translator, node):
+        translator.body.append("</thead><!-- end table head -->")
+
+
+class CustomTableHeadDirective(SphinxDirective):
+    """.. matrix-head::"""
+
+    has_content = True
+
+    def run(self):
+        node = CustomTableHead()
+        self.state.nested_parse(self.content, self.content_offset, node)
+        return [node]
+
+
+class CustomTableRow(nodes.General, nodes.Element):
+    """A table row (<tr> inside <thead> or <tbody>)."""
+
+    @staticmethod
+    def visit_html(translator, node):
+        # handle automatic <tbody> opening for body rows
+        if not node.get("header-row", False):
+            if not getattr(translator, "_in_matrix_body", False):
+                translator.body.append("<!-- start tbody --><tbody>")
+                translator._in_matrix_body = True
+
+        show_when_attr = kv_to_data_attr("show-when", node.get("show-when", ""))
+        disable_when_attr = kv_to_data_attr(
+            "disable-when", node.get("disable-when", "")
+        )
+
+        classes = " ".join(node.get("classes", []))
+        attrs = []
+        if classes:
+            attrs.append(f'class="{classes}"')
+        if show_when_attr:
+            attrs.append(show_when_attr)
+        if disable_when_attr:
+            attrs.append(disable_when_attr)
+
+        attrs_str = "" if not attrs else " " + " ".join(attrs)
+        translator.body.append(f"<!-- start custom-table row --><tr{attrs_str}>")
+
+    @staticmethod
+    def depart_html(translator, node):
+        translator.body.append("</tr><!-- end custom-table row -->")
+
+
+class CustomTableRowDirective(SphinxDirective):
+    """.. matrix-row::"""
+
+    required_arguments = 0
+    final_argument_whitespace = True
+    has_content = True
+    option_spec = {
+        "class": directives.class_option,
+        "show-when": directives.unchanged,
+        "disable-when": directives.unchanged,
+        "header": directives.flag,
+    }
+
+    def run(self):
+        node = CustomTableRow()
+        node["classes"] = self.options.get("class", [])
+        node["show-when"] = self.options.get("show-when", "")
+        node["disable-when"] = self.options.get("disable-when", "")
+        node["header-row"] = self.options.get("header", False) is not False
+
+        # Parse nested cells
+        self.state.nested_parse(self.content, self.content_offset, node)
+
+        # Inherit header status if inside matrix-head
+        parent_in_head = any(
+            isinstance(p, CustomTableHead)
+            for p in self.state.parent.traverse(include_self=True)
+        )
+        if parent_in_head:
+            node["header-row"] = True
+
+        # Mark all cells as headers if this is a header row
+        if node["header-row"]:
+            for cell in node.findall(CustomTableCell):
+                if "header" not in cell:
+                    cell["header"] = True
+
+        # Sanity check
+        parent = getattr(self.state, "parent", None)
+        if not parent or not any(
+            isinstance(p, (CustomTable, CustomTableHead))
+            for p in parent.traverse(include_self=True)
+        ):
+            logger.warning(
+                "'.. matrix-row::' at line %s should be nested under a '.. matrix::'.",
+                self.lineno,
+                location=(self.env.docname, self.lineno),
+            )
+
+        return [node]
+
+
+class CustomTableCell(nodes.General, nodes.Element):
+    """A table cell (<td> or <th>)."""
+
+    @staticmethod
+    def visit_html(translator, node):
+        is_header = bool(node.get("header", False))
+        tag = "th" if is_header else "td"
+
+        classes = " ".join(node.get("classes", []))
+        colspan = node.get("colspan", 1)
+        rowspan = node.get("rowspan", 1)
+        show_when_attr = kv_to_data_attr("show-when", node.get("show-when", ""))
+
+        attrs = []
+        if classes:
+            attrs.append(f'class="{classes}"')
+        if colspan and colspan > 1:
+            attrs.append(f'colspan="{colspan}"')
+        if rowspan and rowspan > 1:
+            attrs.append(f'rowspan="{rowspan}"')
+        if show_when_attr:
+            attrs.append(show_when_attr)
+
+        attrs_str = "" if not attrs else " " + " ".join(attrs)
+        translator.body.append(f"<{tag}{attrs_str}>")
+
+    @staticmethod
+    def depart_html(translator, node):
+        tag = "th" if node.get("header", False) else "td"
+        translator.body.append(f"</{tag}>")
+
+
+class CustomTableCellDirective(SphinxDirective):
+    """.. matrix-cell::"""
+
+    required_arguments = 0
+    optional_arguments = 1
+    final_argument_whitespace = True
+    has_content = True
+    option_spec = {
+        "header": directives.flag,
+        "class": directives.class_option,
+        "colspan": directives.nonnegative_int,
+        "rowspan": directives.nonnegative_int,
+        "show-when": directives.unchanged,
+    }
+
+    def run(self):
+        label = self.arguments[0] if self.arguments else ""
+        node = CustomTableCell()
+
+        # Explicit :header: always wins
+        explicit_header = self.options.get("header", False) is not False
+
+        # Detect if the parent row (matrix-row) or one of its ancestors
+        # (matrix-head) marks this as a header section.
+        parent_header_row = False
+        parent_node = getattr(self.state, "parent", None)
+        if parent_node:
+            for ancestor in parent_node.traverse(include_self=True):
+                if isinstance(ancestor, CustomTableRow) and ancestor.get("header-row", False):
+                    parent_header_row = True
+                    break
+                if isinstance(ancestor, CustomTableHead):
+                    parent_header_row = True
+                    break
+        node["header"] = explicit_header or parent_header_row
+        node["classes"] = self.options.get("class", [])
+        node["colspan"] = self.options.get("colspan", 1)
+        node["rowspan"] = self.options.get("rowspan", 1)
+        node["show-when"] = self.options.get("show-when", "")
+
+        if self.content:
+            self.state.nested_parse(self.content, self.content_offset, node)
+        elif label:
+            node += nodes.Text(label)
+
+        # Sanity check nesting
+        if not parent_node or not any(
+            isinstance(p, CustomTableRow)
+            for p in parent_node.traverse(include_self=True)
+        ):
+            logger.warning(
+                "'.. matrix-cell::' at line %s should be nested under a '.. matrix-row::' directive",
+                self.lineno,
+                location=(self.env.docname, self.lineno),
+            )
+
+        return [node]
+
+
+def setup(app):
+    app.add_node(CustomTable, html=(CustomTable.visit_html, CustomTable.depart_html))
+    app.add_node(CustomTableHead, html=(CustomTableHead.visit_html, CustomTableHead.depart_html))
+    app.add_node(CustomTableRow, html=(CustomTableRow.visit_html, CustomTableRow.depart_html))
+    app.add_node(CustomTableCell, html=(CustomTableCell.visit_html, CustomTableCell.depart_html))
+
+    app.add_directive("matrix", CustomTableDirective)
+    app.add_directive("matrix-head", CustomTableHeadDirective)
+    app.add_directive("matrix-row", CustomTableRowDirective)
+    app.add_directive("matrix-cell", CustomTableCellDirective)
+
+    static_assets_dir = Path(__file__).parent / "static"
+    app.config.html_static_path.append(str(static_assets_dir))
+    app.add_css_file("table.css")
+
+    return {
+        "version": "1.1",
+        "parallel_read_safe": True,
+        "parallel_write_safe": True,
+    }

docs/extension/rocm_docs_custom/selector.py

@@ -1,20 +1,10 @@
 from sphinx.util.docutils import SphinxDirective, directives, nodes
 from pathlib import Path
-from .utils import kv_to_data_attr, normalize_key
-import random
-import string
-
+from .utils import kv_to_data_attr, normalize_key, logger
 
 class SelectorGroup(nodes.General, nodes.Element):
     """
-    A row within a selector container.
-
-    rST usage:
-
-    .. selector-group:: Heading
-       :key:
-       :show-when: os=ubuntu (list of key=value pairs separated by spaces)
-       :heading-width: 4 (defaults to 6)
+    A row or dropdown within a selector container.
     """
 
     @staticmethod
@@ -23,32 +13,46 @@ class SelectorGroup(nodes.General, nodes.Element):
         key = node["key"]
         show_when_attr = kv_to_data_attr("show-when", node["show-when"])
         heading_width = node["heading-width"]
-        icon = node["icon"]
+        list_mode = node.get("list", False)
 
-        icon_html = ""
-        if icon:
-            icon_html = f'<i class="rocm-docs-selector-icon {icon}"></i>'
+        # Standard tile mode
+        info_nodes = list(node.findall(SelectorInfo))
+        info_link = info_nodes[0]["link"] if info_nodes else None
+        info_icon = info_nodes[0]["icon"] if info_nodes else None
+
+        info_icon_html = ""
+        if info_link:
+            info_icon_html = f"""
+            <a href="{info_link}" target="_blank">
+                <i class="rocm-docs-selector-icon {info_icon}"></i>
+            </a>
+            """
 
         translator.body.append(
             "<!-- start selector-group row -->"
             f"""
-            <div id="{nodes.make_id(label)}" class="rocm-docs-selector-group row gx-0 pt-2"
+            <div id="{nodes.make_id(label)}"
+                class="rocm-docs-selector-group row gx-0 pt-2"
                 data-selector-key="{key}"
                 {show_when_attr}
-                role="radiogroup"
+                {'role="radiogroup"' if list_mode else ""}
                 aria-label="{label}"
             >
                 <div class="col-{heading_width} me-1 px-2 rocm-docs-selector-group-heading">
-                    <span class="rocm-docs-selector-group-heading-text">{label}{icon_html}</span>
+                    <span class="rocm-docs-selector-group-heading-text">{label}{info_icon_html}</span>
                 </div>
                 <div class="row col-{12 - heading_width} pe-0">
+                {f'<select class="form-select rocm-docs-selector-dropdown-list" aria-label="{label}">' if list_mode else ""}
             """.strip()
         )
 
     @staticmethod
-    def depart_html(translator, _):
+    def depart_html(translator, node):
+        list_mode = node.get("list", False)
+
         translator.body.append(
-            """
+            f"""
+                {"</select>" if list_mode else ""}
                 </div>
             </div>
             """
@@ -57,14 +61,14 @@ class SelectorGroup(nodes.General, nodes.Element):
 
 
 class SelectorGroupDirective(SphinxDirective):
-    required_arguments = 1  # tile text
+    required_arguments = 1  # title text
     final_argument_whitespace = True
     has_content = True
     option_spec = {
         "key": directives.unchanged,
         "show-when": directives.unchanged,
         "heading-width": directives.nonnegative_int,
-        "icon": directives.unchanged,
+        "list": directives.flag,
     }
 
     def run(self):
@@ -74,73 +78,121 @@ class SelectorGroupDirective(SphinxDirective):
         node["key"] = normalize_key(self.options.get("key", label))
         node["show-when"] = self.options.get("show-when", "")
         node["heading-width"] = self.options.get("heading-width", 3)
-        node["icon"] = self.options.get("icon")
+        node["list"] = "list" in self.options
 
-        # Parse nested content
+        # Parse nested content (selector-info + selector-option)
         self.state.nested_parse(self.content, self.content_offset, node)
 
-        # Find all SelectorOption descendants
         option_nodes = list(node.findall(SelectorOption))
-
         if option_nodes:
-            # Set the group key on all options
             for opt in option_nodes:
                 opt["group_key"] = node["key"]
+                opt["list"] = node["list"]
 
-            # Find all options marked as default
+            # Default marking
             default_options = [opt for opt in option_nodes if opt["default"]]
-
-            if default_options:
-                # Multiple options marked :default: - only keep first as default
-                for i, opt in enumerate(default_options):
-                    if i > 0:
-                        opt["default"] = False
-            else:
-                # No explicit default - make first option default
+            if not default_options:
                 option_nodes[0]["default"] = True
 
         return [node]
 
+class SelectorInfo(nodes.General, nodes.Element):
+    """
+    Represents an informational icon/link associated with a selector group.
+    Appears as a clickable icon in the selector group heading.
+
+    rST usage:
+
+    .. selector:: AMD EPYC Server CPU
+       :key: cpu
+
+       .. selector-info:: https://www.amd.com/en/products/processors/server/epyc.html
+          :icon: fa-solid fa-circle-info fa-lg
+
+       .. selector-option:: EPYC 9005 (5th gen.)
+          :value: 9005
+    """
+
+    @staticmethod
+    def visit_html(translator, node):
+        # Do nothing — rendering handled by SelectorGroup
+        pass
+
+    @staticmethod
+    def depart_html(translator, node):
+        # Do nothing — prevent NotImplementedError
+        pass
+
+
+class SelectorInfoDirective(SphinxDirective):
+    required_arguments = 1  # link URL
+    final_argument_whitespace = True
+    has_content = False
+    option_spec = {"icon": directives.unchanged}
+
+    def run(self):
+        node = SelectorInfo()
+        node["link"] = self.arguments[0]
+        node["icon"] = self.options.get("icon", "fa-solid fa-circle-info fa-lg")
+
+        parent = getattr(self.state, "parent", None)
+        if not parent or not any(isinstance(p, SelectorGroup) for p in parent.traverse(include_self=True)):
+            logger.warning(
+                f"'.. selector-info::' at line {self.lineno} should be nested under a '.. selector::' directive",
+                location=(self.env.docname, self.lineno),
+            )
+
+        return [node]
+
 
 class SelectorOption(nodes.General, nodes.Element):
     """
-    A selectable tile within a selector group.
-
-    rST usage:
-
-    .. selector-option::
+    A selectable tile or list-item option within a selector group.
     """
 
     @staticmethod
     def visit_html(translator, node):
+        label = node["label"]
         value = node["value"]
+        show_when_attr = kv_to_data_attr("show-when", node["show-when"])
         disable_when_attr = kv_to_data_attr("disable-when", node["disable-when"])
         default = node["default"]
         width = node["width"]
-        group_key = node.get("group_key", "")
+        list_mode = node.get("list", False)
+
+        if list_mode:
+            selected_attr = " selected" if default else ""
+            translator.body.append(
+                f'<option value="{value}"{selected_attr} {show_when_attr} {disable_when_attr}>{label}</option>'
+            )
+            return
 
         default_class = "rocm-docs-selector-option-default" if default else ""
-
         translator.body.append(
             "<!-- start selector-option tile -->"
             f"""
             <div class="rocm-docs-selector-option {default_class} col-{width} px-2"
-                data-selector-key="{group_key}"
+                data-selector-key="{node.get('group_key', '')}"
                 data-selector-value="{value}"
+                {show_when_attr}
                 {disable_when_attr}
                 tabindex="0"
                 role="radio"
                 aria-checked="false"
             >
+                <span>{label}</span>
             """.strip()
         )
 
     @staticmethod
     def depart_html(translator, node):
+        list_mode = node.get("list", False)
+        if list_mode:
+            return  # no closing tag needed for <option>
         icon = node["icon"]
         if icon:
             translator.body.append(f'<i class="rocm-docs-selector-icon {icon}"></i>')
-        translator.body.append("</div>" "<!-- end selector-option tile -->")
+        translator.body.append("</div><!-- end selector-option tile -->")
 
 
 class SelectorOptionDirective(SphinxDirective):
@@ -148,6 +200,7 @@ class SelectorOptionDirective(SphinxDirective):
     final_argument_whitespace = True
     option_spec = {
         "value": directives.unchanged,
+        "show-when": directives.unchanged,
         "disable-when": directives.unchanged,
         "default": directives.flag,
         "width": directives.nonnegative_int,
@@ -160,22 +213,31 @@ class SelectorOptionDirective(SphinxDirective):
         node = SelectorOption()
         node["label"] = label
         node["value"] = normalize_key(self.options.get("value", label))
-        # node["show-when"] = self.options.get("show-when", "")
+        node["show-when"] = self.options.get("show-when", "")
         node["disable-when"] = self.options.get("disable-when", "")
         node["default"] = self.options.get("default", False) is not False
         node["width"] = self.options.get("width", 6)
         node["icon"] = self.options.get("icon")
 
         # Content replaces label if provided
-        if self.content:
-            self.state.nested_parse(self.content, self.content_offset, node)
-        else:
-            node += nodes.Text(label)
+        # if self.content:
+        #     self.state.nested_parse(self.content, self.content_offset, node)
+        # else:
+        #     node += nodes.Text(label)
+
+        parent = getattr(self.state, "parent", None)
+        if not parent or not any(isinstance(p, SelectorGroup) for p in parent.traverse(include_self=True)):
+            logger.warning(
+                f"'.. selector-option::' at line {self.lineno} should be nested under a '.. selector::' directive",
+                location=(self.env.docname, self.lineno),
+            )
+
         return [node]
 
+
 class SelectedContent(nodes.General, nodes.Element):
     """
-    A container to hold conditional content.
+    A container to hold documentation content to be shown conditionally.
 
     rST usage::
 
@@ -185,20 +247,21 @@ class SelectedContent(nodes.General, nodes.Element):
 
     @staticmethod
     def visit_html(translator, node):
-        show_when_attr = kv_to_data_attr("show-when", node["show-when"])
+        show_when = node.get("show-when", "")
+        show_when_attr = kv_to_data_attr("show-when", show_when)
         classes = " ".join(node.get("class", []))
         heading = node.get("heading", "")
-        heading_level = node.get("heading-level") or (SelectedContent._get_depth(node) + 1)
-        heading_level = min(heading_level, 6)
+        # default to <h2>
+        heading_level = min(node.get("heading-level") or 2, 6) # maximum depth is <h6>
 
+        id_attr = ""
         heading_elem = ""
+        combined_show_when = node.get("combined-show-when", show_when)
         if heading:
-            # HACK to fix secondary sidebar observer
-            suffix = "".join(random.choices(string.ascii_lowercase + string.digits, k=3))
-            id_attr = nodes.make_id(f"{heading}-{suffix}")
+            id_attr = nodes.make_id(f"{heading}-{combined_show_when}")
 
             heading_elem = (
-                f'<h{heading_level} id="{id_attr}" class="rocm-docs-custom-heading">'
+                f'<h{heading_level} class="rocm-docs-custom-heading">'
                 f'{heading}<a class="headerlink" href="#{id_attr}" title="Link to this heading">#</a>'
                 f'</h{heading_level}>'
             )
@@ -206,24 +269,22 @@ class SelectedContent(nodes.General, nodes.Element):
         translator.body.append(
             f"""
             <!-- start selected-content -->
-            <div class="rocm-docs-selected-content {classes}" {show_when_attr} aria-hidden="true">
+            <{"section" if heading else "div"}
+                id="{id_attr}"
+                class="rocm-docs-selected-content {classes}"
+                {show_when_attr}
+                aria-hidden="true">
                 {heading_elem}
             """.strip()
         )
 
     @staticmethod
-    def depart_html(translator, _):
-        translator.body.append("</div><!-- end selected-content -->")
+    def depart_html(translator, node):
+        heading = node.get("heading", "")
 
-    @staticmethod
-    def _get_depth(node):
-        depth = 1
-        parent = node.parent
-        while parent is not None:
-            if isinstance(parent, SelectedContent) and parent.get("heading"):
-                depth += 1
-            parent = getattr(parent, "parent", None)
-        return depth
+        translator.body.append(f"""
+            </{"section" if heading else "div"}>
+            <!-- end selected-content -->""")
 
 
 class SelectedContentDirective(SphinxDirective):
@@ -242,9 +303,20 @@ class SelectedContentDirective(SphinxDirective):
         node["show-when"] = self.arguments[0]
         node["id"] = self.options.get("id", "")
         node["class"] = self.options.get("class", "")
-        node["heading"] = self.options.get("heading", "")  # empty string if none
+        node["heading"] = self.options.get("heading", "")
         node["heading-level"] = self.options.get("heading-level", None)
 
+        # Collect parent show-whens (if nested)
+        # to create a completely unique id
+        parent_show_whens = []
+        for ancestor in self.state.parent.traverse(include_self=True):
+            if isinstance(ancestor, SelectedContent) and "show-when" in ancestor:
+                parent_show_whens.append(ancestor["show-when"])
+
+        # Compose combined show-when chain
+        combined_show_when = "+".join(parent_show_whens + [node["show-when"]])
+        node["combined-show-when"] = combined_show_when
+
         # Parse nested content
         self.state.nested_parse(self.content, self.content_offset, node)
         return [node]
@@ -255,6 +327,10 @@ def setup(app):
         SelectorGroup,
         html=(SelectorGroup.visit_html, SelectorGroup.depart_html),
     )
+    app.add_node(
+        SelectorInfo,
+        html=(SelectorInfo.visit_html, SelectorInfo.depart_html),
+    )
     app.add_node(
         SelectorOption,
         html=(SelectorOption.visit_html, SelectorOption.depart_html),
@@ -265,13 +341,14 @@ def setup(app):
     )
 
     app.add_directive("selector", SelectorGroupDirective)
+    app.add_directive("selector-info", SelectorInfoDirective)
     app.add_directive("selector-option", SelectorOptionDirective)
     app.add_directive("selected-content", SelectedContentDirective)
     app.add_directive("selected", SelectedContentDirective)
 
-    static_assets_dir = Path(__file__).parent / "static"
+    static_assets_dir = Path(__file__).parent / "static" / "selector"
     app.config.html_static_path.append(str(static_assets_dir))
     app.add_css_file("selector.css")
     app.add_js_file("selector.js", type="module", defer="defer")
 
-    return {"version": "1.0", "parallel_read_safe": True}
+    return {"version": "1.1", "parallel_read_safe": True}

docs/extension/rocm_docs_custom/static/selector.js

@@ -1,300 +0,0 @@
-const READY_EVENT = "ROCmDocsSelectorsReady";
-const STATE_CHANGE_EVENT = "ROCmDocsSelectorStateChanged";
-
-const GROUP_QUERY = ".rocm-docs-selector-group";
-const OPTION_QUERY = ".rocm-docs-selector-option";
-const COND_QUERY = "[data-show-when]";
-const TOC2_OPTIONS_LIST_QUERY = ".rocm-docs-selector-toc2-options";
-const TOC2_CONTENTS_LIST_QUERY = ".rocm-docs-selector-toc2-contents";
-const HEADING_QUERY = ".rocm-docs-selected-content h1,h2,h3,h4,h5,h6[id]";
-
-const isDefaultOption = (elem) =>
-  elem.classList.contains("rocm-docs-selector-option-default");
-
-const DISABLED_CLASS = "rocm-docs-disabled";
-const disable = (elem) => {
-  elem.classList.add(DISABLED_CLASS);
-  elem.setAttribute("aria-disabled", "true");
-  elem.setAttribute("tabindex", "-1");
-};
-// const enable = (elem) => {
-//   elem.classList.remove(DISABLED_CLASS);
-//   elem.setAttribute("aria-disabled", "false");
-//   elem.setAttribute("tabindex", "0");
-// };
-
-const HIDDEN_CLASS = "rocm-docs-hidden";
-const hide = (elem) => {
-  elem.classList.add(HIDDEN_CLASS);
-  elem.setAttribute("aria-hidden", "true");
-};
-const show = (elem) => {
-  elem.classList.remove(HIDDEN_CLASS);
-  elem.setAttribute("aria-hidden", "false");
-};
-
-const SELECTED_CLASS = "rocm-docs-selected";
-const select = (elem) => {
-  elem.classList.add(SELECTED_CLASS);
-  elem.setAttribute("aria-checked", "true");
-};
-const deselect = (elem) => {
-  elem.classList.remove(SELECTED_CLASS);
-  elem.setAttribute("aria-checked", "false");
-};
-
-const state = {};
-
-function getState() {
-  return { ...state };
-}
-
-function setState(updates) {
-  const previousState = getState();
-  Object.assign(state, updates);
-
-  const event = new CustomEvent(STATE_CHANGE_EVENT, {
-    detail: {
-      previousState,
-      currentState: getState(),
-      changes: updates,
-    },
-  });
-  document.dispatchEvent(event);
-}
-
-function validateOptionElem(optionElem) {
-  const key = optionElem.dataset.selectorKey;
-  const value = optionElem.dataset.selectorValue;
-
-  const errors = [];
-  if (!key) errors.push("Missing 'data-selector-key'");
-  if (!value) errors.push("Missing 'data-selector-value'");
-
-  if (errors.length === 0) return;
-
-  const label = optionElem.textContent.trim() || "<unnamed option>";
-  console.error(
-    `[ROCmDocsSelector] Invalid selector option '${label}': ${
-      errors.join(", ")
-    }!`,
-  );
-  disable(optionElem);
-}
-
-function handleOptionSelect(e) {
-  const option = e.currentTarget;
-  const parentGroup = option.closest(GROUP_QUERY);
-  const siblingOptions = parentGroup.querySelectorAll(OPTION_QUERY);
-
-  siblingOptions.forEach((elem) => deselect(elem));
-  select(option);
-
-  // Update global state
-  const key = option.dataset.selectorKey;
-  const value = option.dataset.selectorValue;
-  if (key && value) setState({ [key]: value });
-
-  updateVisibility();
-}
-
-function shouldBeShown(elem) {
-  const conditionsData = elem.dataset.showWhen;
-  if (!conditionsData) return true; // Default visible
-
-  try {
-    const conditions = JSON.parse(conditionsData);
-    // Ensure it's an object
-    if (typeof conditions !== "object" || Array.isArray(conditions)) {
-      console.warn(
-        "[ROCmDocsSelector] Invalid 'show-when' format (must be key/value object):",
-        conditionsData,
-      );
-      return true;
-    }
-
-    for (const [key, value] of Object.entries(conditions)) {
-      const currentValue = state[key];
-
-      if (currentValue === undefined) return false;
-
-      if (Array.isArray(value)) {
-        if (!value.includes(currentValue)) return false;
-        continue;
-      }
-
-      if (state[key] !== value) {
-        return false;
-      }
-    }
-
-    return true;
-  } catch (err) {
-    console.error(
-      "[ROCmDocsSelector] Couldn't parse 'show-when' conditions:",
-      err,
-    );
-    return true;
-  }
-}
-
-function updateTOC2OptionsList() {
-  const tocOptionsList = document.querySelector(TOC2_OPTIONS_LIST_QUERY);
-  if (!tocOptionsList) return;
-
-  // Clear previous entries
-  tocOptionsList.innerHTML = "";
-
-  // Get only visible selector groups
-  const groups = Array.from(document.querySelectorAll(GROUP_QUERY)).filter(
-    (g) => g.offsetParent !== null
-  );
-
-  if (groups.length === 0) {
-    const li = document.createElement("li");
-    li.className =
-      "nav-item toc-entry toc-h3 rocm-docs-selector-toc2-item empty";
-    const span = document.createElement("span");
-    span.textContent = "(no visible selectors)";
-    li.appendChild(span);
-    tocOptionsList.appendChild(li);
-    return;
-  }
-
-  groups.forEach((group) => {
-    // ✅ Find group heading span
-    const headingSpan = group.querySelector(
-      ".rocm-docs-selector-group-heading-text"
-    );
-    const headingText = headingSpan
-      ? headingSpan.textContent.trim()
-      : "(Unnamed Selector)";
-
-    // Find currently selected option
-    const selectedOption = group.querySelector(`.${SELECTED_CLASS}`);
-    const optionText = selectedOption
-      ? selectedOption.textContent.trim()
-      : "(none selected)";
-
-    // Build list item
-    const li = document.createElement("li");
-    li.className = "nav-item toc-entry toc-h3 rocm-docs-selector-toc2-item";
-
-    const link = document.createElement("a");
-    link.className = "nav-link";
-    link.href = `#${group.id}`;
-    link.innerHTML = `<strong>${headingText}</strong>: ${optionText}`;
-
-    li.appendChild(link);
-    tocOptionsList.appendChild(li);
-  });
-}
-
-function updateTOC2ContentsList() {
-  const tocOptionsList = document.querySelector(TOC2_OPTIONS_LIST_QUERY);
-  const tocContentsList = document.querySelector(TOC2_CONTENTS_LIST_QUERY);
-  if (!tocContentsList || !tocOptionsList) return;
-
-  const visibleHeaders = [...document.querySelectorAll(HEADING_QUERY)]
-    .filter((h) => h.offsetParent !== null); // only visible headings
-
-  tocContentsList
-    .querySelectorAll("li.toc-entry.rocm-docs-selector-toc2-item")
-    .forEach((node) => node.remove());
-
-  if (visibleHeaders.length === 0) return;
-
-  let lastH2Li = null;
-
-  visibleHeaders.forEach((h) => {
-    const level = parseInt(h.tagName.substring(1), 10);
-    const li = document.createElement("li");
-    li.className = `nav-item toc-entry toc-${h.tagName.toLowerCase()} rocm-docs-selector-toc2-item`;
-
-    const a = document.createElement("a");
-    a.className = "reference internal nav-link";
-    const section = h.closest("section");
-    const fallbackId = section ? section.id : "";
-    a.href = h.id ? `#${h.id}` : fallbackId ? `#${fallbackId}` : "#";
-    a.textContent = h.cloneNode(true).childNodes[0].textContent.trim();
-    li.appendChild(a);
-
-    // Nest logic: h3+ belong to last h2's inner list
-    if (level === 2) {
-      tocContentsList.appendChild(li);
-      lastH2Li = li;
-    } else if (level === 3 && lastH2Li) {
-      // ensure nested UL exists
-      let innerUl = lastH2Li.querySelector("ul");
-      if (!innerUl) {
-        innerUl = document.createElement("ul");
-        innerUl.className = "nav section-nav flex-column";
-        lastH2Li.appendChild(innerUl);
-      }
-      innerUl.appendChild(li);
-    } else {
-      tocContentsList.appendChild(li);
-    }
-  });
-}
-
-function updateVisibility() {
-  document.querySelectorAll(COND_QUERY).forEach((elem) => {
-    if (shouldBeShown(elem)) {
-      show(elem);
-    } else {
-      hide(elem);
-    }
-  });
-
-  updateTOC2OptionsList();
-  updateTOC2ContentsList();
-}
-
-function init() {
-  const selectorOptions = document.querySelectorAll(OPTION_QUERY);
-
-  const initialState = {};
-
-  // Attach listeners and gather defaults
-  selectorOptions.forEach((option) => {
-    validateOptionElem(option);
-
-    option.addEventListener("click", handleOptionSelect);
-    option.addEventListener("keydown", (e) => {
-      if (e.key === "Enter" || e.key === " ") {
-        e.preventDefault();
-        handleOptionSelect(e);
-      }
-    });
-
-    if (isDefaultOption(option)) {
-      select(option);
-      const { selectorKey: key, selectorValue: value } = option.dataset;
-      if (key && value) {
-        initialState[key] = value;
-      }
-    }
-  });
-
-  setState(initialState);
-  updateVisibility();
-
-  document.dispatchEvent(new CustomEvent(READY_EVENT));
-}
-
-function domReady(callback) {
-  if (document.readyState !== "loading") {
-    callback();
-  } else {
-    document.addEventListener("DOMContentLoaded", callback, { once: true });
-  }
-}
-
-// window.rocmDocsSelector = {
-//   setState,
-//   getState,
-// };
-
-// Initialize when DOM is ready
-domReady(init);

docs/extension/rocm_docs_custom/static/selector/selector-toc.js

@@ -0,0 +1,235 @@
+const GROUP_QUERY = ".rocm-docs-selector-group";
+const SELECTED_CLASS = "rocm-docs-selected";
+const TOC2_OPTIONS_LIST_QUERY = ".rocm-docs-selector-toc2-options";
+const TOC2_CONTENTS_LIST_QUERY = ".rocm-docs-selector-toc2-contents";
+const HEADING_QUERY = ".rocm-docs-selected-content h1,h2,h3,h4,h5,h6";
+
+const TOC_ITEM_CLASS = "rocm-docs-selector-toc2-item";
+const EMPTY_ITEM_CLASS = "empty";
+
+let optionsTocInitialized = false;
+
+function isVisible(el) {
+  return !!(el && el.offsetParent !== null);
+}
+
+function getUniqueGroups(groups) {
+  const seen = new Set();
+  return groups.filter((group) => {
+    // Use group ID as primary identity; fallback to heading text
+    const headingSpan = group.querySelector(
+      ".rocm-docs-selector-group-heading-text"
+    );
+    const headingText = headingSpan
+      ? headingSpan.textContent.trim()
+      : "(Unnamed Selector)";
+    const identifier = group.id ? `id:${group.id}` : `heading:${headingText}`;
+
+    if (seen.has(identifier)) return false;
+    seen.add(identifier);
+    return true;
+  });
+}
+
+function initTOC2OptionsList() {
+  const tocOptionsList = document.querySelector(TOC2_OPTIONS_LIST_QUERY);
+  if (!tocOptionsList) return;
+
+  tocOptionsList.innerHTML = "";
+
+  let groups = Array.from(document.querySelectorAll(GROUP_QUERY)).filter(isVisible);
+  groups = getUniqueGroups(groups);
+
+  if (groups.length === 0) {
+    const li = document.createElement("li");
+    li.className = `nav-item toc-entry toc-h3 ${TOC_ITEM_CLASS} ${EMPTY_ITEM_CLASS}`;
+    const span = document.createElement("span");
+    span.textContent = "(no visible selectors)";
+    li.appendChild(span);
+    tocOptionsList.appendChild(li);
+    optionsTocInitialized = true;
+    return;
+  }
+
+  groups.forEach((group) => {
+    const headingSpan = group.querySelector(
+      ".rocm-docs-selector-group-heading-text"
+    );
+    const headingText = headingSpan
+      ? headingSpan.textContent.trim()
+      : "(Unnamed Selector)";
+
+    const li = document.createElement("li");
+    li.className = `nav-item toc-entry toc-h3 ${TOC_ITEM_CLASS}`;
+    li.dataset.groupId = group.id || "";
+
+    const link = document.createElement("a");
+    link.className = "nav-link";
+    link.href = group.id ? `#${group.id}` : "#";
+    link.dataset.headingText = headingText;
+
+    const selectedOption = group.querySelector(`.${SELECTED_CLASS}`);
+    let optionText = "(none selected)";
+    if (selectedOption) {
+      const clone = selectedOption.cloneNode(true);
+      clone.querySelectorAll("i, svg").forEach((el) => el.remove());
+      optionText = clone.innerHTML.trim();
+    }
+
+    link.innerHTML = `<strong>${headingText}</strong>: ${optionText}`;
+    li.appendChild(link);
+    tocOptionsList.appendChild(li);
+  });
+
+  optionsTocInitialized = true;
+}
+
+export function updateTOC2OptionsList() {
+  const tocOptionsList = document.querySelector(TOC2_OPTIONS_LIST_QUERY);
+  if (!tocOptionsList) return;
+
+  let visibleGroups = Array.from(document.querySelectorAll(GROUP_QUERY)).filter(
+    isVisible
+  );
+  visibleGroups = getUniqueGroups(visibleGroups);
+
+  // Always rebuild fresh (simpler, avoids state drift)
+  tocOptionsList.innerHTML = "";
+
+  if (visibleGroups.length === 0) {
+    const li = document.createElement("li");
+    li.className = `nav-item toc-entry toc-h3 ${TOC_ITEM_CLASS} ${EMPTY_ITEM_CLASS}`;
+    const span = document.createElement("span");
+    span.textContent = "(no visible selectors)";
+    li.appendChild(span);
+    tocOptionsList.appendChild(li);
+    return;
+  }
+
+  visibleGroups.forEach((group) => {
+    const headingSpan = group.querySelector(
+      ".rocm-docs-selector-group-heading-text"
+    );
+    const headingText = headingSpan
+      ? headingSpan.textContent.trim()
+      : "(Unnamed Selector)";
+
+    const li = document.createElement("li");
+    li.className = `nav-item toc-entry toc-h3 ${TOC_ITEM_CLASS}`;
+    li.dataset.groupId = group.id || "";
+
+    const link = document.createElement("a");
+    link.className = "nav-link";
+    link.href = group.id ? `#${group.id}` : "#";
+    link.dataset.headingText = headingText;
+
+    const selectedOption = group.querySelector(`.${SELECTED_CLASS}`);
+    let optionText = "(none selected)";
+    if (selectedOption) {
+      const clone = selectedOption.cloneNode(true);
+      clone.querySelectorAll("i, svg").forEach((el) => el.remove());
+      optionText = clone.innerHTML.trim();
+    }
+
+    link.innerHTML = `<strong>${headingText}</strong>: ${optionText}`;
+    li.appendChild(link);
+    tocOptionsList.appendChild(li);
+  });
+}
+
+let contentsTocInitialized = false;
+
+function initTOC2ContentsList() {
+  const tocContentsList = document.querySelector(TOC2_CONTENTS_LIST_QUERY);
+  if (!tocContentsList) return;
+
+  // Remove any previous dynamic items (idempotent init)
+  tocContentsList
+    .querySelectorAll(`li.toc-entry.${TOC_ITEM_CLASS}`)
+    .forEach((node) => node.remove());
+
+  const headings = Array.from(document.querySelectorAll(HEADING_QUERY));
+  if (headings.length === 0) {
+    contentsTocInitialized = true;
+    return;
+  }
+
+  const lastLiByLevel = {};
+
+  headings.forEach((h) => {
+    const level = parseInt(h.tagName.substring(1), 10);
+    if (Number.isNaN(level) || level < 2 || level > 6) return;
+
+    const li = document.createElement("li");
+    li.className = `nav-item toc-entry toc-${h.tagName.toLowerCase()} ` +
+      TOC_ITEM_CLASS;
+
+    const a = document.createElement("a");
+    a.className = "reference internal nav-link";
+
+    const section = h.closest("section");
+    const targetId = h.id || (section ? section.id : "");
+    a.href = targetId ? `#${targetId}` : "#";
+
+    // Use only the text from the heading (ignore headerlink icon etc.)
+    const clone = h.cloneNode(true);
+    const firstTextNode = clone.childNodes.length > 0
+      ? clone.childNodes[0].textContent
+      : "";
+    a.textContent = (firstTextNode || "").trim();
+
+    li.dataset.targetId = targetId;
+    li.appendChild(a);
+
+    // Nest under closest previous shallower heading
+    let parentUl = null;
+    for (let parentLevel = level - 1; parentLevel >= 2; parentLevel -= 1) {
+      const parentLi = lastLiByLevel[parentLevel];
+      if (parentLi) {
+        parentUl = parentLi.querySelector("ul");
+        if (!parentUl) {
+          parentUl = document.createElement("ul");
+          parentUl.className = "nav section-nav flex-column";
+          parentLi.appendChild(parentUl);
+        }
+        break;
+      }
+    }
+
+    if (parentUl) {
+      parentUl.appendChild(li);
+    } else {
+      tocContentsList.appendChild(li);
+    }
+
+    lastLiByLevel[level] = li;
+    for (let deeper = level + 1; deeper <= 6; deeper += 1) {
+      delete lastLiByLevel[deeper];
+    }
+  });
+
+  contentsTocInitialized = true;
+}
+
+export function updateTOC2ContentsList() {
+  const tocOptionsList = document.querySelector(TOC2_OPTIONS_LIST_QUERY);
+  const tocContentsList = document.querySelector(TOC2_CONTENTS_LIST_QUERY);
+  if (!tocContentsList || !tocOptionsList) return;
+
+  if (!contentsTocInitialized) {
+    initTOC2ContentsList();
+  }
+
+  tocContentsList
+    .querySelectorAll(`li.toc-entry.${TOC_ITEM_CLASS}`)
+    .forEach((li) => {
+      const targetId = li.dataset.targetId;
+      if (!targetId) {
+        li.style.display = "none";
+        return;
+      }
+      const target = document.getElementById(targetId);
+      const visible = target && target.offsetParent !== null;
+      li.style.display = visible ? "" : "none";
+    });
+}

docs/extension/rocm_docs_custom/static/selector/selector.css

@@ -7,6 +7,11 @@ html {
   --rocm-docs-selector-option-hover-color: var(--pst-color-link-hover);
   --rocm-docs-selector-option-selected-color: var(--pst-color-primary);
   --rocm-docs-selector-tile-padding: 0.2rem;
+  --rocm-docs-selector-tile-gap: 0.5rem;
+  --rocm-docs-selector-focus-ring: 2px solid var(
+    --rocm-docs-selector-accent-color
+  );
+  --rocm-docs-selector-focus-offset: 2px;
 }
 
 html[data-theme="light"] {
@@ -21,15 +26,37 @@ html[data-theme="dark"] {
   --rocm-docs-selector-shadow-hover: 0 2px 8px rgba(0, 0, 0, 0.4);
 }
 
-/* Selector container */
-.rocm-docs-selector-container {
-  padding: 0 0 1rem 0;
+/* Avoid odd sizing interactions with Bootstrap width utilities */
+.rocm-docs-selector-group,
+.rocm-docs-selector-group * {
+  box-sizing: border-box;
 }
 
-/* Selector group heading when one of its options is hovered */
-.rocm-docs-selector-group:has(.rocm-docs-selector-option:hover)
-  .rocm-docs-selector-group-heading {
-  border-right-color: var(--rocm-docs-selector-option-hover-color);
+/* Hide selectors during initialization to prevent FOUC */
+.rocm-docs-selector-group:not(.rocm-docs-selector-initialized) {
+  visibility: hidden;
+}
+
+/* Smooth fade-in when ready */
+.rocm-docs-selector-group.rocm-docs-selector-initialized {
+  visibility: visible;
+  animation: rocm-docs-selector-fade-in 0.2s ease-in;
+}
+
+@keyframes rocm-docs-selector-fade-in {
+  from {
+    opacity: 0;
+  }
+  to {
+    opacity: 1;
+  }
+}
+
+@supports selector(.x:has(.y)) {
+  .rocm-docs-selector-group:has(.rocm-docs-selector-option:hover)
+    .rocm-docs-selector-group-heading {
+    border-right-color: var(--rocm-docs-selector-option-hover-color);
+  }
 }
 
 /* Selector group heading box */
@@ -39,7 +66,7 @@ html[data-theme="dark"] {
   font-weight: 600;
   border-right: solid 3px var(--rocm-docs-selector-accent-color);
   border-radius: var(--rocm-docs-selector-border-radius);
-  transition: border-right-color 0.25s ease;
+  transition: border-color 0.25s ease;
   box-shadow: var(--rocm-docs-selector-shadow);
 }
 
@@ -57,16 +84,25 @@ html[data-theme="dark"] {
   justify-content: space-between;
   align-items: center;
   gap: 0.5rem;
+
   background-color: var(--rocm-docs-selector-bg-color);
   padding: var(--rocm-docs-selector-tile-padding);
   border: solid 2px var(--rocm-docs-selector-border-color);
-  cursor: pointer;
-  transition: all 0.2 ease;
   border-radius: var(--rocm-docs-selector-border-radius);
   box-shadow: var(--rocm-docs-selector-shadow);
+
+  cursor: pointer;
+  user-select: none;
+
+  transition:
+    background-color 0.2s ease,
+    color 0.2s ease,
+    transform 0.2s ease,
+    box-shadow 0.2s ease,
+    border-color 0.2s ease;
 }
 
-/* Selector option when hovered */
+/* Hover (not disabled) */
 .rocm-docs-selector-option:hover:not(.rocm-docs-disabled) {
   background-color: var(--rocm-docs-selector-option-hover-color);
   color: var(--rocm-docs-selector-fg-color);
@@ -74,30 +110,94 @@ html[data-theme="dark"] {
   box-shadow: var(--rocm-docs-selector-shadow-hover);
 }
 
-.rocm-docs-selector-option:focus:not(.rocm-docs-disabled) {
-  z-index: 69;
+/* Accessible keyboard focus */
+.rocm-docs-selector-option:focus-visible:not(.rocm-docs-disabled) {
+  outline: var(--rocm-docs-selector-focus-ring);
+  outline-offset: var(--rocm-docs-selector-focus-offset);
 }
 
-/* Selector option when selected */
+/* Keep it above neighbors if it gets an outline/box-shadow */
+.rocm-docs-selector-option:focus:not(.rocm-docs-disabled) {
+  position: relative;
+  z-index: 1;
+}
+
+/* Selected */
 .rocm-docs-selector-option.rocm-docs-selected {
   background-color: var(--rocm-docs-selector-option-selected-color);
   color: var(--rocm-docs-selector-fg-color);
 }
 
-/* Prevent hover effect on selected */
+/* Prevent hover lift on selected (keeps it steady) */
 .rocm-docs-selector-option.rocm-docs-selected:hover {
   transform: none;
 }
 
-/* Selector option when disabled */
+/* Disabled */
 .rocm-docs-selector-option.rocm-docs-disabled {
   background-color: var(--rocm-docs-selector-border-color);
   color: var(--rocm-docs-selector-fg-color);
   cursor: not-allowed;
   pointer-events: none;
+  opacity: 0.75;
 }
 
 /* Hidden state */
 .rocm-docs-hidden {
   display: none;
 }
+
+/* Put selected option summary in secondary page-level TOC on a new line */
+.rocm-docs-selector-toc2-item .nav-link span {
+  display: block;
+  padding-left: 1rem;
+}
+
+@media (max-width: 768px) {
+  .rocm-docs-selector-group {
+    row-gap: var(--rocm-docs-selector-tile-gap);
+  }
+}
+
+@media (max-width: 576px) {
+  .rocm-docs-selector-option {
+    flex: 0 0 100%;
+    max-width: 100%;
+  }
+}
+
+@media (max-width: 440px) {
+  .rocm-docs-selector-group-heading {
+    flex: 0 0 100%;
+    max-width: 100%;
+
+    /* When stacked, puth the accent on the bottom */
+    border-right: 0;
+    border-bottom: solid 3px var(--rocm-docs-selector-accent-color);
+  }
+
+  .rocm-docs-selector-group.row .row {
+    flex: 0 0 100%;
+    max-width: 100%;
+  }
+
+  .rocm-docs-selector-option {
+    flex: 0 0 100%;
+    max-width: 100%;
+  }
+}
+
+/* Motion reduction */
+@media (prefers-reduced-motion: reduce) {
+  .rocm-docs-selector-group.rocm-docs-selector-initialized {
+    animation: none;
+  }
+
+  .rocm-docs-selector-option {
+    transition: none;
+  }
+
+  .rocm-docs-selector-option:hover:not(.rocm-docs-disabled) {
+    transform: none;
+  }
+}

docs/extension/rocm_docs_custom/static/selector/selector.js

@@ -0,0 +1,329 @@
+import { domReady, logDebug } from "./utils.js";
+import {
+  updateTOC2ContentsList,
+  updateTOC2OptionsList,
+} from "./selector-toc.js";
+
+const GROUP_QUERY = ".rocm-docs-selector-group";
+const OPTION_QUERY = ".rocm-docs-selector-option";
+const COND_QUERY = "[data-show-when],[data-disable-when]";
+
+const DEFAULT_OPTION_CLASS = "rocm-docs-selector-option-default";
+const DISABLED_CLASS = "rocm-docs-disabled";
+const HIDDEN_CLASS = "rocm-docs-hidden";
+const SELECTED_CLASS = "rocm-docs-selected";
+
+// Toggle helpers -------------------------------------------------------------
+
+const isDefaultOption = (elem) => elem.classList.contains(DEFAULT_OPTION_CLASS);
+
+const disable = (elem) => {
+  elem.classList.add(DISABLED_CLASS);
+  elem.setAttribute("aria-disabled", "true");
+  elem.setAttribute("tabindex", "-1");
+};
+
+const enable = (elem) => {
+  elem.classList.remove(DISABLED_CLASS);
+  elem.setAttribute("aria-disabled", "false");
+  elem.setAttribute("tabindex", "0");
+};
+
+const hide = (elem) => {
+  elem.classList.add(HIDDEN_CLASS);
+  elem.setAttribute("aria-hidden", "true");
+};
+
+const show = (elem) => {
+  elem.classList.remove(HIDDEN_CLASS);
+  elem.setAttribute("aria-hidden", "false");
+};
+
+const select = (elem) => {
+  elem.classList.add(SELECTED_CLASS);
+  elem.setAttribute("aria-checked", "true");
+};
+
+const deselect = (elem) => {
+  elem.classList.remove(SELECTED_CLASS);
+  elem.setAttribute("aria-checked", "false");
+};
+
+// Global selector state ------------------------------------------------------
+
+const state = {};
+
+function getState() {
+  return { ...state };
+}
+
+function setState(updates) {
+  Object.assign(state, updates);
+  logDebug("State updated:", state);
+}
+
+// Condition handling ---------------------------------------------------------
+
+/**
+ * Safely parse JSON-encoded conditions from a data-* attribute.
+ * Expects a key/value object, where values may be strings or arrays of strings.
+ */
+function parseConditions(attrName, raw) {
+  if (!raw) return null;
+
+  try {
+    const conditions = JSON.parse(raw);
+    if (typeof conditions !== "object" || Array.isArray(conditions)) {
+      console.warn(
+        `[ROCmDocsSelector] Invalid '${attrName}' format ` +
+          "(must be a key/value object):",
+        raw,
+      );
+      return null;
+    }
+    return conditions;
+  } catch (err) {
+    console.error(
+      `[ROCmDocsSelector] Couldn't parse '${attrName}' conditions:`,
+      err,
+    );
+    return null;
+  }
+}
+
+/**
+ * Return true iff all conditions match the current state.
+ * - Values can be a string or an array of strings.
+ * - A condition with an undefined state key is treated as not matching.
+ */
+function matchesConditions(conditions, currentState) {
+  for (const [key, expected] of Object.entries(conditions)) {
+    const actual = currentState[key];
+
+    // If no value yet, this condition does not match.
+    if (actual === undefined) return false;
+
+    if (Array.isArray(expected)) {
+      if (!expected.includes(actual)) return false;
+    } else if (actual !== expected) {
+      return false;
+    }
+  }
+  return true;
+}
+
+function shouldBeDisabled(elem) {
+  const raw = elem.dataset.disableWhen;
+  if (!raw) return false; // no conditions => never disabled
+
+  const conditions = parseConditions("disable-when", raw);
+  if (!conditions) {
+    console.warn(
+      "[ROCmDocsSelector] Invalid 'show-when' conditions; " +
+        "hiding affected element.",
+    );
+    return false;
+  }
+
+  return matchesConditions(conditions, state);
+}
+
+function shouldBeShown(elem) {
+  const raw = elem.dataset.showWhen;
+  if (!raw) return true; // no conditions => always visible
+
+  const conditions = parseConditions("show-when", raw);
+  if (!conditions) return true;
+
+  return matchesConditions(conditions, state);
+}
+
+// Event handlers -------------------------------------------------------------
+
+function handleOptionSelect(e) {
+  const option = e.currentTarget;
+
+  // Ignore interaction with disabled or already selected options
+  if (
+    option.classList.contains(DISABLED_CLASS) ||
+    option.classList.contains(SELECTED_CLASS)
+  ) {
+    return;
+  }
+
+  const { selectorKey: key, selectorValue: value } = option.dataset;
+  if (!key || !value) return;
+
+  // Update all selectors sharing the same key
+  const allOptions = document.querySelectorAll(
+    `${OPTION_QUERY}[data-selector-key="${key}"]`,
+  );
+
+  allOptions.forEach((opt) => {
+    if (opt.dataset.selectorValue === value) {
+      select(opt);
+    } else {
+      deselect(opt);
+    }
+  });
+
+  // Update global state
+  setState({ [key]: value });
+
+  // Re-run visibility rules and TOC sync
+  updateVisibility();
+}
+
+function handleOptionKeydown(e) {
+  if (e.key === "Enter" || e.key === " ") {
+    e.preventDefault();
+    handleOptionSelect(e);
+  }
+}
+
+// Visibility / enablement update --------------------------------------------
+
+// Ensure each selector group always has a valid selected option.
+// If the current selection becomes disabled/hidden due to another selector's
+// change, automatically pick a replacement.
+function reconcileGroupSelections() {
+  const currentState = getState();
+  const updates = {};
+
+  document.querySelectorAll(GROUP_QUERY).forEach((group) => {
+    // Skip groups that are themselves hidden
+    if (group.classList.contains(HIDDEN_CLASS)) return;
+
+    const options = Array.from(group.querySelectorAll(OPTION_QUERY));
+    if (!options.length) return;
+
+    const groupKey = group.dataset.selectorKey ||
+      options[0].dataset.selectorKey;
+    if (!groupKey) return;
+
+    // Options that are both enabled and visible
+    const enabledVisible = options.filter(
+      (opt) =>
+        !opt.classList.contains(DISABLED_CLASS) &&
+        !opt.classList.contains(HIDDEN_CLASS),
+    );
+
+    if (!enabledVisible.length) {
+      // No valid options left; just clear visual selection.
+      options.forEach(deselect);
+      return;
+    }
+
+    const currentlySelected = options.find((opt) =>
+      opt.classList.contains(SELECTED_CLASS)
+    );
+
+    const selectedStillValid = currentlySelected &&
+      enabledVisible.includes(currentlySelected);
+
+    if (selectedStillValid) {
+      const selectedValue = currentlySelected.dataset.selectorValue;
+      if (selectedValue && currentState[groupKey] !== selectedValue) {
+        updates[groupKey] = selectedValue;
+      }
+      return;
+    }
+
+    // Need a new selection: prefer a default option, otherwise the first
+    // enabled+visible option in DOM order.
+    const replacement = enabledVisible.find(isDefaultOption) ||
+      enabledVisible[0];
+    if (!replacement) return;
+
+    options.forEach(deselect);
+    select(replacement);
+
+    const newValue = replacement.dataset.selectorValue;
+    if (newValue && currentState[groupKey] !== newValue) {
+      updates[groupKey] = newValue;
+    }
+  });
+
+  const changedKeys = Object.keys(updates);
+  if (changedKeys.length > 0) {
+    setState(updates);
+    return true;
+  }
+  return false;
+}
+
+let isUpdatingVisibility = false;
+
+function updateVisibility() {
+  // Prevent re-entrancy if something triggers updateVisibility
+  // while it is already running.
+  if (isUpdatingVisibility) return;
+  isUpdatingVisibility = true;
+
+  try {
+    let stateChanged = false;
+    let iterations = 0;
+
+    // We may need multiple passes: reconciling selections can change the
+    // global state, which in turn affects show/disable conditions.
+    do {
+      document.querySelectorAll(COND_QUERY).forEach((elem) => {
+        // Show/hide only if element has show-when
+        if (elem.dataset.showWhen !== undefined) {
+          if (shouldBeShown(elem)) {
+            show(elem);
+          } else {
+            hide(elem);
+          }
+        }
+
+        // Enable/disable only if element has disable-when
+        if (elem.dataset.disableWhen !== undefined) {
+          if (shouldBeDisabled(elem)) {
+            disable(elem);
+          } else {
+            enable(elem);
+          }
+        }
+      });
+
+      stateChanged = reconcileGroupSelections();
+      iterations += 1;
+      // Hard stop to avoid infinite loops in case of conflicting rules.
+    } while (stateChanged && iterations < 5);
+
+    updateTOC2OptionsList();
+    updateTOC2ContentsList();
+  } finally {
+    isUpdatingVisibility = false;
+  }
+}
+
+// Initialization -------------------------------------------------------------
+
+domReady(() => {
+  const selectorOptions = document.querySelectorAll(OPTION_QUERY);
+  const initialState = {};
+
+  // Attach listeners and gather defaults
+  selectorOptions.forEach((option) => {
+    option.addEventListener("click", handleOptionSelect);
+    option.addEventListener("keydown", handleOptionKeydown);
+
+    if (isDefaultOption(option)) {
+      select(option);
+      const { selectorKey: key, selectorValue: value } = option.dataset;
+      if (key && value && initialState[key] === undefined) {
+        initialState[key] = value;
+      }
+    }
+  });
+
+  setState(initialState);
+  updateVisibility();
+
+  // Mark all selector groups as initialized to make them visible
+  document.querySelectorAll(GROUP_QUERY).forEach((group) => {
+    group.classList.add("rocm-docs-selector-initialized");
+  });
+});

docs/extension/rocm_docs_custom/static/selector/utils.js

@@ -0,0 +1,14 @@
+export function domReady(callback) {
+  if (document.readyState !== "loading") {
+    callback();
+  } else {
+    document.addEventListener("DOMContentLoaded", callback, { once: true });
+  }
+}
+
+const DEBUG = true;
+export const logDebug = (...args) => {
+  if (DEBUG) {
+    console.debug("[ROCmDocsSelector]", ...args);
+  }
+};

docs/extension/rocm_docs_custom/table.py

@@ -1,9 +0,0 @@
-from sphinx.util.docutils import SphinxDirective, directives, nodes
-from pathlib import Path
-
-def setup(app):
-    static_assets_dir = Path(__file__).parent / "static"
-    app.config.html_static_path.append(str(static_assets_dir))
-    app.add_css_file("table.css")
-
-    return {"version": "1.0", "parallel_read_safe": True}

docs/extension/rocm_docs_custom/templates/selector-toc2.html

@@ -1,6 +1,7 @@
+<!-- Summary of selected options in secondary TOC -->
 <div class="tocsection onthispage">
-  <i class="fa-solid fa-filter"></i>
-  Options
+  <i class="fa-solid fa-computer"></i>
+  Installation environment
 </div>
 <nav class="page-toc rocm-docs-selector-toc2">
   <ul
@@ -8,6 +9,7 @@
   >
   </ul>
 </nav>
+<!-- Summary of page contents in secondary TOC -->
 <div class="page-toc tocsection onthispage">
   <i class="fa-solid fa-list"></i>
   Contents

docs/extension/rocm_docs_custom/utils.py

@@ -1,5 +1,6 @@
 import json
 import html
+from sphinx.util import logging
 
 def normalize_key(key):
     return key.replace(" ", "_").lower().strip()
@@ -28,3 +29,6 @@ def kv_to_data_attr(name, kv_str, separator="="):
 
     return f'data-{name}="{html.escape(json.dumps(pairs))}"' if pairs else ""
 
+
+logger = logging.getLogger(__name__)
+

docs/how-to/rocm-for-ai/training/benchmark-docker/previous-versions/jax-maxtext-v25.7.rst

@@ -1,366 +0,0 @@
-:orphan:
-
-.. meta::
-   :description: How to train a model using JAX MaxText for ROCm.
-   :keywords: ROCm, AI, LLM, train, jax, torch, Llama, flux, tutorial, docker
-
-******************************************
-Training a model with JAX MaxText on ROCm
-******************************************
-
-.. caution::
-
-   This documentation does not reflect the latest version of ROCm JAX MaxText
-   training performance documentation. See :doc:`../jax-maxtext` for the latest version.
-
-MaxText is a high-performance, open-source framework built on the Google JAX
-machine learning library to train LLMs at scale. The MaxText framework for
-ROCm is an optimized fork of the upstream
-`<https://github.com/AI-Hypercomputer/maxtext>`__ enabling efficient AI workloads
-on AMD MI300X series GPUs.
-
-The MaxText for ROCm training Docker image
-provides a prebuilt environment for training on AMD Instinct MI300X and MI325X GPUs,
-including essential components like JAX, XLA, ROCm libraries, and MaxText utilities.
-It includes the following software components:
-
-.. datatemplate:yaml:: /data/how-to/rocm-for-ai/training/jax-maxtext-benchmark-models.yaml
-
-   {% set dockers = data.dockers %}
-   .. tab-set::
-
-      {% for docker in dockers %}
-      {% set jax_version = docker.components["JAX"] %}
-
-      .. tab-item:: ``{{ docker.pull_tag }}``
-         :sync: {{ docker.pull_tag }}
-
-         .. list-table::
-            :header-rows: 1
-
-            * - Software component
-              - Version
-
-            {% for component_name, component_version in docker.components.items() %}
-            * - {{ component_name }}
-              - {{ component_version }}
-
-            {% endfor %}
-         {% if jax_version == "0.6.0" %}
-         .. note::
-
-            Shardy is a new config in JAX 0.6.0. You might get related errors if it's
-            not configured correctly. For now you can turn it off by setting
-            ``shardy=False`` during the training run. You can also follow the `migration
-            guide <https://docs.jax.dev/en/latest/shardy_jax_migration.html>`__ to enable
-            it.
-         {% endif %}
-
-      {% endfor %}
-
-MaxText with on ROCm provides the following key features to train large language models efficiently:
-
-- Transformer Engine (TE)
-
-- Flash Attention (FA) 3 -- with or without sequence input packing
-
-- GEMM tuning
-
-- Multi-node support
-
-- NANOO FP8 quantization support
-
-.. _amd-maxtext-model-support-v257:
-
-Supported models
-================
-
-The following models are pre-optimized for performance on AMD Instinct MI300
-series GPUs. Some instructions, commands, and available training
-configurations in this documentation might vary by model -- select one to get
-started.
-
-.. datatemplate:yaml:: /data/how-to/rocm-for-ai/training/jax-maxtext-benchmark-models.yaml
-
-   {% set model_groups = data.model_groups %}
-   .. raw:: html
-
-      <div id="vllm-benchmark-ud-params-picker" class="container-fluid">
-         <div class="row gx-0">
-            <div class="col-2 me-1 px-2 model-param-head">Model</div>
-            <div class="row col-10 pe-0">
-      {% for model_group in model_groups %}
-               <div class="col-4 px-2 model-param" data-param-k="model-group" data-param-v="{{ model_group.tag }}" tabindex="0">{{ model_group.group }}</div>
-      {% endfor %}
-            </div>
-         </div>
-
-         <div class="row gx-0 pt-1">
-            <div class="col-2 me-1 px-2 model-param-head">Variant</div>
-            <div class="row col-10 pe-0">
-      {% for model_group in model_groups %}
-         {% set models = model_group.models %}
-         {% for model in models %}
-            {% if models|length % 3 == 0 %}
-               <div class="col-4 px-2 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 px-2 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>
-
-.. note::
-
-   Some models, such as Llama 3, require an external license agreement through
-   a third party (for example, Meta).
-
-System validation
-=================
-
-Before running AI workloads, it's important to validate that your AMD hardware is configured
-correctly and performing optimally.
-
-If you have already validated your system settings, including aspects like NUMA auto-balancing, you
-can skip this step. Otherwise, complete the procedures in the :ref:`System validation and
-optimization <rocm-for-ai-system-optimization>` guide to properly configure your system settings
-before starting training.
-
-To test for optimal performance, consult the recommended :ref:`System health benchmarks
-<rocm-for-ai-system-health-bench>`. This suite of tests will help you verify and fine-tune your
-system's configuration.
-
-Environment setup
-=================
-
-This Docker image is optimized for specific model configurations outlined
-as follows. Performance can vary for other training workloads, as AMD
-doesn’t validate configurations and run conditions outside those described.
-
-Pull the Docker image
----------------------
-
-Use the following command to pull the Docker image from Docker Hub.
-
-.. datatemplate:yaml:: /data/how-to/rocm-for-ai/training/jax-maxtext-benchmark-models.yaml
-
-   {% set dockers = data.dockers %}
-   .. tab-set::
-
-      {% for docker in dockers %}
-      {% set jax_version = docker.components["JAX"] %}
-
-      .. tab-item:: JAX {{ jax_version }}
-         :sync: {{ docker.pull_tag }}
-
-         .. code-block:: shell
-
-            docker pull {{ docker.pull_tag }}
-
-      {% endfor %}
-
-.. _amd-maxtext-multi-node-setup-v257:
-
-Multi-node configuration
-------------------------
-
-See :doc:`/how-to/rocm-for-ai/system-setup/multi-node-setup` to configure your
-environment for multi-node training.
-
-.. _amd-maxtext-get-started-v257:
-
-Benchmarking
-============
-
-Once the setup is complete, choose between two options to reproduce the
-benchmark results:
-
-.. datatemplate:yaml:: /data/how-to/rocm-for-ai/training/jax-maxtext-benchmark-models.yaml
-
-   .. _vllm-benchmark-mad:
-
-   {% set dockers = data.dockers %}
-   {% set model_groups = data.model_groups %}
-   {% for model_group in model_groups %}
-      {% for model in model_group.models %}
-
-   .. container:: model-doc {{model.mad_tag}}
-
-      .. tab-set::
-
-         {% if model.mad_tag and "single-node" in model.doc_options %}
-         .. tab-item:: MAD-integrated benchmarking
-
-            1. 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.
-
-               .. code-block:: shell
-
-                  git clone https://github.com/ROCm/MAD
-                  cd MAD
-                  pip install -r requirements.txt
-
-            2. Use this command to run the performance benchmark test on the {{ model.model }} model
-               using one GPU with the :literal:`{{model.precision}}` data type on the host machine.
-
-               .. code-block:: shell
-
-                  export MAD_SECRETS_HFTOKEN="your personal Hugging Face token to access gated models"
-                  madengine run \
-                      --tags {{model.mad_tag}} \
-                      --keep-model-dir \
-                      --live-output \
-                      --timeout 28800
-
-            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/perf.csv/``.
-         {% endif %}
-
-         .. tab-item:: Standalone benchmarking
-
-            .. rubric:: Download the Docker image and required scripts
-
-            Run the JAX MaxText benchmark tool independently by starting the
-            Docker container as shown in the following snippet.
-
-            .. tab-set::
-               {% for docker in dockers %}
-               {% set jax_version = docker.components["JAX"] %}
-
-               .. tab-item:: JAX {{ jax_version }}
-                  :sync: {{ docker.pull_tag }}
-
-                  .. code-block:: shell
-
-                     docker pull {{ docker.pull_tag }}
-               {% endfor %}
-
-            {% if model.model_repo and "single-node" in model.doc_options %}
-            .. rubric:: Single node training
-
-            1. Set up environment variables.
-
-               .. code-block:: shell
-
-                  export MAD_SECRETS_HFTOKEN=<Your Hugging Face token>
-                  export HF_HOME=<Location of saved/cached Hugging Face models>
-
-               ``MAD_SECRETS_HFTOKEN`` is your Hugging Face access token to access models, tokenizers, and data.
-               See `User access tokens <https://huggingface.co/docs/hub/en/security-tokens>`__.
-
-               ``HF_HOME`` is where ``huggingface_hub`` will store local data. See `huggingface_hub CLI <https://huggingface.co/docs/huggingface_hub/main/en/guides/cli#huggingface-cli-download>`__.
-               If you already have downloaded or cached Hugging Face artifacts, set this variable to that path.
-               Downloaded files typically get cached to ``~/.cache/huggingface``.
-
-            2. Launch the Docker container.
-
-               .. tab-set::
-                  {% for docker in dockers %}
-                  {% set jax_version = docker.components["JAX"] %}
-
-                  .. tab-item:: JAX {{ jax_version }}
-                     :sync: {{ docker.pull_tag }}
-
-                     .. 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 \
-                            -v $HF_HOME:/hf_cache \
-                            -e HF_HOME=/hf_cache \
-                            -e MAD_SECRETS_HFTOKEN=$MAD_SECRETS_HFTOKEN
-                            --shm-size 64G \
-                            --name training_env \
-                            {{ docker.pull_tag }}
-                  {% endfor %}
-
-            3. In the Docker container, clone the ROCm MAD repository and navigate to the
-               benchmark scripts directory at ``MAD/scripts/jax-maxtext``.
-
-               .. code-block:: shell
-
-                  git clone https://github.com/ROCm/MAD
-                  cd MAD/scripts/jax-maxtext
-
-            4. Run the setup scripts to install libraries and datasets needed
-               for benchmarking.
-
-               .. code-block:: shell
-
-                  ./jax-maxtext_benchmark_setup.sh -m {{ model.model_repo }}
-
-            5. To run the training benchmark without quantization, use the following command:
-
-               .. code-block:: shell
-
-                  ./jax-maxtext_benchmark_report.sh -m {{ model.model_repo }}
-
-               For quantized training, use the following command:
-
-               .. code-block:: shell
-
-                  ./jax-maxtext_benchmark_report.sh -m {{ model.model_repo }} -q nanoo_fp8
-
-            {% endif %}
-            {% if model.multinode_training_script and "multi-node" in model.doc_options %}
-            .. rubric:: Multi-node training
-
-            The following examples use SLURM to run on multiple nodes.
-
-            .. note::
-
-               The following scripts will launch the Docker container and run the
-               benchmark. Run them outside of any Docker container.
-
-            1. Make sure ``$HF_HOME`` is set before running the test. See
-               `ROCm benchmarking <https://github.com/ROCm/MAD/blob/develop/scripts/jax-maxtext/gpu-rocm/readme.md>`__
-               for more details on downloading the Llama models before running the
-               benchmark.
-
-            2. To run multi-node training for {{ model.model }},
-               use the
-               `multi-node training script <https://github.com/ROCm/MAD/blob/develop/scripts/jax-maxtext/gpu-rocm/{{ model.multinode_training_script }}>`__
-               under the ``scripts/jax-maxtext/gpu-rocm/`` directory.
-
-            3. Run the multi-node training benchmark script.
-
-               .. code-block:: shell
-
-                  sbatch -N <num_nodes> {{ model.multinode_training_script }}
-
-         {% else %}
-            .. rubric:: Multi-node training
-
-            For multi-node training examples, choose a model from :ref:`amd-maxtext-model-support-v257`
-            with an available `multi-node training script <https://github.com/ROCm/MAD/tree/develop/scripts/jax-maxtext/gpu-rocm>`__.
-         {% endif %}
-      {% endfor %}
-   {% endfor %}
-
-Further reading
-===============
-
-- To learn more about MAD and the ``madengine`` CLI, see the `MAD usage guide <https://github.com/ROCm/MAD?tab=readme-ov-file#usage-guide>`__.
-
-- To learn more about system settings and management practices to configure your system for
-  AMD Instinct MI300X series GPUs, see `AMD Instinct MI300X system optimization <https://instinct.docs.amd.com/projects/amdgpu-docs/en/latest/system-optimization/mi300x.html>`_.
-
-- For a list of other ready-made Docker images for AI with ROCm, see
-  `AMD Infinity Hub <https://www.amd.com/en/developer/resources/infinity-hub.html#f-amd_hub_category=AI%20%26%20ML%20Models>`_.
-
-Previous versions
-=================
-
-See :doc:`jax-maxtext-history` to find documentation for previous releases
-of the ``ROCm/jax-training`` Docker image.

docs/how-to/rocm-for-ai/training/benchmark-docker/previous-versions/primus-megatron-v25.8.rst

@@ -1,667 +0,0 @@
-: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 Primus and Megatron-LM
-********************************************
-
-.. caution::
-
-   This documentation does not reflect the latest version of ROCm Megatron-LM
-   training performance documentation. See :doc:`../primus-megatron` for the latest version.
-
-`Primus <https://github.com/AMD-AGI/Primus>`__ is a unified and flexible
-LLM training framework designed to streamline training. It streamlines LLM
-training on AMD Instinct GPUs using a modular, reproducible configuration paradigm.
-Primus is backend-agnostic and supports multiple training engines -- including Megatron.
-
-.. note::
-
-   Primus with Megatron is designed to replace the :doc:`ROCm Megatron-LM training <../megatron-lm>` workflow.
-   To learn how to migrate workloads from Megatron-LM to Primus with Megatron,
-   see :doc:`megatron-lm-primus-migration-guide`.
-
-For ease of use, AMD provides a ready-to-use Docker image for MI300 series GPUs
-containing essential components for Primus and Megatron-LM. This Docker is powered by Primus
-Turbo optimizations for performance; this release adds support for Primus Turbo
-with optimized attention and grouped GEMM kernels.
-
-.. note::
-
-   This Docker environment is based on Python 3.10 and Ubuntu 22.04. For an alternative environment with
-   Python 3.12 and Ubuntu 24.04, see the :doc:`previous ROCm Megatron-LM v25.6 Docker release <megatron-lm-v25.6>`.
-
-.. datatemplate:yaml:: /data/how-to/rocm-for-ai/training/previous-versions/primus-megatron-v25.8-benchmark-models.yaml
-
-   {% set dockers = data.dockers %}
-   {% set docker = dockers[0] %}
-   .. list-table::
-      :header-rows: 1
-
-      * - Software component
-        - Version
-
-      {% for component_name, component_version in docker.components.items() %}
-      * - {{ component_name }}
-        - {{ component_version }}
-      {% endfor %}
-
-.. _amd-primus-megatron-lm-model-support:
-
-Supported models
-================
-
-The following models are pre-optimized for performance on AMD Instinct MI300X series GPUs.
-Some instructions, commands, and training examples in this documentation might
-vary by model -- select one to get started.
-
-.. datatemplate:yaml:: /data/how-to/rocm-for-ai/training/previous-versions/primus-megatron-v25.8-benchmark-models.yaml
-
-   {% set model_groups = data.model_groups %}
-   .. raw:: html
-
-      <div id="vllm-benchmark-ud-params-picker" class="container-fluid">
-         <div class="row gx-0">
-            <div class="col-2 me-1 px-2 model-param-head">Model</div>
-            <div class="row col-10 pe-0">
-      {% for model_group in model_groups %}
-               <div class="col-3 px-2 model-param" data-param-k="model-group" data-param-v="{{ model_group.tag }}" tabindex="0">{{ model_group.group }}</div>
-      {% endfor %}
-            </div>
-         </div>
-
-         <div class="row gx-0 pt-1">
-            <div class="col-2 me-1 px-2 model-param-head">Variant</div>
-            <div class="row col-10 pe-0">
-      {% for model_group in model_groups %}
-         {% set models = model_group.models %}
-         {% for model in models %}
-            {% if models|length % 3 == 0 %}
-               <div class="col-4 px-2 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 px-2 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>
-
-.. note::
-
-   Some models, such as Llama, require an external license agreement through
-   a third party (for example, Meta).
-
-System validation
-=================
-
-Before running AI workloads, it's important to validate that your AMD hardware is configured
-correctly and performing optimally.
-
-If you have already validated your system settings, including aspects like NUMA auto-balancing, you
-can skip this step. Otherwise, complete the procedures in the :ref:`System validation and
-optimization <rocm-for-ai-system-optimization>` guide to properly configure your system settings
-before starting training.
-
-To test for optimal performance, consult the recommended :ref:`System health benchmarks
-<rocm-for-ai-system-health-bench>`. This suite of tests will help you verify and fine-tune your
-system's configuration.
-
-.. _mi300x-amd-primus-megatron-lm-training:
-
-.. datatemplate:yaml:: /data/how-to/rocm-for-ai/training/previous-versions/primus-megatron-v25.8-benchmark-models.yaml
-
-   {% set dockers = data.dockers %}
-      {% set docker = dockers[0] %}
-
-   Environment setup
-   =================
-
-   Use the following instructions to set up the environment, configure the script to train models, and
-   reproduce the benchmark results on MI300X series GPUs with the ``{{ docker.pull_tag }}`` image.
-
-   .. _amd-primus-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 {{ docker.pull_tag }}
-
-   2. Launch the Docker container.
-
-      .. code-block:: shell
-
-         docker run -it \
-             --device /dev/dri \
-             --device /dev/kfd \
-             --device /dev/infiniband \
-             --network host --ipc host \
-             --group-add video \
-             --cap-add SYS_PTRACE \
-             --security-opt seccomp=unconfined \
-             --privileged \
-             -v $HOME:$HOME \
-             --shm-size 128G \
-             --name primus_training_env \
-             {{ docker.pull_tag }}
-
-3. Use these commands if you exit the ``primus_training_env`` container and need to return to it.
-
-   .. code-block:: shell
-
-      docker start primus_training_env
-      docker exec -it primus_training_env bash
-
-The Docker container hosts verified commit ``927a717`` of the `Primus
-<https://github.com/AMD-AGI/Primus/tree/927a71702784347a311ca48fd45f0f308c6ef6dd>`__ repository.
-
-.. _amd-primus-megatron-lm-environment-setup:
-
-Configuration
-=============
-
-Primus defines a training configuration in YAML for each model in
-`examples/megatron/configs <https://github.com/AMD-AGI/Primus/tree/927a71702784347a311ca48fd45f0f308c6ef6dd/examples/megatron/configs>`__.
-
-.. datatemplate:yaml:: /data/how-to/rocm-for-ai/training/previous-versions/primus-megatron-v25.8-benchmark-models.yaml
-
-   {% set model_groups = data.model_groups %}
-   {% for model_group in model_groups %}
-      {% for model in model_group.models %}
-   .. container:: model-doc {{ model.mad_tag }}
-
-      To update training parameters for {{ model.model }}, you can update ``examples/megatron/configs/{{ model.config_name }}``.
-      Note that training configuration YAML files for other models follow this naming convention.
-
-      {% endfor %}
-   {% endfor %}
-
-.. note::
-
-   See :ref:`Key options <amd-primus-megatron-lm-benchmark-test-vars>` for more information on configuration options.
-
-Dataset options
----------------
-
-You can use either mock data or real data for training.
-
-* Mock data can be useful for testing and validation. Use the ``mock_data`` field to toggle between mock and real data. The default
-  value is ``true`` for enabled.
-
-  .. code-block:: yaml
-
-     mock_data: true
-
-* If you're using a real dataset, update the ``train_data_path`` field to point to the location of your dataset.
-
-  .. code-block:: bash
-
-     mock_data: false
-     train_data_path: /path/to/your/dataset
-
-  Ensure that the files are accessible inside the Docker container.
-
-.. _amd-primus-megatron-lm-tokenizer:
-
-Tokenizer
----------
-
-Set the ``HF_TOKEN`` environment variable with
-right permissions to access the tokenizer for each model.
-
-.. code-block:: bash
-
-   # Export your HF_TOKEN in the workspace
-   export HF_TOKEN=<your_hftoken>
-
-.. note::
-
-   In Primus, each model uses a tokenizer from Hugging Face. For example, Llama
-   3.1 8B model uses ``tokenizer_model: meta-llama/Llama-3.1-8B`` and
-   ``tokenizer_type: Llama3Tokenizer`` defined in the `llama3.1-8B model
-   <https://github.com/AMD-AGI/Primus/blob/927a71702784347a311ca48fd45f0f308c6ef6dd/examples/megatron/configs/llama3.1_8B-pretrain.yaml>`__
-   definition.
-
-.. _amd-primus-megatron-lm-run-training:
-
-Run training
-============
-
-Use the following example commands to set up the environment, configure
-:ref:`key options <amd-primus-megatron-lm-benchmark-test-vars>`, and run training on
-MI300X series GPUs with the AMD Megatron-LM environment.
-
-Single node training
---------------------
-
-To run training on a single node, navigate to ``/workspace/Primus`` and use the following setup command:
-
-.. code-block:: shell
-
-   pip install -r requirements.txt
-   export HSA_NO_SCRATCH_RECLAIM=1
-   export NVTE_CK_USES_BWD_V3=1
-
-.. container:: model-doc primus_pyt_megatron_lm_train_llama-3.3-70b
-
-   Once setup is complete, run the appropriate training command.
-   The following run commands are tailored to Llama 3.3 70B.
-   See :ref:`amd-primus-megatron-lm-model-support` to switch to another available model.
-
-   To run pre-training for Llama 3.3 70B BF16, run:
-
-   .. code-block:: shell
-
-      EXP=examples/megatron/configs/llama3.3_70B-pretrain.yaml \
-      bash ./examples/run_pretrain.sh \
-          --micro_batch_size 2 \
-          --global_batch_size 16 \
-          --train_iters 50
-
-.. container:: model-doc primus_pyt_megatron_lm_train_llama-3.1-8b
-
-   Once setup is complete, run the appropriate training command.
-   The following run commands are tailored to Llama 3.1 8B.
-   See :ref:`amd-primus-megatron-lm-model-support` to switch to another available model.
-
-   To run pre-training for Llama 3.1 8B FP8, run:
-
-   .. code-block:: shell
-
-      EXP=examples/megatron/configs/llama3.1_8B-pretrain.yaml \
-      bash ./examples/run_pretrain.sh \
-          --train_iters 50 \
-          --fp8 hybrid
-
-   For Llama 3.1 8B BF16, use the following command:
-
-   .. code-block:: shell
-
-      EXP=examples/megatron/configs/llama3.1_8B-pretrain.yaml \
-      bash ./examples/run_pretrain.sh --train_iters 50
-
-.. container:: model-doc primus_pyt_megatron_lm_train_llama-3.1-70b
-
-   Once setup is complete, run the appropriate training command.
-   The following run commands are tailored to Llama 3.1 70B.
-   See :ref:`amd-primus-megatron-lm-model-support` to switch to another available model.
-
-   To run pre-training for Llama 3.1 70B BF16, run:
-
-   .. code-block:: shell
-
-      EXP=examples/megatron/configs/llama3.1_70B-pretrain.yaml \
-      bash ./examples/run_pretrain.sh \
-           --train_iters 50
-
-   To run the training on a single node for Llama 3.1 70B FP8 with proxy, use the following command:
-
-   .. code-block:: shell
-
-      EXP=examples/megatron/configs/llama3.1_70B-pretrain.yaml \
-      bash ./examples/run_pretrain.sh \
-          --train_iters 50 \
-          --num_layers 40 \
-          --fp8 hybrid
-
-   .. note::
-
-      Use two or more nodes to run the *full* Llama 70B model with FP8 precision.
-
-.. container:: model-doc primus_pyt_megatron_lm_train_llama-2-7b
-
-   Once setup is complete, run the appropriate training command.
-   The following run commands are tailored to Llama 2 7B.
-   See :ref:`amd-primus-megatron-lm-model-support` to switch to another available model.
-
-   To run pre-training for Llama 2 7B FP8, run:
-
-   .. code-block:: shell
-
-      EXP=examples/megatron/configs/llama2_7B-pretrain.yaml \
-      bash ./examples/run_pretrain.sh \
-          --train_iters 50 \
-          --fp8 hybrid
-
-   To run pre-training for Llama 2 7B BF16, run:
-
-   .. code-block:: shell
-
-      EXP=examples/megatron/configs/llama2_7B-pretrain.yaml \
-      bash ./examples/run_pretrain.sh --train_iters 50
-
-.. container:: model-doc primus_pyt_megatron_lm_train_llama-2-70b
-
-   Once setup is complete, run the appropriate training command.
-   The following run commands are tailored to Llama 2 70B.
-   See :ref:`amd-primus-megatron-lm-model-support` to switch to another available model.
-
-   To run pre-training for Llama 2 70B BF16, run:
-
-   .. code-block:: shell
-
-      EXP=examples/megatron/configs/llama2_70B-pretrain.yaml \
-      bash ./examples/run_pretrain.sh --train_iters 50
-
-.. container:: model-doc primus_pyt_megatron_lm_train_deepseek-v3-proxy
-
-   Once setup is complete, run the appropriate training command.
-   The following run commands are tailored to DeepSeek-V3.
-   See :ref:`amd-primus-megatron-lm-model-support` to switch to another available model.
-
-   To run training on a single node for DeepSeek-V3 (MoE with expert parallel) with 3-layer proxy,
-   use the following command:
-
-   .. code-block:: shell
-
-      EXP=examples/megatron/configs/deepseek_v3-pretrain.yaml \
-      bash examples/run_pretrain.sh \
-          --num_layers 3 \
-          --moe_layer_freq 1 \
-          --train_iters 50
-
-.. container:: model-doc primus_pyt_megatron_lm_train_deepseek-v2-lite-16b
-
-   Once setup is complete, run the appropriate training command.
-   The following run commands are tailored to DeepSeek-V2-Lite.
-   See :ref:`amd-primus-megatron-lm-model-support` to switch to another available model.
-
-   To run training on a single node for DeepSeek-V2-Lite (MoE with expert parallel),
-   use the following command:
-
-   .. code-block:: shell
-
-      EXP=examples/megatron/configs/deepseek_v2_lite-pretrain.yaml \
-      bash examples/run_pretrain.sh \
-          --global_batch_size 256 \
-          --train_iters 50
-
-.. container:: model-doc primus_pyt_megatron_lm_train_mixtral-8x7b
-
-   Once setup is complete, run the appropriate training command.
-   The following run commands are tailored to Mixtral 8x7B.
-   See :ref:`amd-primus-megatron-lm-model-support` to switch to another available model.
-
-   To run training on a single node for Mixtral 8x7B (MoE with expert parallel),
-   use the following command:
-
-   .. code-block:: shell
-
-      EXP=examples/megatron/configs/mixtral_8x7B_v0.1-pretrain.yaml \
-      bash examples/run_pretrain.sh --train_iters 50
-
-.. container:: model-doc primus_pyt_megatron_lm_train_mixtral-8x22b-proxy
-
-   Once setup is complete, run the appropriate training command.
-   The following run commands are tailored to Mixtral 8x22B.
-   See :ref:`amd-primus-megatron-lm-model-support` to switch to another available model.
-
-   To run training on a single node for Mixtral 8x22B (MoE with expert parallel) with 4-layer proxy,
-   use the following command:
-
-   .. code-block:: shell
-
-      EXP=examples/megatron/configs/mixtral_8x22B_v0.1-pretrain.yaml \
-      bash examples/run_pretrain.sh \
-          --num_layers 4 \
-          --pipeline_model_parallel_size 1 \
-          --micro_batch_size 1 \
-          --global_batch_size 16 \
-          --train_iters 50
-
-.. container:: model-doc primus_pyt_megatron_lm_train_qwen2.5-7b
-
-   Once setup is complete, run the appropriate training command.
-   The following run commands are tailored to Qwen 2.5 7B.
-   See :ref:`amd-primus-megatron-lm-model-support` to switch to another available model.
-
-   To run training on a single node for Qwen 2.5 7B BF16, use the following
-   command:
-
-   .. code-block:: shell
-
-      EXP=examples/megatron/configs/qwen2.5_7B-pretrain.yaml \
-      bash examples/run_pretrain.sh --train_iters 50
-
-   For FP8, use the following command.
-
-   .. code-block:: shell
-
-      EXP=examples/megatron/configs/qwen2.5_7B-pretrain.yaml \
-      bash examples/run_pretrain.sh \
-          --train_iters 50 \
-          --fp8 hybrid
-
-.. container:: model-doc primus_pyt_megatron_lm_train_qwen2.5-72b
-
-   Once setup is complete, run the appropriate training command.
-   The following run commands are tailored to Qwen 2.5 72B.
-   See :ref:`amd-primus-megatron-lm-model-support` to switch to another available model.
-
-   To run the training on a single node for Qwen 2.5 72B BF16, use the following command.
-
-   .. code-block:: shell
-
-      EXP=examples/megatron/configs/qwen2.5_72B-pretrain.yaml \
-      bash examples/run_pretrain.sh --train_iters 50
-
-.. _amd-primus-megatron-multi-node-examples:
-
-Multi-node training examples
-----------------------------
-
-Refer to :doc:`/how-to/rocm-for-ai/system-setup/multi-node-setup` to configure your environment for multi-node
-training.
-
-To run training on multiple nodes, you can use the
-`run_slurm_pretrain.sh <https://github.com/AMD-AGI/Primus/blob/927a71702784347a311ca48fd45f0f308c6ef6dd/examples/run_slurm_pretrain.sh>`__
-to launch the multi-node workload. Use the following steps to setup your environment:
-
-.. datatemplate:yaml:: /data/how-to/rocm-for-ai/training/previous-versions/primus-megatron-v25.8-benchmark-models.yaml
-
-   {% set dockers = data.dockers %}
-   {% set docker = dockers[0] %}
-
-   .. code-block:: shell
-
-      cd /workspace/Primus/
-      export DOCKER_IMAGE={{ docker.pull_tag }}
-      export HF_TOKEN=<your_HF_token>
-      export HSA_NO_SCRATCH_RECLAIM=1
-      export NVTE_CK_USES_BWD_V3=1
-      export NCCL_IB_HCA=<your_NCCL_IB_HCA> # specify which RDMA interfaces to use for communication
-      export NCCL_SOCKET_IFNAME=<your_NCCL_SOCKET_IFNAME> # your Network Interface
-      export GLOO_SOCKET_IFNAME=<your_GLOO_SOCKET_IFNAME> # your Network Interface
-      export NCCL_IB_GID_INDEX=3 # Set InfiniBand GID index for NCCL communication. Default is 3 for ROCE
-
-.. note::
-
-   * Make sure 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 Docker container.
-   * If ``NCCL_IB_HCA`` and ``NCCL_SOCKET_IFNAME`` are not set, Primus will try to auto-detect. However, since NICs can vary accross different cluster, it is encouraged to explicitly export your NCCL parameters for the cluster.
-   * To find your network interface, you can use ``ip a``.
-   * To find RDMA interfaces, you can use ``ibv_devices`` to get the list of all the RDMA/IB  devices.
-
-.. container:: model-doc primus_pyt_megatron_lm_train_llama-3.3-70b
-
-   To train Llama 3.3 70B FP8 on 8 nodes, run:
-
-   .. code-block:: shell
-
-      NNODES=8 EXP=examples/megatron/configs/llama3.3_70B-pretrain.yaml \
-      bash examples/run_slurm_pretrain.sh \
-          --micro_batch_size 1 \
-          --global_batch_size 256 \
-          --recompute_num_layers 80 \
-          --fp8 hybrid
-
-   To train Llama 3.3 70B BF16 on 8 nodes, run:
-
-   .. code-block:: shell
-
-      NNODES=8 EXP=examples/megatron/configs/llama3.3_70B-pretrain.yaml \
-      bash examples/run_slurm_pretrain.sh \
-          --micro_batch_size 1 \
-          --global_batch_size 256 \
-          --recompute_num_layers 12
-
-.. container:: model-doc primus_pyt_megatron_lm_train_llama-3.1-8b
-
-   To train Llama 3.1 8B FP8 on 8 nodes, run:
-
-   .. code-block:: shell
-
-      # Adjust the training parameters. For e.g., `global_batch_size: 8 * #single_node_bs` for 8 nodes in this case
-      NNODES=8 EXP=examples/megatron/configs/llama3.1_8B-pretrain.yaml \
-      bash ./examples/run_slurm_pretrain.sh \
-          --global_batch_size 1024 \
-          --fp8 hybrid
-
-.. container:: model-doc primus_pyt_megatron_lm_train_llama-3.1-70b
-
-   To train Llama 3.1 70B FP8 on 8 nodes, run:
-
-   .. code-block:: shell
-
-      NNODES=8 EXP=examples/megatron/configs/llama3.1_70B-pretrain.yaml \
-      bash examples/run_slurm_pretrain.sh \
-          --micro_batch_size 1 \
-          --global_batch_size 256 \
-          --recompute_num_layers 80 \
-          --fp8 hybrid
-
-   To train Llama 3.1 70B BF16 on 8 nodes, run:
-
-   .. code-block:: shell
-
-      NNODES=8 EXP=examples/megatron/configs/llama3.1_70B-pretrain.yaml \
-      bash examples/run_slurm_pretrain.sh \
-          --micro_batch_size 1 \
-          --global_batch_size 256 \
-          --recompute_num_layers 12
-
-.. container:: model-doc primus_pyt_megatron_lm_train_llama-2-7b
-
-   To train Llama 2 8B FP8 on 8 nodes, run:
-
-   .. code-block:: shell
-
-      # Adjust the training parameters. For e.g., `global_batch_size: 8 * #single_node_bs` for 8 nodes in this case
-      NNODES=8 EXP=examples/megatron/configs/llama2_7B-pretrain.yaml bash ./examples/run_slurm_pretrain.sh --global_batch_size 2048 --fp8 hybrid
-
-.. container:: model-doc primus_pyt_megatron_lm_train_llama-2-70b
-
-   To train Llama 2 70B FP8 on 8 nodes, run:
-
-   .. code-block:: shell
-
-      NNODES=8 EXP=examples/megatron/configs/llama2_70B-pretrain.yaml \
-      bash examples/run_slurm_pretrain.sh \
-          --micro_batch_size 2 \
-          --global_batch_size 256 \
-          --recompute_num_layers 80 \
-          --fp8 hybrid
-
-   To train Llama 2 70B BF16 on 8 nodes, run:
-
-   .. code-block:: shell
-
-      NNODES=8 EXP=examples/megatron/configs/llama2_70B-pretrain.yaml \
-      bash ./examples/run_slurm_pretrain.sh \
-          --micro_batch_size 2 \
-          --global_batch_size 1536 \
-          --recompute_num_layers 12
-
-.. container:: model-doc primus_pyt_megatron_lm_train_mixtral-8x7b
-
-   To train Mixtral 8x7B BF16 on 8 nodes, run:
-
-   .. code-block:: shell
-
-      NNODES=8 EXP=examples/megatron/configs/mixtral_8x7B_v0.1-pretrain.yaml \
-      bash examples/run_slurm_pretrain.sh \
-          --micro_batch_size 2 \
-          --global_batch_size 256
-
-.. container:: model-doc primus_pyt_megatron_lm_train_qwen2.5-72b
-
-   To train Qwen2.5 72B FP8 on 8 nodes, run:
-
-   .. code-block:: shell
-
-      NNODES=8 EXP=examples/megatron/configs/qwen2.5_72B-pretrain.yaml \
-      bash examples/run_slurm_pretrain.sh \
-          --micro_batch_size 4 \
-          --global_batch_size 256 \
-          --recompute_num_layers 80 \
-          --fp8 hybrid
-
-.. _amd-primus-megatron-lm-benchmark-test-vars:
-
-Key options
------------
-
-The following are key options to take note of
-
-fp8
-  ``hybrid`` enables FP8 GEMMs.
-
-use_torch_fsdp2
-  ``use_torch_fsdp2: 1``  enables torch fsdp-v2. If FSDP is enabled,
-  set ``use_distributed_optimizer`` and ``overlap_param_gather`` to ``false``.
-
-profile
-  To enable PyTorch profiling, set these parameters:
-
-  .. code-block:: yaml
-
-     profile: true
-     use_pytorch_profiler: true
-     profile_step_end: 7
-     profile_step_start: 6
-
-train_iters
-  The total number of iterations (default: 50).
-
-mock_data
-  True by default.
-
-micro_batch_size
-  Micro batch size.
-
-global_batch_size
-  Global batch size.
-
-recompute_granularity
-  For activation checkpointing.
-
-num_layers
-  For using a reduced number of layers as with proxy models.
-
-Further reading
-===============
-
-- For an introduction to Primus, see `Primus: A Lightweight, Unified Training
-  Framework for Large Models on AMD GPUs <https://rocm.blogs.amd.com/software-tools-optimization/primus/README.html>`__.
-
-- To learn more about system settings and management practices to configure your system for
-  AMD Instinct MI300X series GPUs, see `AMD Instinct MI300X system optimization <https://instinct.docs.amd.com/projects/amdgpu-docs/en/latest/system-optimization/mi300x.html>`_.
-
-- For a list of other ready-made Docker images for AI with ROCm, see
-  `AMD Infinity Hub <https://www.amd.com/en/developer/resources/infinity-hub.html#f-amd_hub_category=AI%20%26%20ML%20Models>`_.
-
-Previous versions
-=================
-
-See :doc:`megatron-lm-history` to find documentation for previous releases
-of the ``ROCm/megatron-lm`` Docker image.
-
-This training environment now uses Primus with Megatron as the primary
-configuration. Limited support for the legacy ROCm Megatron-LM is still
-available; see the :doc:`../megatron-lm` documentation.

docs/how-to/rocm-for-ai/training/benchmark-docker/previous-versions/primus-pytorch-v25.8.rst

@@ -1,312 +0,0 @@
-: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 Primus and PyTorch
-****************************************
-
-.. caution::
-
-   This documentation does not reflect the latest version of ROCm Primus PyTorch training
-   performance benchmark documentation. See :doc:`../primus-pytorch` for the latest version.
-
-`Primus <https://github.com/AMD-AGI/Primus>`__ is a unified and flexible
-LLM training framework designed to streamline training. It streamlines LLM
-training on AMD Instinct GPUs using a modular, reproducible configuration paradigm.
-Primus now supports the PyTorch torchtitan backend.
-
-.. note::
-
-   Primus with the PyTorch torchtitan backend is designed to replace the :doc:`ROCm PyTorch training <../pytorch-training>` workflow.
-   See :doc:`../pytorch-training` to see steps to run workloads without Primus.
-
-.. datatemplate:yaml:: /data/how-to/rocm-for-ai/training/previous-versions/primus-pytorch-v25.8-benchmark-models.yaml
-
-   {% set dockers = data.dockers %}
-   {% set docker = dockers[0] %}
-   For ease of use, AMD provides a ready-to-use Docker image -- ``{{
-   docker.pull_tag }}`` -- for MI300X series GPUs containing essential
-   components for Primus and PyTorch training with
-   Primus Turbo optimizations.
-
-   .. list-table::
-      :header-rows: 1
-
-      * - Software component
-        - Version
-
-      {% for component_name, component_version in docker.components.items() %}
-      * - {{ component_name }}
-        - {{ component_version }}
-      {% endfor %}
-
-.. _amd-primus-pytorch-model-support-v258:
-
-Supported models
-================
-
-The following models are pre-optimized for performance on the AMD Instinct MI325X and MI300X GPUs.
-Some instructions, commands, and training recommendations in this documentation might
-vary by model -- select one to get started.
-
-.. datatemplate:yaml:: /data/how-to/rocm-for-ai/training/previous-versions/primus-pytorch-v25.8-benchmark-models.yaml
-
-   {% set unified_docker = data.dockers[0] %}
-   {% set model_groups = data.model_groups %}
-   .. raw:: html
-
-      <div id="vllm-benchmark-ud-params-picker" class="container-fluid">
-         <div class="row gx-0" style="display: none;">
-            <div class="col-2 me-1 px-2 model-param-head">Model</div>
-            <div class="row col-10 pe-0">
-      {% for model_group in model_groups %}
-               <div class="col-3 px-2 model-param" data-param-k="model-group" data-param-v="{{ model_group.tag }}" tabindex="0">{{ model_group.group }}</div>
-      {% endfor %}
-            </div>
-         </div>
-
-         <div class="row gx-0 pt-1">
-            <div class="col-2 me-1 px-2 model-param-head">Model</div>
-            <div class="row col-10 pe-0">
-      {% for model_group in model_groups %}
-         {% set models = model_group.models %}
-         {% for model in models %}
-            {% if models|length % 3 == 0 %}
-               <div class="col-4 px-2 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 px-2 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>
-
-.. seealso::
-
-   For additional workloads, including Llama 3.3, Llama 3.2, Llama 2, GPT OSS, Qwen, and Flux models,
-   see the documentation :doc:`../pytorch-training` (without Primus)
-
-.. _amd-primus-pytorch-performance-measurements-v258:
-
-System validation
-=================
-
-Before running AI workloads, it's important to validate that your AMD hardware is configured
-correctly and performing optimally.
-
-If you have already validated your system settings, including aspects like NUMA auto-balancing, you
-can skip this step. Otherwise, complete the procedures in the :ref:`System validation and
-optimization <rocm-for-ai-system-optimization>` guide to properly configure your system settings
-before starting training.
-
-To test for optimal performance, consult the recommended :ref:`System health benchmarks
-<rocm-for-ai-system-health-bench>`. This suite of tests will help you verify and fine-tune your
-system's configuration.
-
-This Docker image is optimized for specific model configurations outlined
-below. Performance can vary for other training workloads, as AMD
-doesn’t test configurations and run conditions outside those described.
-
-.. datatemplate:yaml:: /data/how-to/rocm-for-ai/training/previous-versions/primus-pytorch-v25.8-benchmark-models.yaml
-
-   {% set unified_docker = data.dockers[0] %}
-
-   Pull the Docker image
-   =====================
-
-   Use the following command to pull the `Docker image <{{ unified_docker.docker_hub_url }}>`_ from Docker Hub.
-
-   .. code-block:: shell
-
-      docker pull {{ unified_docker.pull_tag }}
-
-   Run training
-   ============
-
-   {% set model_groups = data.model_groups %}
-
-   Once the setup is complete, choose between the following two workflows to start benchmarking training.
-   For fine-tuning workloads and multi-node training examples, see :doc:`../pytorch-training` (without Primus).
-
-   .. tab-set::
-
-      .. tab-item:: MAD-integrated benchmarking
-
-   {% for model_group in model_groups %}
-      {% for model in model_group.models %}
-
-         .. container:: model-doc {{ model.mad_tag }}
-
-            The following run command is tailored to {{ model.model }}.
-            See :ref:`amd-primus-pytorch-model-support-v258` to switch to another available model.
-
-            1. 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.
-
-               .. code-block:: shell
-
-                  git clone https://github.com/ROCm/MAD
-                  cd MAD
-                  pip install -r requirements.txt
-
-            2. For example, use this command to run the performance benchmark test on the {{ model.model }} model
-               using one node with the {{ model.precision }} data type on the host machine.
-
-               .. code-block:: shell
-
-                  export MAD_SECRETS_HFTOKEN="your personal Hugging Face token to access gated models"
-                  madengine run \
-                      --tags {{ model.mad_tag }} \
-                      --keep-model-dir \
-                      --live-output \
-                      --timeout 28800
-
-               MAD launches a Docker container with the name
-               ``container_ci-{{ model.mad_tag }}``. The latency and throughput reports of the
-               model are collected in ``~/MAD/perf.csv``.
-
-      {% endfor %}
-   {% endfor %}
-
-      .. tab-item:: Standalone benchmarking
-
-   {% for model_group in model_groups %}
-      {% for model in model_group.models %}
-
-         .. container:: model-doc {{ model.mad_tag }}
-
-            The following run commands are tailored to {{ model.model }}.
-            See :ref:`amd-primus-pytorch-model-support-v258` to switch to another available model.
-
-            .. rubric:: 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 {{ unified_docker.pull_tag }}
-
-            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 $HOME/.ssh:/root/.ssh \
-                      --shm-size 64G \
-                      --name training_env \
-                      {{ unified_docker.pull_tag }}
-
-               Use these commands if you exit the ``training_env`` container and need to return to it.
-
-               .. code-block:: shell
-
-                  docker start training_env
-                  docker exec -it training_env bash
-
-            3. In the Docker container, clone the `<https://github.com/ROCm/MAD>`__
-               repository and navigate to the benchmark scripts directory
-               ``/workspace/MAD/scripts/pytorch_train``.
-
-               .. code-block:: shell
-
-                  git clone https://github.com/ROCm/MAD
-                  cd MAD/scripts/pytorch_train
-
-            .. rubric:: Prepare training datasets and dependencies
-
-            1. The following benchmarking examples require downloading models and datasets
-               from Hugging Face. To ensure successful access to gated repos, set your
-               ``HF_TOKEN``.
-
-               .. code-block:: shell
-
-                  export HF_TOKEN=$your_personal_hugging_face_access_token
-
-            2. Run the setup script to install libraries and datasets needed for benchmarking.
-
-               .. code-block:: shell
-
-                  ./pytorch_benchmark_setup.sh
-
-            .. rubric:: 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 pretrain \
-                   -m {{ model.model_repo }} \
-                   -p $datatype \
-                   -s $sequence_length
-
-
-            .. list-table::
-               :header-rows: 1
-
-               * - Name
-                 - Options
-                 - Description
-
-               {% for mode in available_modes %}
-               * - {% if loop.first %}``$training_mode``{% endif %}
-                 - ``{{ mode }}``
-                 - {{ training_mode_descs[mode] }}
-               {% endfor %}
-
-               * - ``$datatype``
-                 - ``BF16``{% if model.mad_tag == "primus_pyt_train_llama-3.1-8b" %} or ``FP8``{% endif %}
-                 - Currently, only Llama 3.1 8B supports FP8 precision.
-
-               * - ``$sequence_length``
-                 - Sequence length for the language model.
-                 - Between 2048 and 8192. 8192 by default.
-
-            .. rubric:: Benchmarking examples
-
-            Use the following command to run train {{ model.model }} with BF16 precision using Primus torchtitan.
-
-            .. code-block:: shell
-
-               ./pytorch_benchmark_report.sh -m {{ model.model_repo }}
-
-            To train {{ model.model }} with FP8 precision, use the following command.
-
-            .. code-block:: shell
-
-               ./pytorch_benchmark_report.sh -m {{ model.model_repo }} -p FP8
-      {% endfor %}
-   {% endfor %}
-
-Further reading
-===============
-
-- For an introduction to Primus, see `Primus: A Lightweight, Unified Training
-  Framework for Large Models on AMD GPUs <https://rocm.blogs.amd.com/software-tools-optimization/primus/README.html>`__.
-
-- To learn more about MAD and the ``madengine`` CLI, see the `MAD usage guide <https://github.com/ROCm/MAD?tab=readme-ov-file#usage-guide>`__.
-
-- To learn more about system settings and management practices to configure your system for
-  AMD Instinct MI300X series GPUs, see `AMD Instinct MI300X system optimization <https://instinct.docs.amd.com/projects/amdgpu-docs/en/latest/system-optimization/mi300x.html>`_.
-
-- For a list of other ready-made Docker images for AI with ROCm, see
-  `AMD Infinity Hub <https://www.amd.com/en/developer/resources/infinity-hub.html#f-amd_hub_category=AI%20%26%20ML%20Models>`_.
-
-Previous versions
-=================
-
-See :doc:`pytorch-training-history` to find documentation for previous releases
-of the ``ROCm/pytorch-training`` Docker image.

docs/how-to/rocm-for-ai/training/benchmark-docker/previous-versions/pytorch-training-v25.8.rst

@@ -1,588 +0,0 @@
-: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 on ROCm
-**************************************
-
-.. caution::
-
-   This documentation does not reflect the latest version of ROCm PyTorch training
-   performance benchmark documentation. See :doc:`../pytorch-training` for the latest version.
-
-PyTorch is an open-source machine learning framework that is widely used for
-model training with GPU-optimized components for transformer-based models.
-
-.. datatemplate:yaml:: /data/how-to/rocm-for-ai/training/previous-versions/pytorch-training-v25.8-benchmark-models.yaml
-
-   {% set dockers = data.dockers %}
-   {% set docker = dockers[0] %}
-   The `PyTorch for ROCm training Docker <{{ docker.docker_hub_url }}>`__
-   (``{{ docker.pull_tag }}``) image provides a prebuilt optimized environment for fine-tuning and pretraining a
-   model on AMD Instinct MI325X and MI300X GPUs. It includes the following software components to accelerate
-   training workloads:
-
-   .. list-table::
-      :header-rows: 1
-
-      * - Software component
-        - Version
-
-      {% for component_name, component_version in docker.components.items() %}
-      * - {{ component_name }}
-        - {{ component_version }}
-      {% endfor %}
-
-.. _amd-pytorch-training-model-support:
-
-Supported models
-================
-
-The following models are pre-optimized for performance on the AMD Instinct MI325X and MI300X GPUs.
-Some instructions, commands, and training recommendations in this documentation might
-vary by model -- select one to get started.
-
-.. datatemplate:yaml:: /data/how-to/rocm-for-ai/training/previous-versions/pytorch-training-v25.8-benchmark-models.yaml
-
-   {% set unified_docker = data.dockers[0] %}
-   {% set model_groups = data.model_groups %}
-   .. raw:: html
-
-      <div id="vllm-benchmark-ud-params-picker" class="container-fluid">
-         <div class="row gx-0">
-            <div class="col-2 me-1 px-2 model-param-head">Model</div>
-            <div class="row col-10 pe-0">
-      {% for model_group in model_groups %}
-               <div class="col-4 px-2 model-param" data-param-k="model-group" data-param-v="{{ model_group.tag }}" tabindex="0">{{ model_group.group }}</div>
-      {% endfor %}
-            </div>
-         </div>
-
-         <div class="row gx-0 pt-1">
-            <div class="col-2 me-1 px-2 model-param-head">Variant</div>
-            <div class="row col-10 pe-0">
-      {% for model_group in model_groups %}
-         {% set models = model_group.models %}
-         {% for model in models %}
-            {% if models|length % 3 == 0 %}
-               <div class="col-4 px-2 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 px-2 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>
-
-
-   .. _amd-pytorch-training-supported-training-modes:
-
-   The following table lists supported training modes per model.
-
-   .. dropdown:: Supported training modes
-
-      .. list-table::
-         :header-rows: 1
-
-         * - Model
-           - Supported training modes
-
-      {% for model_group in model_groups %}
-         {% set models = model_group.models %}
-         {% for model in models %}
-         {% if model.training_modes %}
-         * - {{ model.model }}
-           - ``{{ model.training_modes | join('``, ``') }}``
-
-         {% endif %}
-         {% endfor %}
-      {% endfor %}
-
-      .. note::
-
-         Some model and fine-tuning combinations are not listed. This is
-         because the `upstream torchtune repository <https://github.com/pytorch/torchtune>`__
-         doesn't provide default YAML configurations for them.
-         For advanced usage, you can create a custom configuration to enable
-         unlisted fine-tuning methods by using an existing file in the
-         ``/workspace/torchtune/recipes/configs`` directory as a template.
-
-.. _amd-pytorch-training-performance-measurements:
-
-Performance measurements
-========================
-
-To evaluate performance, the
-`Performance results with AMD ROCm software <https://www.amd.com/en/developer/resources/rocm-hub/dev-ai/performance-results.html#tabs-a8deaeb413-item-21cea50186-tab>`_
-page provides reference throughput and latency measurements for training
-popular AI models.
-
-.. note::
-
-   The performance data presented in
-   `Performance results with AMD ROCm software <https://www.amd.com/en/developer/resources/rocm-hub/dev-ai/performance-results.html#tabs-a8deaeb413-item-21cea50186-tab>`_
-   should not be interpreted as the peak performance achievable by AMD
-   Instinct MI325X and MI300X GPUs or ROCm software.
-
-System validation
-=================
-
-Before running AI workloads, it's important to validate that your AMD hardware is configured
-correctly and performing optimally.
-
-If you have already validated your system settings, including aspects like NUMA auto-balancing, you
-can skip this step. Otherwise, complete the procedures in the :ref:`System validation and
-optimization <rocm-for-ai-system-optimization>` guide to properly configure your system settings
-before starting training.
-
-To test for optimal performance, consult the recommended :ref:`System health benchmarks
-<rocm-for-ai-system-health-bench>`. This suite of tests will help you verify and fine-tune your
-system's configuration.
-
-This Docker image is optimized for specific model configurations outlined
-below. Performance can vary for other training workloads, as AMD
-doesn’t test configurations and run conditions outside those described.
-
-Run training
-============
-
-.. datatemplate:yaml:: /data/how-to/rocm-for-ai/training/previous-versions/pytorch-training-v25.8-benchmark-models.yaml
-
-   {% set unified_docker = data.dockers[0] %}
-   {% set model_groups = data.model_groups %}
-
-   Once the setup is complete, choose between two options to start benchmarking training:
-
-   .. tab-set::
-
-      .. tab-item:: MAD-integrated benchmarking
-
-   {% for model_group in model_groups %}
-      {% for model in model_group.models %}
-
-         .. container:: model-doc {{ model.mad_tag }}
-
-            The following run command is tailored to {{ model.model }}.
-            See :ref:`amd-pytorch-training-model-support` to switch to another available model.
-
-            1. 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.
-
-               .. code-block:: shell
-
-                  git clone https://github.com/ROCm/MAD
-                  cd MAD
-                  pip install -r requirements.txt
-
-            2. For example, use this command to run the performance benchmark test on the {{ model.model }} model
-               using one node with the {{ model.precision }} data type on the host machine.
-
-               .. code-block:: shell
-
-                  export MAD_SECRETS_HFTOKEN="your personal Hugging Face token to access gated models"
-                  madengine run \
-                      --tags {{ model.mad_tag }} \
-                      --keep-model-dir \
-                      --live-output \
-                      --timeout 28800
-
-               MAD launches a Docker container with the name
-               ``container_ci-{{ model.mad_tag }}``. The latency and throughput reports of the
-               model are collected in ``~/MAD/perf.csv``.
-
-      {% endfor %}
-   {% endfor %}
-
-      .. tab-item:: Standalone benchmarking
-
-   {% for model_group in model_groups %}
-      {% for model in model_group.models %}
-
-         .. container:: model-doc {{ model.mad_tag }}
-
-            The following commands are tailored to {{ model.model }}.
-            See :ref:`amd-pytorch-training-model-support` to switch to another available model.
-
-      {% endfor %}
-   {% endfor %}
-
-         .. rubric:: 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 {{ unified_docker.pull_tag }}
-
-         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 $HOME/.ssh:/root/.ssh \
-                   --shm-size 64G \
-                   --name training_env \
-                   {{ unified_docker.pull_tag }}
-
-            Use these commands if you exit the ``training_env`` container and need to return to it.
-
-            .. code-block:: shell
-
-               docker start training_env
-               docker exec -it training_env bash
-
-         3. In the Docker container, clone the `<https://github.com/ROCm/MAD>`__
-            repository and navigate to the benchmark scripts directory
-            ``/workspace/MAD/scripts/pytorch_train``.
-
-            .. code-block:: shell
-
-               git clone https://github.com/ROCm/MAD
-               cd MAD/scripts/pytorch_train
-
-         .. rubric:: Prepare training datasets and dependencies
-
-         1. The following benchmarking examples require downloading models and datasets
-            from Hugging Face. To ensure successful access to gated repos, set your
-            ``HF_TOKEN``.
-
-            .. code-block:: shell
-
-               export HF_TOKEN=$your_personal_hugging_face_access_token
-
-         2. Run the setup script to install libraries and datasets needed for benchmarking.
-
-            .. code-block:: shell
-
-               ./pytorch_benchmark_setup.sh
-
-            .. container:: model-doc pyt_train_llama-3.1-8b
-
-               ``pytorch_benchmark_setup.sh`` installs the following libraries for Llama 3.1 8B:
-
-               .. list-table::
-                  :header-rows: 1
-
-                  * - Library
-                    - Reference
-
-                  * - ``accelerate``
-                    - `Hugging Face Accelerate <https://huggingface.co/docs/accelerate/en/index>`_
-
-                  * - ``datasets``
-                    - `Hugging Face Datasets <https://huggingface.co/docs/datasets/v3.2.0/en/index>`_ 3.2.0
-
-            .. container:: model-doc pyt_train_llama-3.1-70b
-
-               ``pytorch_benchmark_setup.sh`` installs the following libraries for Llama 3.1 70B:
-
-               .. list-table::
-                  :header-rows: 1
-
-                  * - Library
-                    - Reference
-
-                  * - ``datasets``
-                    - `Hugging Face Datasets <https://huggingface.co/docs/datasets/v3.2.0/en/index>`_ 3.2.0
-
-                  * - ``torchdata``
-                    - `TorchData <https://meta-pytorch.org/data/beta/index.html#torchdata>`__
-
-                  * - ``tomli``
-                    - `Tomli <https://pypi.org/project/tomli/>`__
-
-                  * - ``tiktoken``
-                    - `tiktoken <https://github.com/openai/tiktoken>`__
-
-                  * - ``blobfile``
-                    - `blobfile <https://pypi.org/project/blobfile/>`__
-
-                  * - ``tabulate``
-                    - `tabulate <https://pypi.org/project/tabulate/>`__
-
-                  * - ``wandb``
-                    - `Weights & Biases <https://github.com/wandb/wandb>`__
-
-                  * - ``sentencepiece``
-                    - `SentencePiece <https://github.com/google/sentencepiece>`__ 0.2.0
-
-                  * - ``tensorboard``
-                    - `TensorBoard <https://www.tensorflow.org/tensorboard>`__ 2.18.0
-
-            .. container:: model-doc pyt_train_flux
-
-               ``pytorch_benchmark_setup.sh`` installs the following libraries for FLUX:
-
-               .. list-table::
-                  :header-rows: 1
-
-                  * - Library
-                    - Reference
-
-                  * - ``accelerate``
-                    - `Hugging Face Accelerate <https://huggingface.co/docs/accelerate/en/index>`_
-
-                  * - ``datasets``
-                    - `Hugging Face Datasets <https://huggingface.co/docs/datasets/v3.2.0/en/index>`__ 3.2.0
-
-                  * - ``sentencepiece``
-                    - `SentencePiece <https://github.com/google/sentencepiece>`__ 0.2.0
-
-                  * - ``tensorboard``
-                    - `TensorBoard <https://www.tensorflow.org/tensorboard>`__ 2.18.0
-
-                  * - ``csvkit``
-                    - `csvkit <https://csvkit.readthedocs.io/en/latest/>`__ 2.0.1
-
-                  * - ``deepspeed``
-                    - `DeepSpeed <https://github.com/deepspeedai/DeepSpeed>`__ 0.16.2
-
-                  * - ``diffusers``
-                    - `Hugging Face Diffusers <https://huggingface.co/docs/diffusers/en/index>`__ 0.31.0
-
-                  * - ``GitPython``
-                    - `GitPython <https://github.com/gitpython-developers/GitPython>`__ 3.1.44
-
-                  * - ``opencv-python-headless``
-                    - `opencv-python-headless <https://pypi.org/project/opencv-python-headless/>`__ 4.10.0.84
-
-                  * - ``peft``
-                    - `PEFT <https://huggingface.co/docs/peft/en/index>`__ 0.14.0
-
-                  * - ``protobuf``
-                    - `Protocol Buffers <https://github.com/protocolbuffers/protobuf>`__ 5.29.2
-
-                  * - ``pytest``
-                    - `PyTest <https://docs.pytest.org/en/stable/>`__ 8.3.4
-
-                  * - ``python-dotenv``
-                    - `python-dotenv <https://pypi.org/project/python-dotenv/>`__ 1.0.1
-
-                  * - ``seaborn``
-                    - `Seaborn <https://seaborn.pydata.org/>`__ 0.13.2
-
-                  * - ``transformers``
-                    - `Transformers <https://huggingface.co/docs/transformers/en/index>`__ 4.47.0
-
-            ``pytorch_benchmark_setup.sh`` downloads the following datasets from Hugging Face:
-
-            * `bghira/pseudo-camera-10k <https://huggingface.co/datasets/bghira/pseudo-camera-10k>`__
-
-   {% for model_group in model_groups %}
-      {% for model in model_group.models %}
-         {% set training_modes = model.training_modes %}
-         {% set training_mode_descs = {
-            "pretrain": "Benchmark pre-training.",
-            "HF_pretrain": "Llama 3.1 8B pre-training with FP8 precision."
-         } %}
-         {% set available_modes = training_modes | select("in", ["pretrain", "HF_pretrain"]) | list %}
-         {% if available_modes %}
-
-         .. container:: model-doc {{ model.mad_tag }}
-
-            .. rubric:: Pre-training
-
-            To start the pre-training 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 {% if available_modes | length == 1 %}{{ available_modes[0] }}{% else %}$training_mode{% endif %} \
-                   -m {{ model.model_repo }} \
-                   -p $datatype \
-                   -s $sequence_length
-
-            {% if model.mad_tag == "pyt_train_flux" %}
-            .. container:: model-doc {{ model.mad_tag }}
-
-               .. note::
-
-                  Currently, FLUX models are not supported out-of-the-box on {{ unified_docker.pull_tag }}.
-                  To use FLUX, refer to ``rocm/pytorch-training`` Docker: :doc:`pytorch-training-v25.6`
-
-                  Occasionally, downloading the Flux dataset might fail. In the event of this
-                  error, manually download it from Hugging Face at
-                  `black-forest-labs/FLUX.1-dev <https://huggingface.co/black-forest-labs/FLUX.1-dev>`_
-                  and save it to `/workspace/FluxBenchmark`. This ensures that the test script can access
-                  the required dataset.
-            {% endif %}
-
-            .. list-table::
-               :header-rows: 1
-
-               * - Name
-                 - Options
-                 - Description
-
-               {% for mode in available_modes %}
-               * - {% if loop.first %}``$training_mode``{% endif %}
-                 - ``{{ mode }}``
-                 - {{ training_mode_descs[mode] }}
-               {% endfor %}
-
-               * - ``$datatype``
-                 - ``BF16``{% if model.mad_tag == "pyt_train_llama-3.1-8b" %} or ``FP8``{% endif %}
-                 - Only Llama 3.1 8B supports FP8 precision.
-
-               * - ``$sequence_length``
-                 - Sequence length for the language model.
-                 - Between 2048 and 8192. 8192 by default.
-         {% endif %}
-
-         {% set training_mode_descs = {
-            "finetune_fw": "Full weight fine-tuning (BF16 and FP8 supported).",
-            "finetune_lora": "LoRA fine-tuning (BF16 supported).",
-            "finetune_qlora": "QLoRA fine-tuning (BF16 supported).",
-            "HF_finetune_lora": "LoRA fine-tuning with Hugging Face PEFT.",
-         } %}
-         {% set available_modes = training_modes | select("in", ["finetune_fw", "finetune_lora", "finetune_qlora", "HF_finetune_lora"]) | list %}
-         {% if available_modes %}
-         .. container:: model-doc {{ model.mad_tag }}
-
-            .. rubric:: Fine-tuning
-
-            To start the fine-tuning benchmark, use the following command with the
-            appropriate options. See the following list of options and their descriptions.
-            See :ref:`supported training modes <amd-pytorch-training-supported-training-modes>`.
-
-            .. code-block:: shell
-
-               ./pytorch_benchmark_report.sh -t $training_mode \
-                   -m {{ model.model_repo }} \
-                   -p $datatype \
-                   -s $sequence_length
-
-            .. list-table::
-               :header-rows: 1
-
-               * - Name
-                 - Options
-                 - Description
-
-               {% for mode in available_modes %}
-               * - {% if loop.first %}``$training_mode``{% endif %}
-                 - ``{{ mode }}``
-                 - {{ training_mode_descs[mode] }}
-               {% endfor %}
-
-               * - ``$datatype``
-                 - ``BF16``{% if "finetune_fw" in available_modes %} or ``FP8``{% endif %}
-                 - All models support BF16.{% if "finetune_fw" in available_modes %} FP8 is only available for full weight fine-tuning.{% endif %}
-
-               * - ``$sequence_length``
-                 - Between 2048 and 16384.
-                 - Sequence length for the language model.
-
-            {% if model.mad_tag in ["pyt_train_llama3.2-vision-11b", "pyt_train_llama-3.2-vision-90b"] %}
-            .. note::
-
-               For LoRA and QLoRA support with vision models (Llama 3.2 11B and 90B),
-               use the following torchtune commit for compatibility:
-
-               .. code-block:: shell
-
-                  git checkout 48192e23188b1fc524dd6d127725ceb2348e7f0e
-
-            {% elif model.mad_tag in ["pyt_train_llama-2-7b", "pyt_train_llama-2-13b", "pyt_train_llama-2-70b"] %}
-            .. note::
-
-               You might encounter the following error with Llama 2: ``ValueError: seq_len (16384) of
-               input tensor should be smaller than max_seq_len (4096)``.
-               This error indicates that an input sequence is longer than the model's maximum context window.
-
-               Ensure your tokenized input does not exceed the model's ``max_seq_len`` (4096
-               tokens in this case). You can resolve this by truncating the input or splitting
-               it into smaller chunks before passing it to the model.
-
-               Note on reproducibility: The results in this guide are based on
-               commit ``b4c98ac`` from the upstream
-               `<https://github.com/pytorch/torchtune>`__ repository. For the
-               latest updates, you can use the main branch.
-
-            {% endif %}
-         {% endif %}
-      {% endfor %}
-   {% endfor %}
-
-            .. rubric:: Benchmarking examples
-
-            For examples of benchmarking commands, see `<https://github.com/ROCm/MAD/tree/develop/benchmark/pytorch_train#benchmarking-examples>`__.
-
-.. _amd-pytorch-training-multinode-examples:
-
-Multi-node training
--------------------
-
-Refer to :doc:`/how-to/rocm-for-ai/system-setup/multi-node-setup` to configure your environment for multi-node
-training. See :ref:`rocm-for-ai-multi-node-setup-pyt-train-example` for example Slurm run commands.
-
-Pre-training
-~~~~~~~~~~~~
-
-Multi-node training with torchtitan is supported. The provided SLURM script is pre-configured for Llama 3 70B.
-
-To launch the training job on a SLURM cluster for Llama 3 70B, run the following commands from the MAD repository.
-
-.. code-block:: shell
-
-   # In the MAD repository
-   cd scripts/pytorch_train
-   sbatch run_slurm_train.sh
-
-Fine-tuning
-~~~~~~~~~~~
-
-Multi-node training with torchtune is supported. The provided SLURM script is pre-configured for Llama 3.3 70B.
-
-To launch the training job on a SLURM cluster for Llama 3.3 70B, run the following commands from the MAD repository.
-
-.. code-block:: shell
-
-   huggingface-cli login # Get access to HF Llama model space
-   huggingface-cli download meta-llama/Llama-3.3-70B-Instruct --local-dir ./models/Llama-3.3-70B-Instruct # Download the Llama 3.3 model locally
-   # In the MAD repository
-   cd scripts/pytorch_train
-   sbatch Torchtune_Multinode.sh
-
-.. note::
-
-   Information regarding benchmark setup:
-
-   * By default, Llama 3.3 70B is fine-tuned using ``alpaca_dataset``.
-   * You can adjust the torchtune `YAML configuration file
-     <https://github.com/pytorch/torchtune/blob/main/recipes/configs/llama3_3/70B_full_multinode.yaml>`__
-     if you're using a different model.
-   * The number of nodes and other parameters can be tuned in the SLURM script ``Torchtune_Multinode.sh``.
-   * Set the ``mounting_paths`` inside the SLURM script.
-
-Once the run is finished, you can find the log files in the ``result_torchtune/`` directory.
-
-Further reading
-===============
-
-- To learn more about MAD and the ``madengine`` CLI, see the `MAD usage guide <https://github.com/ROCm/MAD?tab=readme-ov-file#usage-guide>`__.
-
-- To learn more about system settings and management practices to configure your system for
-  AMD Instinct MI300X series GPUs, see `AMD Instinct MI300X system optimization <https://instinct.docs.amd.com/projects/amdgpu-docs/en/latest/system-optimization/mi300x.html>`_.
-
-- For a list of other ready-made Docker images for AI with ROCm, see
-  `AMD Infinity Hub <https://www.amd.com/en/developer/resources/infinity-hub.html#f-amd_hub_category=AI%20%26%20ML%20Models>`_.
-
-Previous versions
-=================
-
-See :doc:`pytorch-training-history` to find documentation for previous releases
-of the ``ROCm/pytorch-training`` Docker image.

docs/index.rst

@@ -2,16 +2,16 @@
    :description: Learn what ROCm is – AMD's open software stack for GPU programming, including runtimes, compilers, libraries, and tools for Linux and Windows.
    :keywords: ROCm, AMD, GPU computing, ROCm Core SDK, ROCm components, TheRock, ROCm architecture, HPC, AI, machine learning, ROCm runtime
 
-**********************
-AMD ROCm 7.9.0 preview
-**********************
+*******************************
+AMD ROCm |ROCM_VERSION| preview
+*******************************
 
-AMD ROCm is an open, modular, and high‑performance GPU software ecosystem —
-built collaboratively with the community, maintained transparently, and
-optimized for consistent, scalable performance across data centers, desktops,
-and edge devices.
+AMD ROCm is an open, modular, and high‑performance GPU software ecosystem
+— built collaboratively with the community, maintained transparently, and
+optimized for consistent, scalable performance across data centers,
+workstations, and edge devices.
 
-ROCm 7.9.0 is a technology preview release built with
+ROCm |ROCM_VERSION| is a technology preview release built with
 `TheRock <https://github.com/ROCm/TheRock>`__, AMD’s new open build and release
 system.
 This preview introduces a new modular build workflow that will become standard
@@ -34,11 +34,12 @@ frameworks such as PyTorch.
 * **Open source** -- Transparent development driven by community feedback
 * **Cross‑platform** -- Supports Linux and Windows environments
 * **Comprehensive** -- End‑to‑end toolchain from compilers to libraries
-* **Performance‑focused** -- Tuned for AMD Instinct™, AMD Radeon™, and AMD Ryzen™ architectures
+* **Performance‑focused** -- Tuned for AMD Instinct™, AMD Radeon™, and AMD Ryzen™ devices
 
 .. image:: data/rocm-ontology.png
    :align: center
    :alt: ROCm software ecosystem and components illustration
+   :width: 1000
 
 ROCm supports AMD GPU architectures spanning data center, workstation, and APU
 categories. TheRock enables a unified ROCm user‑space experience across
@@ -59,8 +60,8 @@ What’s changing
 ROCm is evolving to improve flexibility, maintainability, and
 use‑case alignment.
 
-* **Leaner core** – The Core SDK focuses on essential runtime and development components.  
-* **Use case‑specific expansions** – Optional domain‑specific SDKs for AI, data science, and HPC.  
+* **Leaner core** – The Core SDK focuses on essential runtime and development components.
+* **Use case‑specific expansions** – Optional domain‑specific SDKs for AI, data science, and HPC.
 * **Modular installation** – Install only the components required for your workflow.
 
 This approach streamlines installation, reduces footprint, and accelerates
@@ -76,6 +77,7 @@ computing.
 .. image:: data/rocm-sdk-arch.png
    :align: center
    :alt: ROCm Core SDK internal architecture illustration
+   :width: 1000
 
 The TheRock infrastructure keeps these components modular, consistent, and easy
 to integrate across configurations.

docs/install/build-from-source.rst

@@ -0,0 +1,118 @@
+.. meta::
+   :description: Learn how to build the ROCm Core SDK from source using TheRock. Includes references to environment setup guides for Ubuntu 24.04 and Windows 11, plus links to official instructions and compatibility guidance.
+   :keywords: AMD ROCm Core SDK, build ROCm from source, TheRock, ROCm SDK, Ubuntu 24.04, Windows 11, ROCm environment setup, AMD GPU compute, ROCm compatibility, ROCm build guide
+
+***********************************
+Build the ROCm Core SDK from source
+***********************************
+
+You can build the ROCm Core SDK from source using the open-source unified build
+system `TheRock <https://github.com/ROCm/TheRock>`__. To learn more about the
+motivation and architecture behind this system, see `ROCm Technology Preview:
+ROCm Core SDK and TheRock Build System
+<https://rocm.blogs.amd.com/software-tools-optimization/therock/README.html>`__.
+
+This page consists mainly of key references to `TheRock's README
+<https://github.com/ROCm/TheRock?tab=readme-ov-file#building-from-source>`__
+and `supporting development manuals
+<https://github.com/ROCm/TheRock/blob/main/README.md#development-manuals>`__
+which provide up-to-date build instructions and guidance for supported
+platforms. See `TheRock Development Guide
+<https://github.com/ROCm/TheRock/blob/main/docs/development/development_guide.md#therock-development-guide>`__
+to learn about the overall build architecture.
+
+.. tip::
+
+   Building from source is recommended only if you need custom builds or are
+   contributing to ROCm development.
+   For most users, installing from official AMD releases is faster and easier.
+   See :doc:`/install/rocm` for installation instructions.
+
+Prerequisites
+=============
+
+A successful build depends on a correctly configured environment. Before you
+begin, ensure your system meets all hardware and software requirements. Review the
+following resources:
+
+* `Environment setup guide
+  <https://github.com/ROCm/TheRock/blob/main/docs/environment_setup_guide.md>`__
+
+* :doc:`/compatibility/compatibility-matrix`
+
+High-level build process
+========================
+
+For an overview of the build architecture, start with TheRock's `development guide
+<https://github.com/ROCm/TheRock/blob/main/docs/development/development_guide.md>`__.
+While specific commands vary by platform, the general workflow for building
+from source involves these stages:
+
+1. Clone the repository and install build dependencies. See :ref:`build-from-src-plat-setup`.
+
+2. Configure the build: Use CMake feature flags to configure the build. This
+   step allows you to target specific platforms, build subsets of ROCm Core SDK
+   components, and toggle component features. See `Build configuration
+   <https://github.com/ROCm/TheRock/blob/main/README.md#build-configuration>`__
+   for optional and required build flags.
+
+3. Execute the build: Run the build command to compile the source code. This
+   can be a time- and resource-intensive process. See `CMake build usage
+   <https://github.com/ROCm/TheRock/blob/main/README.md#cmake-build-usage>`__.
+
+4. After a successful build, the outputs are available for use in downstream
+   tasks. To learn more about build outputs, see the relevant `TheRock
+   documentation <https://github.com/ROCm/TheRock/blob/main/docs/development/artifacts.md>`__.
+   Common post-build tasks include:
+
+   * Using ``build/dist/rocm``: When the build completes, you should have a
+     build of ROCm in the ``build/dist/rocm/`` directory. See `Using installed
+     tarballs <https://github.com/ROCm/TheRock/blob/main/RELEASES.md#using-installed-tarballs>`__
+     for more information.
+
+   * Building Python packages: Prepare the build artifacts for distribution as
+     Python packages. See `Building Python packages
+     <https://github.com/ROCm/TheRock/blob/main/docs/packaging/python_packaging.md#building-packages>`__.
+
+   * Building PyTorch: Build a compatible PyTorch version against ROCm wheels.
+     See the `PyTorch build instructions
+     <https://github.com/ROCm/TheRock/tree/main/external-builds/pytorch#build-instructions>`__.
+
+.. _build-from-src-plat-setup:
+
+Platform-specific setup
+=======================
+
+ManyLinux
+---------
+
+On Linux, it's recommended to build with ManyLinux to produce binaries that are
+portable across Ubuntu and other Linux distributions. To learn more about what
+a ROCm ManyLinux build entails, see `ManyLinux builds
+<https://github.com/ROCm/TheRock/blob/main/docs/design/manylinux_builds.md>`__.
+Refer to `ManyLinux x86_84
+<https://github.com/ROCm/TheRock/blob/main/docs/environment_setup_guide.md#manylinux-x84-64>`__
+in the environment setup guide.
+
+Ubuntu 24.04
+------------
+
+TheRock provides detailed instructions and scripts for preparing an Ubuntu
+24.04 system, including installing necessary packages via apt.
+Refer to `Setup — Ubuntu 24.04 <https://github.com/ROCm/TheRock?tab=readme-ov-file#setup---ubuntu-2404>`__
+in the TheRock repository for guidance.
+
+Windows 11
+----------
+
+For setup instructions on Windows 11 using Visual Studio 2022, see
+`Setup — Windows 11
+<https://github.com/ROCm/TheRock?tab=readme-ov-file#setup---windows-11-vs-2022>`__
+in the TheRock repository.
+
+.. seealso::
+
+   For details on supported configurations, known issues, and other
+   Windows-specific considerations, review the `Windows support
+   <https://github.com/ROCm/TheRock/blob/main/docs/development/windows_support.md>`__
+   documentation.

docs/install/includes/1_rhel-10.0-tar-prerequisites.md

@@ -1,21 +0,0 @@
-1. Register your Enterprise Linux.
-
-   ```bash
-   subscription-manager register --username <username> --password <password>
-   ```
-
-2. Update your Enterprise Linux.
-
-   ```bash
-   sudo dnf update --releasever=10.0 --exclude=\*release\*
-   ```
-
-3. Configure permissions for GPU access.
-
-   ```bash
-   sudo usermod -a -G render,video $LOGNAME # Add the current user to the render and video groups
-   ```
-
-   ```{note}
-   To apply all settings, reboot your system.
-   ```

docs/install/includes/1_rhel-10.0-wheel-prerequisites.md

@@ -1,27 +0,0 @@
-1. Register your Enterprise Linux.
-
-   ```bash
-   subscription-manager register --username <username> --password <password>
-   ```
-
-2. Update your Enterprise Linux.
-
-   ```bash
-   sudo dnf update --releasever=10.0 --exclude=\*release\*
-   ```
-
-3. Install Python 3.12 or 3.13.
-
-   ```bash
-   sudo dnf install python3.12 python3.12-pip
-   ```
-
-4. Configure permissions for GPU access.
-
-   ```bash
-   sudo usermod -a -G render,video $LOGNAME # Add the current user to the render and video groups
-   ```
-
-   ```{note}
-   To apply all settings, reboot your system.
-   ```

docs/install/includes/1_rhel-9.6-tar-prerequisites.md

@@ -1,22 +0,0 @@
-1. Register your Enterprise Linux.
-
-   ```bash
-   subscription-manager register --username <username> --password <password>
-   sudo subscription-manager attach --auto
-   ```
-
-2. Update your Enterprise Linux.
-
-   ```bash
-   sudo dnf update --releasever=9.6 --exclude=\*release\*
-   ```
-
-3. Configure permissions for GPU access.
-
-   ```bash
-   sudo usermod -a -G render,video $LOGNAME # Add the current user to the render and video groups
-   ```
-
-   ```{note}
-   To apply all settings, reboot your system.
-   ```

docs/install/includes/1_rhel-9.6-wheel-prerequisites.md

@@ -1,28 +0,0 @@
-1. Register your Enterprise Linux.
-
-   ```bash
-   subscription-manager register --username <username> --password <password>
-   sudo subscription-manager attach --auto
-   ```
-
-2. Update your Enterprise Linux.
-
-   ```bash
-   sudo dnf update --releasever=9.6 --exclude=\*release\*
-   ```
-
-3. Install Python 3.11, 3.12 or 3.13.
-
-   ```bash
-   sudo dnf install python3.11 python3.11-pip
-   ```
-
-4. Configure permissions for GPU access.
-
-   ```bash
-   sudo usermod -a -G render,video $LOGNAME # Add the current user to the render and video groups
-   ```
-
-   ```{note}
-   To apply all settings, reboot your system.
-   ```

docs/install/includes/1_ubuntu-22.04-wheel-prerequisites.md

@@ -1,15 +0,0 @@
-1. Install Python 3.11.
-
-   ```bash
-   sudo apt install python3.11 python3.11-venv
-   ```
-
-2. Configure permissions for GPU access.
-
-   ```bash
-   sudo usermod -a -G render,video $LOGNAME # Add the current user to the render and video groups
-   ```
-
-   ```{note}
-   To apply all settings, reboot your system.
-   ```

docs/install/includes/1_ubuntu-24.04-wheel-prerequisites.md

@@ -1,15 +0,0 @@
-1. Install Python 3.12 or 3.13.
-
-   ```bash
-   sudo apt install python3.12 python3.12-venv
-   ```
-
-2. Configure permissions for GPU access.
-
-   ```bash
-   sudo usermod -a -G render,video $LOGNAME # Add the current user to the render and video groups
-   ```
-
-   ```{note}
-   To apply all settings, reboot your system.
-   ```

docs/install/includes/1_ubuntu-tar-prerequisites.md

@@ -1,9 +0,0 @@
-Configure permissions for GPU access.
-
-```bash
-sudo usermod -a -G render,video $LOGNAME # Add the current user to the render and video groups
-```
-
-```{note}
-To apply all settings, reboot your system.
-```

docs/install/includes/1_windows-tar-prerequisites.md

@@ -1,6 +0,0 @@
-1. Remove any existing HIP SDK installations and other
-   conflicting AMD graphics software.
-
-2. Install the [Adrenalin Driver version
-   25.9.2](https://www.amd.com/en/resources/support-articles/release-notes/RN-RAD-WIN-25-9-2.html).
-   For instructions, see [Install AMD Software: Adrenalin Edition](https://www.amd.com/en/resources/support-articles/faqs/RSX2-INSTALL.html).

docs/install/includes/1_windows-wheel-prerequisites.md

@@ -1,8 +0,0 @@
-1. Remove any existing HIP SDK installations and other
-   conflicting AMD graphics software.
-
-2. Install the [Adrenalin Driver version
-   25.9.2](https://www.amd.com/en/resources/support-articles/release-notes/RN-RAD-WIN-25-9-2.html).
-   For instructions, see [Install AMD Software: Adrenalin Edition](https://www.amd.com/en/resources/support-articles/faqs/RSX2-INSTALL.html).
-
-3. Install a supported Python version: 3.11, 3.12, or 3.13.

docs/install/includes/2_rhel-instinct-install-kmd.md

@@ -1,6 +0,0 @@
-For information about driver compatibility, see the {doc}`compatibility-matrix`.
-
-For information about the AMD GPU driver installation, see the
-[RHEL native
-installation](https://instinct.docs.amd.com/projects/amdgpu-docs/en/latest/install/detailed-install/package-manager/package-manager-rhel.html)
-in the AMD Instinct Data Center GPU Documentation.

docs/install/includes/2_ubuntu-instinct-install-kmd.md

@@ -1,6 +0,0 @@
-For information about driver compatibility, see the {doc}`compatibility-matrix`.
-
-For information about the AMD GPU driver installation, see the
-[Ubuntu native
-installation](https://instinct.docs.amd.com/projects/amdgpu-docs/en/latest/install/detailed-install/package-manager/package-manager-ubuntu.html)
-in the AMD Instinct Data Center GPU Documentation.

docs/install/includes/3_tar-gfx1151-install-rocm.md

@@ -1,20 +0,0 @@
-1. Create the installation directory. For example:
-
-   ```bash
-   mkdir therock-tarball && cd therock-tarball
-   ```
-
-   ```{note}
-   Subsequent commands assume you're working with the
-   `therock-tarball` directory.
-   If you choose a different directory name, adjust the
-   subsequent commands accordingly.
-   ```
-
-2. Download and unpack the tarball.
-
-   ```bash
-   wget https://repo.amd.com/rocm/tarball/therock-dist-linux-gfx1151-7.9.0rc1.tar.gz
-   mkdir install
-   tar -xf *.tar.gz -C install
-   ```

docs/install/includes/3_tar-gfx942-install-rocm.md

@@ -1,20 +0,0 @@
-1. Create the installation directory. For example:
-
-   ```bash
-   mkdir therock-tarball && cd therock-tarball
-   ```
-
-   ```{note}
-   Subsequent commands assume you're working with the
-   `therock-tarball` directory.
-   If you choose a different directory name, adjust the
-   subsequent commands accordingly.
-   ```
-
-2. Download and unpack the tarball.
-
-   ```bash
-   wget https://repo.amd.com/rocm/tarball/therock-dist-linux-gfx94X-dcgpu-7.9.0rc1.tar.gz
-   mkdir install
-   tar -xf *.tar.gz -C install
-   ```

docs/install/includes/3_tar-gfx950-install-rocm.md

@@ -1,20 +0,0 @@
-1. Create the installation directory. For example:
-
-   ```bash
-   mkdir therock-tarball && cd therock-tarball
-   ```
-
-   ```{note}
-   Subsequent commands assume you're working with the
-   `therock-tarball` directory.
-   If you choose a different directory name, adjust the
-   subsequent commands accordingly.
-   ```
-
-2. Download and unpack the tarball.
-
-   ```bash
-   wget https://repo.amd.com/rocm/tarball/therock-dist-linux-gfx950-dcgpu-7.9.0rc1.tar.gz
-   mkdir install
-   tar -xf *.tar.gz -C install
-   ```

docs/install/includes/3_wheel-gfx1151-install-rocm.md

@@ -1,12 +0,0 @@
-1. Set up your Python virtual environment.
-
-   ```bash
-   python3.12 -m venv .venv
-   source .venv/bin/activate
-   ```
-
-2. Install ROCm wheels packages.
-
-   ```bash
-   python -m pip install --index-url https://repo.amd.com/rocm/whl/gfx1151/ "rocm[libraries,devel]"
-   ```

docs/install/includes/3_wheel-gfx942-install-rocm-py311.md

@@ -1,12 +0,0 @@
-1. Set up your Python virtual environment.
-
-   ```bash
-   python3.11 -m venv .venv
-   source .venv/bin/activate
-   ```
-
-2. Install ROCm wheels packages.
-
-   ```bash
-   python -m pip install --index-url https://repo.amd.com/rocm/whl/gfx94X-dcgpu/ "rocm[libraries,devel]"
-   ```

docs/install/includes/3_wheel-gfx942-install-rocm.md

@@ -1,12 +0,0 @@
-1. Set up your Python virtual environment.
-
-   ```bash
-   python3.12 -m venv .venv
-   source .venv/bin/activate
-   ```
-
-2. Install ROCm wheels packages.
-
-   ```bash
-   python -m pip install --index-url https://repo.amd.com/rocm/whl/gfx94X-dcgpu/ "rocm[libraries,devel]"
-   ```

docs/install/includes/3_wheel-gfx950-install-rocm-py311.md

@@ -1,12 +0,0 @@
-1. Set up your Python virtual environment.
-
-   ```bash
-   python3.11 -m venv .venv
-   source .venv/bin/activate
-   ```
-
-2. Install ROCm wheels packages.
-
-   ```bash
-   python -m pip install --index-url https://repo.amd.com/rocm/whl/gfx950-dcgpu/ "rocm[libraries,devel]"
-   ```

docs/install/includes/3_wheel-gfx950-install-rocm.md

@@ -1,12 +0,0 @@
-1. Set up your Python virtual environment.
-
-   ```bash
-   python3.12 -m venv .venv
-   source .venv/bin/activate
-   ```
-
-2. Install ROCm wheels packages.
-
-   ```bash
-   python -m pip install --index-url https://repo.amd.com/rocm/whl/gfx950-dcgpu/ "rocm[libraries,devel]"
-   ```

docs/install/includes/3_windows-tar-install-rocm.md

@@ -1,47 +0,0 @@
-```{important}
-- Do not copy/replace the ROCm-SDK compiler and runtime DLLs to
-`System32` as this can cause conflicts.
-
-- Disable the following Windows security features as they
-  can interfere with ROCm functionality:
-
-  - Turn off WDAG (Windows Defender Application Guard)
-    - Control Panel > Programs > Programs and Features > Turn Windows features on or off > **Deselect** "Microsoft Defender Application Guard"
-  - Turn off SAC (Smart App Control)
-    - Settings > Privacy & security > Windows Security > App & browser control > Smart App Control settings > **Off**
-```
-
-1. Create the installation directory in `C:\TheRock\build`.
-
-   ```{note}
-   Subsequent commands assume you're working with the
-   `C:\TheRock\build` directory.
-   ```
-
-2. Download the tarball and extract the contents to
-   `C:\TheRock\build`.
-
-   - Download link: [https://repo.amd.com/rocm/tarball/therock-dist-windows-gfx1151-7.9.0rc1.tar.gz](https://repo.amd.com/rocm/tarball/therock-dist-windows-gfx1151-7.9.0rc1.tar.gz)
-
-3. Set the following environment variables using the command
-   prompt as an administrator:
-
-   ```bat
-   setx HIP_DEVICE_LIB_PATH “C:\TheRock\build\lib\llvm\amdgcn\bitcode” /M
-   setx HIP_PATH “C:\TheRock\build” /M
-   setx HIP_PLATFORM “amd” /M
-   setx LLVM_PATH “C:\TheRock\build\lib\llvm” /M
-   ```
-
-4. Add the following paths into PATH environment variable using your system settings GUI.
-
-   - `C:\TheRock\build\bin`
-
-   - `C:\TheRock\build\lib\llvm\bin`
-
-5. Open a new command prompt window for the environment variables to take effect. Run `set`
-   to see the list of active variables.
-
-   ```bat
-   set
-   ````

docs/install/includes/3_windows-wheel-install-rocm.md

@@ -1,12 +0,0 @@
-1. Set up your Python virtual environment.
-
-   ```bash
-   python3.12 -m venv .venv
-   .venv\Scripts\activate
-   ```
-
-2. Install ROCm wheels packages.
-
-   ```bash
-   python -m pip install --index-url https://repo.amd.com/rocm/whl/gfx1151/ "rocm[libraries,devel]"
-   ```

docs/install/includes/4_linux-tar-post-install.md

@@ -1,53 +0,0 @@
-1. Configure ROCm PATH. Make sure you're in the `therock-tarball` directory before proceeding.
-
-   ```bash
-   export ROCM_PATH=$PWD
-   export PATH=$PATH:$ROCM_PATH/install/bin
-   ```
-
-2. Configure `LD_LIBRARY_PATH`.
-
-   ```bash
-   export LD_LIBRARY_PATH=$ROCM_PATH/install/lib
-   ```
-
-3. Verify the ROCm installation.
-
-   ```bash
-   rocminfo
-   amd-smi
-   ```
-
-   ```{eval-rst}
-   .. dropdown:: Example output of ``rocminfo``
-
-      .. code-block:: shell-session
-
-         ROCk module is loaded
-         =====================    
-         HSA System Attributes    
-         =====================    
-         Runtime Version:         1.18
-         Runtime Ext Version:     1.14
-         System Timestamp Freq.:  1000.000000MHz
-         Sig. Max Wait Duration:  18446744073709551615 (0xFFFFFFFFFFFFFFFF) (timestamp count)
-         Machine Model:           LARGE                              
-         System Endianness:       LITTLE                             
-         Mwaitx:                  DISABLED
-         XNACK enabled:           NO
-         DMAbuf Support:          YES
-         VMM Support:             YES
-
-         ==========               
-         HSA Agents               
-         ==========               
-         *******                  
-         Agent 1                  
-         *******                  
-           Name:                    AMD RYZEN AI MAX+ PRO 395 w/ Radeon 8060S
-           Uuid:                    CPU-XX                             
-           Marketing Name:          AMD RYZEN AI MAX+ PRO 395 w/ Radeon 8060S
-           Vendor Name:             CPU
-
-         [output truncated]
-   ```

docs/install/includes/4_rhel-wheel-post-install.md

@@ -1,72 +0,0 @@
-1. Verify the ROCm installation.
-
-   ```bash
-   rocminfo
-   amd-smi
-   ```
-
-   ```{eval-rst}
-
-   .. dropdown:: Example output of ``rocminfo``
-
-      .. code-block:: shell-session
-
-         ROCk module is loaded
-         =====================    
-         HSA System Attributes    
-         =====================    
-         Runtime Version:         1.18
-         Runtime Ext Version:     1.14
-         System Timestamp Freq.:  1000.000000MHz
-         Sig. Max Wait Duration:  18446744073709551615 (0xFFFFFFFFFFFFFFFF) (timestamp count)
-         Machine Model:           LARGE                              
-         System Endianness:       LITTLE                             
-         Mwaitx:                  DISABLED
-         XNACK enabled:           NO
-         DMAbuf Support:          YES
-         VMM Support:             YES
-     
-         ==========               
-         HSA Agents               
-         ==========               
-         *******                  
-         Agent 1                  
-         *******                  
-           Name:                    AMD RYZEN AI MAX+ PRO 395 w/ Radeon 8060S
-           Uuid:                    CPU-XX                             
-           Marketing Name:          AMD RYZEN AI MAX+ PRO 395 w/ Radeon 8060S
-           Vendor Name:             CPU
-
-         [output truncated]
-   ```
-
-2. Inspect your ROCm installation in your Python environment.
-
-   ```bash
-   pip freeze | grep rocm
-   which rocm-sdk
-   ls .venv/bin
-   ```
-
-3. Test your ROCm installation.
-
-   ```bash
-   rocm-sdk targets
-   rocm-sdk path --cmake
-   rocm-sdk path --bin
-   rocm-sdk path --root
-   rocm-sdk test
-   ```
-
-   To learn more about the `rocm-sdk` tool and to see example expected outputs,
-   see [Using ROCm Python packages
-   (TheRock)](https://github.com/ROCm/TheRock/blob/main/RELEASES.md#using-rocm-python-packages).
-
-````{tip}
-If you need to deactivate your Python virtual environment when finished,
-run:
-
-```shell
-deactivate
-```
-````

docs/install/includes/4_ubuntu-wheel-post-install.md

@@ -1,72 +0,0 @@
-1. Verify the ROCm installation.
-
-   ```bash
-   rocminfo
-   amd-smi
-   ```
-
-   ```{eval-rst}
-
-   .. dropdown:: Example output of ``rocminfo``
-
-      .. code-block:: shell-session
-
-         ROCk module is loaded
-         =====================    
-         HSA System Attributes    
-         =====================    
-         Runtime Version:         1.18
-         Runtime Ext Version:     1.14
-         System Timestamp Freq.:  1000.000000MHz
-         Sig. Max Wait Duration:  18446744073709551615 (0xFFFFFFFFFFFFFFFF) (timestamp count)
-         Machine Model:           LARGE                              
-         System Endianness:       LITTLE                             
-         Mwaitx:                  DISABLED
-         XNACK enabled:           NO
-         DMAbuf Support:          YES
-         VMM Support:             YES
-     
-         ==========               
-         HSA Agents               
-         ==========               
-         *******                  
-         Agent 1                  
-         *******                  
-           Name:                    AMD RYZEN AI MAX+ PRO 395 w/ Radeon 8060S
-           Uuid:                    CPU-XX                             
-           Marketing Name:          AMD RYZEN AI MAX+ PRO 395 w/ Radeon 8060S
-           Vendor Name:             CPU
-
-         [output truncated]
-   ```
-
-2. Inspect your ROCm installation in your Python environment.
-
-   ```bash
-   pip freeze | grep rocm
-   which rocm-sdk
-   ls .venv/bin
-   ```
-
-3. Test your ROCm installation.
-
-   ```bash
-   rocm-sdk targets
-   rocm-sdk path --cmake
-   rocm-sdk path --bin
-   rocm-sdk path --root
-   rocm-sdk test
-   ```
-
-   To learn more about the `rocm-sdk` tool and to see example expected outputs,
-   see [Using ROCm Python packages
-   (TheRock)](https://github.com/ROCm/TheRock/blob/main/RELEASES.md#using-rocm-python-packages).
-
-````{tip}
-If you need to deactivate your Python virtual environment when finished,
-run:
-
-```shell
-deactivate
-```
-````

docs/install/includes/4_windows-tar-post-install.md

@@ -1,21 +0,0 @@
-Verify the ROCm installation.
-
-```shell
-hipinfo
-```
-
-```{eval-rst}
-.. dropdown:: Example output of ``hipinfo``
-
-   .. code-block:: shell-session
-
-      --------------------------------------------------------------------------------
-      device#                           0
-      Name:                             AMD Radeon(TM) 8060S Graphics
-      pciBusID:                         197
-      pciDeviceID:                      0
-      pciDomainID:                      0
-      multiProcessorCount:              20
-
-      [output truncated]
-```

docs/install/includes/4_windows-wheel-post-install.md

@@ -1,48 +0,0 @@
-1. Verify the ROCm installation.
-
-   ```bash
-   hipinfo
-   ```
-
-   ```{eval-rst}
-   .. dropdown:: Example output of ``hipinfo``
-
-      .. code-block:: shell-session
-
-         --------------------------------------------------------------------------------
-         device#                           0
-         Name:                             AMD Radeon(TM) 8060S Graphics
-         pciBusID:                         197
-         pciDeviceID:                      0
-         pciDomainID:                      0
-         multiProcessorCount:              20
-
-         [output truncated]
-   ```
-
-2. Inspect your ROCm installation in your Python environment.
-
-   ```bash
-   pip freeze
-   where rocm-sdk
-   dir .venv\Scripts
-   ```
-
-3. Test your ROCm installation.
-
-   ```bash
-   rocm-sdk test
-   ```
-
-   To learn more about the `rocm-sdk` tool and to see example expected outputs,
-   see [Using ROCm Python packages
-   (TheRock)](https://github.com/ROCm/TheRock/blob/main/RELEASES.md#using-rocm-python-packages).
-
-````{tip}
-If you need to deactivate your Python virtual environment when finished,
-run:
-
-```shell
-deactivate
-```
-````

docs/install/includes/5_tar-uninstall.md

@@ -1,5 +0,0 @@
-To uninstall ROCm, remove your installation directory.
-
-```bash
-sudo rm -rf therock-tarball
-```

docs/install/includes/5_wheel-linux-uninstall.md

@@ -1,11 +0,0 @@
-1. Clear the pip cache.
-
-   ```bash
-   sudo rm -rf ~/.cache/pip
-   ```
-
-2. Remove your local Python virtual environment.
-
-   ```bash
-   sudo rm -rf .venv
-   ```

docs/install/includes/5_wheel-windows-uninstall.md

@@ -1,11 +0,0 @@
-1. Clear the pip cache.
-
-   ```bash
-   pip cache purge
-   ```
-
-2. Remove your local Python virtual environment.
-
-   ```bash
-   rmdir /s /q .venv
-   ```

docs/install/includes/5_windows-tar-uninstall.md

@@ -1,18 +0,0 @@
-1. Delete the `C:\TheRock\build` and its contents.
-
-2. Delete the environment variables. For example, using PowerShell as an administrator:
-
-   ```powershell
-   [Environment]::SetEnvironmentVariable("HIP_PATH", $null, "Machine")
-   [Environment]::SetEnvironmentVariable("HIP_DEVICE_LIB_PATH", $null, "Machine")
-   [Environment]::SetEnvironmentVariable("HIP_PLATFORM", $null, "Machine")
-   [Environment]::SetEnvironmentVariable("LLVM_PATH", $null, "Machine")
-   ```
-
-3. Remove the following paths from your PATH environment variable using your system settings GUI.
-
-   - `C:\TheRock\build\bin`
-
-   - `C:\TheRock\build\lib\llvm\bin`
-
-4. If you want to uninstall the Adrenalin driver, see [Uninstall AMD Software](https://www.amd.com/en/resources/support-articles/faqs/RSX2-UNINSTALL.html).

docs/install/includes/6_rhel-troubleshooting.md

@@ -1,8 +0,0 @@
-Docker images often only include a minimal set of installations, meaning some
-essential packages might be missing. When installing ROCm within a Docker
-container, you might need to install additional packages for a successful
-installation. Use the following commands to install the prerequisite packages.
-
-```bash
-dnf install sudo libatomic
-```

docs/install/includes/6_ubuntu-troubleshooting.md

@@ -1,9 +0,0 @@
-Docker images often only include a minimal set of installations, meaning some
-essential packages might be missing. When installing ROCm within a Docker
-container, you might need to install additional packages for a successful
-installation. Use the following commands to install the prerequisite packages.
-
-```bash
-apt update
-apt install sudo wget
-```

docs/install/includes/install.rst

@@ -0,0 +1,395 @@
+Installation
+============
+
+.. dropdown:: Installation environment
+   :animate: fade-in-slide-down
+   :icon: desktop-download
+   :chevron: down-up
+
+   .. include:: ./includes/selector.rst
+
+Before getting started, make sure you've completed the :ref:`rocm-prerequisites`.
+For information about supported operating systems and compatible AMD devices,
+see the :doc:`Compatibility matrix </compatibility/compatibility-matrix>`.
+
+.. selected:: os=windows
+
+   .. caution::
+
+      Do not copy/replace the ROCm compiler and runtime DLLs to System32 as
+      this can cause conflicts.
+
+.. selected:: fam=instinct fam=radeon-pro fam=radeon
+
+   .. selected:: os=ubuntu
+      :heading: Install kernel driver
+      :heading-level: 3
+
+      For information about AMD GPU Driver (amdgpu) compatibility, see the
+      :doc:`Compatibility matrix </compatibility/compatibility-matrix>`.
+
+      For instructions on installing the AMD GPU Driver (amdgpu), see `Ubuntu native
+      installation
+      <https://instinct.docs.amd.com/projects/amdgpu-docs/en/latest/install/detailed-install/package-manager/package-manager-ubuntu.html>`__
+      in the AMD Instinct Data Center GPU Documentation.
+
+   .. selected:: os=rhel
+      :heading: Install kernel driver
+      :heading-level: 3
+
+      For information about AMD GPU Driver (amdgpu) compatibility, see the
+      :doc:`Compatibility matrix </compatibility/compatibility-matrix>`.
+
+      For instructions on installing the AMD GPU Driver (amdgpu), see `RHEL native
+      installation
+      <https://instinct.docs.amd.com/projects/amdgpu-docs/en/latest/install/detailed-install/package-manager/package-manager-rhel.html>`__
+      in the AMD Instinct Data Center GPU Documentation.
+
+   .. selected:: os=sles
+      :heading: Install kernel driver
+      :heading-level: 3
+
+      For information about AMD GPU Driver (amdgpu) compatibility, see the
+      :doc:`Compatibility matrix </compatibility/compatibility-matrix>`.
+
+      For instructions on installing the AMD GPU Driver (amdgpu), see `SLES native
+      installation
+      <https://instinct.docs.amd.com/projects/amdgpu-docs/en/latest/install/detailed-install/package-manager/package-manager-sles.html>`__
+      in the AMD Instinct Data Center GPU Documentation.
+
+.. selected:: fam=ryzen
+
+   .. selected:: os=ubuntu os=rhel os=sles
+      :heading: About the kernel driver
+      :heading-level: 3
+
+      Supported Ryzen AI APUs require the inbox kernel driver included with
+      Ubuntu 24.04.3.
+
+Install ROCm
+------------
+
+Use the following instructions to install the ROCm Core SDK on your system.
+
+.. selected:: i=pip
+   :heading: Set up your Python virtual environment
+   :heading-level: 4
+
+   Create and activate the Python virtual environment where you'll install
+   ROCm packages.
+
+   .. selected:: os=ubuntu
+
+      .. selected:: os-version=24
+
+         For example, to create and activate a Python 3.12 virtual environment,
+         run the following command:
+
+         .. code-block:: bash
+
+            python3.12 -m venv .venv
+            source .venv/bin/activate
+
+      .. selected:: os-version=22
+
+         For example, to create and activate a Python 3.11 virtual environment,
+         run the following command:
+
+         .. code-block:: bash
+
+            python3.11 -m venv .venv
+            source .venv/bin/activate
+
+   .. selected:: os=rhel
+
+      .. selected:: os-version=10.1 os-version=10.0
+
+         For example, to create and activate a Python 3.12 virtual environment,
+         run the following command:
+
+         .. code-block:: bash
+
+            python3.12 -m venv .venv
+            source .venv/bin/activate
+
+      .. selected:: os-version=9.7 os-version=9.6
+
+         For example, to create and activate a Python 3.11 virtual environment,
+         run the following command:
+
+         .. code-block:: bash
+
+            python3.11 -m venv .venv
+            source .venv/bin/activate
+
+      .. selected:: os-version=8
+
+         For example, to create and activate a Python 3.11 virtual environment,
+         run the following command:
+
+         .. code-block:: bash
+
+            python3.11 -m venv .venv
+            source .venv/bin/activate
+
+   .. selected:: os=sles
+
+      For example, to create and activate a Python 3.11 virtual environment,
+      run the following command:
+
+      .. code-block:: bash
+
+         python3.11 -m venv .venv
+         source .venv/bin/activate
+
+   .. selected:: os=windows
+
+      For example, to create and activate a Python 3.12 virtual environment,
+      run the following command:
+
+      .. code-block:: bat
+
+         python3.12 -m venv .venv
+         .venv\Scripts\activate
+
+.. selected:: i=pip
+   :heading: Install ROCm wheel packages
+   :heading-level: 4
+
+   .. selected:: gfx=950
+
+      Use pip to install the ROCm Core SDK libraries and development tools for
+      your ``gfx950`` GPU.
+
+      Run the following command:
+
+      .. code-block:: bash
+
+         python -m pip install --index-url https://repo.amd.com/rocm/whl/gfx950-dcgpu/ "rocm[libraries,devel]"
+
+   .. selected:: gfx=942
+
+      Use pip to install the ROCm Core SDK libraries and development tools for
+      your ``gfx942`` device.
+
+      Run the following command:
+
+      .. code-block:: bash
+
+         python -m pip install --index-url https://repo.amd.com/rocm/whl/gfx94X-dcgpu/ "rocm[libraries,devel]"
+
+   .. selected:: gfx=90a
+
+      Use pip to install the ROCm Core SDK libraries and development tools for
+      your ``gfx90a`` GPU.
+
+      Run the following command:
+
+      .. code-block:: bash
+
+         python -m pip install --index-url https://repo.amd.com/rocm/whl/gfx90X-dcgpu/ "rocm[libraries,devel]"
+
+   .. selected:: gfx=1100
+
+      Use pip to install the ROCm Core SDK libraries and development tools for
+      your ``gfx1100`` GPU.
+
+      Run the following command:
+
+      .. code-block:: bash
+
+         python -m pip install --index-url https://repo.amd.com/rocm/whl/gfx110X-dgpu/ "rocm[libraries,devel]"
+
+   .. selected:: gfx=1101
+
+      Use pip to install the ROCm Core SDK libraries and development tools for
+      your ``gfx1101`` GPU.
+
+      Run the following command:
+
+      .. code-block:: bash
+
+         python -m pip install --index-url https://repo.amd.com/rocm/whl/gfx110X-dgpu/ "rocm[libraries,devel]"
+
+   .. selected:: gfx=1150
+
+      Use pip to install the ROCm Core SDK libraries and development tools for
+      your ``gfx1150`` GPU.
+
+      Run the following command:
+
+      .. code-block:: bash
+
+         python -m pip install --index-url https://repo.amd.com/rocm/whl/gfx1150/ "rocm[libraries,devel]"
+
+   .. selected:: gfx=1151
+
+      Use pip to install the ROCm Core SDK libraries and development tools for
+      your ``gfx1151`` GPU.
+
+      Run the following command:
+
+      .. code-block:: bash
+
+         python -m pip install --index-url https://repo.amd.com/rocm/whl/gfx1151/ "rocm[libraries,devel]"
+
+.. selected:: i=tar
+   :heading: Create the installation directory
+   :heading-level: 4
+
+   .. selected:: os=ubuntu os=rhel os=sles
+
+      Run the following command in your desired location to create your
+      installation directory:
+
+      .. code-block:: bash
+
+         mkdir therock-tarball && cd therock-tarball
+
+      .. important::
+
+         Subsequent commands assume you're working with the ``therock-tarball``
+         directory. If you choose a different directory name, adjust the
+         commands accordingly.
+
+
+   .. selected:: os=windows
+
+      Create the installation directory in ``C:\TheRock\build``.
+
+      .. important::
+
+         Subsequent commands assume you're working with the
+         ``C:\TheRock\build`` directory. If you choose a different directory
+         name, adjust the commands accordingly.
+
+.. selected:: i=tar
+   :heading: Download and unpack the tarball
+   :heading-level: 4
+
+   .. selected:: os=ubuntu os=rhel os=sles
+
+      .. selected:: gfx=950
+
+         Use the following commands to download and untar the ROCm tarball for
+         your ``gfx950`` GPU.
+
+         Run the following command:
+
+         .. code-block:: bash
+
+            wget https://repo.amd.com/rocm/tarball/therock-dist-linux-gfx950-dcgpu-7.10.0.tar.gz
+            mkdir install
+            tar -xf *.tar.gz -C install
+
+      .. selected:: gfx=942
+
+         Use the following commands to download and untar the ROCm tarball for
+         your ``gfx942`` GPU.
+
+         Run the following command:
+
+         .. code-block:: bash
+
+            wget https://repo.amd.com/rocm/tarball/therock-dist-linux-gfx94X-dcgpu-7.10.0.tar.gz
+            mkdir install
+            tar -xf *.tar.gz -C install
+
+      .. selected:: gfx=90a
+
+         Use the following commands to download and untar the ROCm tarball for
+         your ``gfx90a`` GPU.
+
+         Run the following command:
+
+         .. code-block:: bash
+
+            wget https://repo.amd.com/rocm/tarball/therock-dist-linux-gfx90X-dcgpu-7.10.0.tar.gz
+            mkdir install
+            tar -xf *.tar.gz -C install
+
+      .. selected:: gfx=1100
+
+         Use the following commands to download and untar the ROCm tarball for
+         your ``gfx1100`` GPU.
+
+         Run the following command:
+
+         .. code-block:: bash
+
+            wget https://repo.amd.com/rocm/tarball/therock-dist-linux-gfx110X-dgpu-7.10.0.tar.gz
+            mkdir install
+            tar -xf *.tar.gz -C install
+
+      .. selected:: gfx=1101
+
+         Use the following commands to download and untar the ROCm tarball for
+         your ``gfx1101`` GPU.
+
+         Run the following command:
+
+         .. code-block:: bash
+
+            wget https://repo.amd.com/rocm/tarball/therock-dist-linux-gfx110X-dgpu-7.10.0.tar.gz
+            mkdir install
+            tar -xf *.tar.gz -C install
+
+      .. selected:: gfx=1150
+
+         Use the following commands to download and untar the ROCm tarball for
+         your ``gfx1150`` GPU.
+
+         Run the following command:
+
+         .. code-block:: bash
+
+            wget https://repo.amd.com/rocm/tarball/therock-dist-linux-gfx1150-7.10.0.tar.gz
+            mkdir install
+            tar -xf *.tar.gz -C install
+
+      .. selected:: gfx=1151
+
+         Use the following commands to download and untar the ROCm tarball for
+         your ``gfx1151`` GPU.
+
+         Run the following command:
+
+         .. code-block:: bash
+
+            wget https://repo.amd.com/rocm/tarball/therock-dist-linux-gfx1151-7.10.0.tar.gz
+            mkdir install
+            tar -xf *.tar.gz -C install
+
+   .. selected:: os=windows
+
+      .. selected:: gfx=1100
+
+         Download the tarball for your ``gfx1100`` GPU and extract the contents
+         to ``C:\TheRock\build``.
+
+         - Download link: `therock-dist-windows-gfx110X-dgpu-7.10.0.tar.gz
+           <https://repo.amd.com/rocm/tarball/therock-dist-windows-gfx110X-dgpu-7.10.0.tar.gz>`__
+
+      .. selected:: gfx=1101
+
+         Download the tarball for your ``gfx1101`` GPU and extract the contents
+         to ``C:\TheRock\build``.
+
+         - Download link: `therock-dist-windows-gfx110X-dgpu-7.10.0.tar.gz
+           <https://repo.amd.com/rocm/tarball/therock-dist-windows-gfx110X-dgpu-7.10.0.tar.gz>`__
+
+      .. selected:: gfx=1150
+
+         Download the tarball for your ``gfx1150`` GPU and extract the contents
+         to ``C:\TheRock\build``.
+
+         - Download link: `therock-dist-windows-gfx1150-7.10.0.tar.gz
+           <https://repo.amd.com/rocm/tarball/therock-dist-windows-gfx1150-7.10.0.tar.gz>`__
+
+      .. selected:: gfx=1151
+
+         Download the tarball for your ``gfx1151`` GPU and extract the contents
+         to ``C:\TheRock\build``.
+
+         - Download link: `therock-dist-windows-gfx1151-7.10.0.tar.gz
+           <https://repo.amd.com/rocm/tarball/therock-dist-windows-gfx1151-7.10.0.tar.gz>`__

docs/install/includes/post-install.rst

@@ -0,0 +1,316 @@
+Post-installation
+=================
+
+.. dropdown:: Installation environment
+   :animate: fade-in-slide-down
+   :icon: desktop-download
+   :chevron: down-up
+
+   .. include:: ./includes/selector.rst
+
+After installing the ROCm Core SDK |ROCM_VERSION| -- see :ref:`rocm-install` --
+complete these post-installation steps to complete your system configuration
+and validate the installation.
+
+.. selected:: i=tar
+
+   .. selected:: os=ubuntu os=rhel os=sles
+      :heading: Configure your environment
+      :heading-level: 3
+
+      Configure environment variables so that ROCm libraries and tools are
+      available either to all users on the system or only to your user account.
+
+      .. _rocm-post-install-system-wide:
+
+      .. tab-set::
+
+         .. tab-item:: System-wide setup
+
+            Create a profile script so that all users inherit the ROCm
+            environment variables when they start a shell session. Make sure
+            you're in the ``therock-tarball`` directory before proceeding.
+
+            .. code-block:: bash
+
+               # Configure ROCm PATH. Make sure you're in the therock-tarball directory before proceeding.
+               ROCM_INSTALL_PATH=$(pwd)/install
+               sudo tee /etc/profile.d/set-rocm-env.sh << EOF
+               export ROCM_PATH=$ROCM_INSTALL_PATH
+               export PATH=\$PATH:\$ROCM_PATH/bin
+               export LD_LIBRARY_PATH=\$ROCM_PATH/lib
+               EOF
+               sudo chmod +x /etc/profile.d/set-rocm-env.sh
+               source /etc/profile.d/set-rocm-env.sh
+
+         .. tab-item:: User setup
+
+            Configure the ROCm environment for your user by updating your shell
+            configuration file.
+
+            1. Add the following to your shell configuration file
+               (``~/.bashrc``, ``~/.profile``). Make sure you're in the
+               ``therock-tarball`` directory before proceeding.
+
+               .. code-block:: bash
+
+                  # Configure ROCm PATH. Make sure you're in the therock-tarball directory before proceeding.
+                  export ROCM_PATH=$PWD/install
+                  export PATH=$PATH:$ROCM_PATH/bin
+                  export LD_LIBRARY_PATH=$ROCM_PATH/lib
+
+            2. After modifying your shell configuration, apply the change to
+               your current session by sourcing your updated shell
+               configuration file.
+
+               .. tab-set::
+
+                  .. tab-item:: .bashrc
+
+                     .. code-block:: bash
+
+                        source ~/.bashrc
+
+                  .. tab-item:: .profile
+
+                     .. code-block:: bash
+
+                        source ~/.profile
+
+   .. selected:: os=windows
+      :heading: Configure your environment
+      :heading-level: 3
+
+      Configure environment variables so that ROCm libraries and tools are
+      available on your Windows system.
+
+      1. Set the following environment variables using the command
+         prompt as an administrator:
+
+         .. code-block:: bat
+
+            setx HIP_DEVICE_LIB_PATH “C:\TheRock\build\lib\llvm\amdgcn\bitcode” /M
+            setx HIP_PATH “C:\TheRock\build” /M
+            setx HIP_PLATFORM “amd” /M
+            setx LLVM_PATH “C:\TheRock\build\lib\llvm” /M
+
+      2. Add the following paths into PATH environment variable using your system settings GUI.
+
+         - ``C:\TheRock\build\bin``
+
+         - ``C:\TheRock\build\lib\llvm\bin``
+
+      3. Open a new command prompt window for the environment variables to take effect. Run ``set``
+         to see the list of active variables.
+
+         .. code-block:: bat
+
+            set
+
+.. selected:: os=ubuntu os=rhel os=sles
+   :heading: Verify your installation
+   :heading-level: 3
+
+   Use the following ROCm tools to verify that the ROCm stack is correctly
+   installed and that your AMD GPU is visible to the system.
+
+   1. Use ``rocminfo`` to list detected AMD GPUs and confirm that the ROCm
+      runtimes and drivers are correctly installed and loaded.
+
+      .. code-block:: bash
+
+         rocminfo
+
+      .. dropdown:: Example output of ``rocminfo``
+         :animate: fade-in-slide-down
+         :color: success
+         :icon: note
+         :chevron: down-up
+
+         .. code-block:: shell-session
+
+            ROCk module is loaded
+            =====================
+            HSA System Attributes
+            =====================
+            Runtime Version:         1.18
+            Runtime Ext Version:     1.14
+            System Timestamp Freq.:  1000.000000MHz
+            Sig. Max Wait Duration:  18446744073709551615 (0xFFFFFFFFFFFFFFFF) (timestamp count)
+            Machine Model:           LARGE
+            System Endianness:       LITTLE
+            Mwaitx:                  DISABLED
+            XNACK enabled:           NO
+            DMAbuf Support:          YES
+            VMM Support:             YES
+
+            ==========
+            HSA Agents
+            ==========
+            *******
+            Agent 1
+            *******
+              Name:                    AMD RYZEN AI MAX+ PRO 395 w/ Radeon 8060S
+              Uuid:                    CPU-XX
+              Marketing Name:          AMD RYZEN AI MAX+ PRO 395 w/ Radeon 8060S
+              Vendor Name:             CPU
+
+            ... [output truncated]
+
+   2. Use the AMD SMI CLI ``amd-smi`` to validate system information.
+
+      .. code-block:: bash
+
+         amd-smi version
+
+      .. dropdown:: Example output of ``amd-smi version``
+         :animate: fade-in-slide-down
+         :color: success
+         :icon: note
+         :chevron: down-up
+
+         .. code-block:: shell-session
+
+            AMDSMI Tool: 26.1.0+cd50d9e0 | AMDSMI Library version: 26.1.0 | ROCm version: 7.10.0 | amdgpu version: 6.16.6 | amd_hsmp version: N/A
+
+   .. selected:: i=pip
+
+      3. Inspect your installation in your Python environment and confirm that
+         ROCm packages, including the ``rocm-sdk`` CLI, are available.
+
+         .. code-block:: bash
+
+            pip freeze | grep rocm
+            which rocm-sdk
+            ls .venv/bin
+
+.. selected:: os=windows
+   :heading: Verify your installation
+   :heading-level: 3
+
+   Use the following ROCm tools to verify that the ROCm stack is correctly
+   installed and that your AMD GPU is visible to the system.
+
+   .. selected:: i=pip
+
+      1. Use ``hipinfo`` to list detected AMD GPUs and confirm that the ROCm
+         runtimes and drivers are correctly installed and loaded.
+
+         .. code-block:: bash
+
+            hipinfo
+
+         .. dropdown:: Example output of ``hipinfo``
+            :animate: fade-in-slide-down
+            :color: success
+            :icon: note
+            :chevron: down-up
+
+            .. code-block:: shell-session
+
+               --------------------------------------------------------------------------------
+               device#                           0
+               Name:                             AMD Radeon(TM) 8060S Graphics
+               pciBusID:                         197
+               pciDeviceID:                      0
+               pciDomainID:                      0
+               multiProcessorCount:              20
+
+               ... [output truncated]
+
+      2. Inspect your installation in your Python environment and confirm that
+         ROCm packages, including the ``rocm-sdk`` CLI, are available.
+
+         .. code-block:: bash
+
+            pip freeze
+            where rocm-sdk
+            dir .venv\Scripts
+
+   .. selected:: i=tar
+
+      Use ``hipinfo`` to list detected AMD GPUs and confirm that the ROCm
+      runtimes and drivers are correctly installed and loaded.
+
+      .. code-block:: bash
+
+         hipinfo
+
+      .. dropdown:: Example output of ``hipinfo``
+         :animate: fade-in-slide-down
+         :color: success
+         :icon: note
+         :chevron: down-up
+
+         .. code-block:: shell-session
+
+            --------------------------------------------------------------------------------
+            device#                           0
+            Name:                             AMD Radeon(TM) 8060S Graphics
+            pciBusID:                         197
+            pciDeviceID:                      0
+            pciDomainID:                      0
+            multiProcessorCount:              20
+
+            ... [output truncated]
+
+.. selected:: os=ubuntu os=rhel os=sles
+
+   .. selected:: i=pip
+      :heading: Test your installation
+      :heading-level: 3
+
+      Run the following commands from your Python virtual environment to confirm
+      that the ROCm SDK is correctly configured and that basic checks complete
+      successfully.
+
+      .. code-block:: bash
+
+         rocm-sdk targets
+         rocm-sdk path --cmake
+         rocm-sdk path --bin
+         rocm-sdk path --root
+
+      To learn more about the ``rocm-sdk`` tool and to see example expected
+      outputs, see `Using ROCm Python packages (TheRock)
+      <https://github.com/ROCm/TheRock/blob/main/RELEASES.md#using-rocm-python-packages>`__.
+
+   .. selected:: i=tar
+      :heading: Test your installation
+      :heading-level: 3
+
+      Run the ``test_hip_api`` tool to verify that the HIP runtime can access
+      your GPU and execute a simple workload.
+
+      .. code-block:: bash
+
+         test_hip_api
+
+.. selected:: os=windows
+
+   .. selected:: i=pip
+      :heading: Test your installation
+      :heading-level: 3
+
+      Run the following commands from your Python virtual environment to confirm
+      that the ROCm SDK is correctly configured and that basic checks complete
+      successfully.
+
+      .. code-block:: bash
+
+         rocm-sdk test
+
+      To learn more about the ``rocm-sdk`` tool and to see example expected
+      outputs, see `Using ROCm Python packages (TheRock)
+      <https://github.com/ROCm/TheRock/blob/main/RELEASES.md#using-rocm-python-packages>`__.
+
+.. selected:: i=pip
+
+   .. tip::
+
+      If you need to deactivate your Python virtual environment when finished, run:
+
+      .. code-block::
+
+         deactivate
+

docs/install/includes/prerequisites.rst

@@ -0,0 +1,301 @@
+Prerequisites
+=============
+
+Before installing the ROCm Core SDK |ROCM_VERSION|, ensure your system meets
+all prerequisites. This includes installing the required dependencies and
+configuring permissions for GPU access. To confirm that your system is
+supported, see the :doc:`Compatibility matrix
+</compatibility/compatibility-matrix>`.
+
+.. selected:: os=ubuntu os=rhel os=sles
+
+   .. dropdown:: Install essential packages for Docker containers
+      :animate: fade-in-slide-down
+      :color: info
+      :icon: tools
+      :chevron: down-up
+
+      Docker images often include only a minimal set of installations, so some
+      essential packages might be missing. When installing ROCm within a Docker
+      container, you might need to install additional packages for a successful
+      installation.
+
+      If applicable, run the following command to install essential packages:
+
+      .. selected:: os=ubuntu
+
+         .. code-block:: bash
+
+            apt update
+            apt install sudo wget python3 libatomic1
+
+      .. selected:: os=rhel
+
+         .. selected:: os-version=10.1 os-version=10.0 os-version=9.7 os-version=9.6
+
+            .. code-block:: bash
+
+               dnf install sudo wget libatomic
+
+         .. selected:: os-version=8
+
+            .. code-block:: bash
+
+               dnf install sudo wget libatomic python3
+
+      .. selected:: os=sles
+
+         .. code-block:: bash
+
+            zypper install sudo libatomic1 libgfortran5 wget SUSEConnect python3
+
+
+.. selected:: os=windows
+
+    1. Remove any existing HIP SDK for Windows installations and other
+       conflicting AMD graphics software.
+
+    2. Install the Adrenalin Driver for Windows.
+
+       * For general use cases, use the Adrenalin Driver version 25.11.1. For
+         details and the download link, see `AMD Software: Adrenalin
+         Edition 25.11.1
+         <https://www.amd.com/en/resources/support-articles/release-notes/RN-RAD-WIN-25-11-1.html>`__.
+
+       * If you intend to run :ref:`ComfyUI workloads
+         <install-comfyui-windows>`, use driver version 25.20.01.17. For
+         details and the download link, see `AMD Software: PyTorch on Windows
+         Edition 7.1.1
+         <https://www.amd.com/en/resources/support-articles/release-notes/RN-AMDGPU-WINDOWS-PYTORCH-7-1-1.html>`__.
+
+    3. Disable the following Windows security features as they can interfere
+       with ROCm functionality:
+
+       * Turn off WDAG (Windows Defender Application Guard)
+
+         * Control Panel > Programs > Programs and Features > Turn Windows
+           features on or off > **Clear** “Microsoft Defender Application
+           Guard”
+
+       * Turn off SAC (Smart App Control)
+
+         * Settings > Privacy & security > Windows Security > App & browser
+           control > Smart App Control settings > **Off**
+
+.. selected:: os=rhel
+   :heading: Register your Red Hat Enterprise Linux system
+   :heading-level: 3
+
+   Register your Red Hat Enterprise Linux (RHEL) system to enable access to Red
+   Hat repositories and ensure you’re able to download and install packages.
+
+   Run the following command to register your system:
+
+   .. selected:: os-version=10.1 os-version=10.0
+
+      .. code-block:: bash
+
+         subscription-manager register --username <username> --password <password>
+
+   .. selected:: os-version=9.7 os-version=9.6 os-version=8
+
+      .. code-block:: bash
+
+         subscription-manager register --username <username> --password <password>
+         subscription-manager attach --auto
+
+.. selected:: os=sles
+   :heading: Register your SUSE Linux Enterprise Server system
+   :heading-level: 3
+
+   Register your SUSE Linux Enterprise Server (SLES) system to enable access to
+   SUSE repositories and ensure you’re able to download and install packages.
+
+   Run the following command to register your system:
+
+   .. code-block:: bash
+
+      sudo SUSEConnect -r <REGCODE>
+
+.. selected:: os=rhel
+   :heading: Update your system
+   :heading-level: 3
+
+   After registering your system, update RHEL to the latest packages. This is
+   particularly important for newer hardware on older versions of RHEL.
+
+   Run the following command to update your system:
+
+   .. selected:: os-version=10.1
+
+      .. code-block:: bash
+
+         sudo dnf update --releasever=10.1 --exclude=\*release\*
+
+   .. selected:: os-version=10.0
+
+      .. code-block:: bash
+
+         sudo dnf update --releasever=10.0 --exclude=\*release\*
+
+   .. selected:: os-version=9.7
+
+      .. code-block:: bash
+
+         sudo dnf update --releasever=9.7 --exclude=\*release\*
+
+   .. selected:: os-version=9.6
+
+      .. code-block:: bash
+
+         sudo dnf update --releasever=9.6 --exclude=\*release\*
+
+   .. selected:: os-version=8
+
+      .. code-block:: bash
+
+         sudo dnf update --releasever=8.10 --exclude=\*release\*
+
+.. selected:: os=sles
+   :heading: Update your system
+   :heading-level: 3
+
+   After registering your system, update SLES to the latest available packages.
+   This is particularly important for newer hardware on older versions of SLES.
+
+   Run the following command to update your system:
+
+   .. code-block:: bash
+
+      sudo zypper update
+
+.. selected:: i=pip
+
+   .. selected:: os=ubuntu
+
+      .. selected:: os-version=24
+         :heading: Install Python
+         :heading-level: 3
+
+         Install a supported Python version. For example, to install Python
+         3.12, run the following command:
+
+         .. code-block:: bash
+
+            sudo apt install python3.12 python3.12-venv
+
+      .. selected:: os-version=22
+         :heading: Install Python
+         :heading-level: 3
+
+         Install a supported Python version. For example, to install Python
+         3.11, run the following command:
+
+         .. code-block:: bash
+
+            sudo apt install python3.11 python3.11-venv
+
+   .. selected:: os=rhel
+
+      .. selected:: os-version=10.1 os-version=10.0
+         :heading: Install Python
+         :heading-level: 3
+
+         Install a supported Python version. For example, to install Python
+         3.12, run the following command:
+
+         .. code-block:: bash
+
+            sudo dnf install python3.12 python3.12-pip
+
+      .. selected:: os-version=9.7 os-version=9.6 os-version=8
+         :heading: Install Python
+         :heading-level: 3
+
+         Install a supported Python version. For example, to install Python
+         3.11, run the following command:
+
+         .. code-block:: bash
+
+            sudo dnf install python3.11 python3.11-pip
+
+   .. selected:: os=sles
+      :heading: Install Python
+      :heading-level: 3
+
+      Install a supported Python version. For example, to install Python 3.11,
+      run the following command:
+
+      .. code-block:: bash
+
+         sudo zypper install -y python311 python311-pip
+
+   .. selected:: os=windows
+      :heading: Install Python
+      :heading-level: 3
+
+      Install a supported Python version: 3.11, 3.12, or 3.13.
+
+.. selected:: os=rhel
+
+   .. selected:: os-version=10.0 os-version=8
+      :heading: Install additional development packages
+      :heading-level: 3
+
+      .. code-block:: bash
+
+         sudo dnf install libatomic
+
+.. selected:: os=sles
+
+   .. selected:: os-version=15
+      :heading: Install additional development packages
+      :heading-level: 3
+
+      .. code-block:: bash
+
+         sudo zypper install libatomic1
+
+.. selected:: os=ubuntu os=rhel os=sles
+   :heading: Configure permissions for GPU access
+   :heading-level: 3
+
+   There are two primary methods of configuring GPU access for ROCm: group
+   membership or udev rules. Each method has its own advantages. The choice
+   depends on your specific requirements and system management preferences.
+
+   .. tab-set::
+
+      .. tab-item:: Group membership
+
+         By default, GPU access is controlled by membership in the ``video`` and
+         ``render`` Linux system groups. The ``video`` group traditionally handles
+         video device access, while the ``render`` group manages GPU rendering
+         through DRM render nodes.
+
+         .. code-block:: bash
+
+            # Add the current user to the render and video groups
+            sudo usermod -a -G render,video $LOGNAME
+
+      .. tab-item:: udev rules
+
+         udev rules are a flexible, system-wide approach for managing device
+         permissions, eliminating the need for user group management while
+         allowing granular GPU access. To enable them and grant GPU access to
+         all users, run the following command:
+
+         .. code-block:: bash
+
+            sudo tee /etc/udev/rules.d/70-amdgpu.rules << EOF
+            KERNEL=="kfd", GROUP="render", MODE="0666"
+            SUBSYSTEM=="drm", KERNEL=="renderD*", GROUP="render", MODE="0666"
+            EOF
+
+            sudo udevadm control --reload-rules
+            sudo udevadm trigger
+
+   .. note::
+
+      To apply all settings, reboot your system.
+

docs/install/includes/selector.rst

@@ -0,0 +1,73 @@
+.. include:: /compatibility/includes/selector.rst
+
+.. selected:: fam=instinct fam=radeon-pro fam=radeon
+
+   .. selector:: Ubuntu version
+      :key: os-version
+      :show-when: os=ubuntu
+
+      .. selector-option:: 24.04.3
+         :value: 24
+
+      .. selector-option:: 22.04.5
+         :value: 22
+
+.. selected:: fam=ryzen
+
+   .. selector:: Ubuntu version
+      :key: os-version
+      :show-when: os=ubuntu
+
+      .. selector-option:: 24.04.3
+         :value: 24
+         :width: 12
+
+.. selector:: RHEL version
+   :key: os-version
+   :show-when: os=rhel
+
+   .. selector-option:: 10.1
+      :value: 10.1
+      :width: 3
+
+   .. selector-option:: 10.0
+      :value: 10.0
+      :width: 3
+
+   .. selector-option:: 9.7
+      :value: 9.7
+      :width: 2
+
+   .. selector-option:: 9.6
+      :value: 9.6
+      :width: 2
+
+   .. selector-option:: 8.10
+      :value: 8
+      :width: 2
+
+.. selector:: SLES version
+   :key: os-version
+   :show-when: os=sles
+
+   .. selector-option:: 15.7
+      :value: 15
+      :width: 12
+
+.. selector:: Windows version
+   :key: os-version
+   :show-when: os=windows
+
+   .. selector-option:: 11 25H2
+      :value: 11-25h2
+      :width: 12
+
+.. selector:: Installation method
+   :key: i
+
+   .. selector-option:: pip
+      :value: pip
+
+   .. selector-option:: Tarball
+      :value: tar
+