Docs: Ray release 25.12 and compatibility version format standardization (#5845) (#5846)

Cherrypicking amd-smi updates from ROCm internal

.azuredevops/components/HIP.yml

@@ -34,7 +34,6 @@ parameters:
   default:
     - cmake
     - libnuma-dev
-    - libsimde-dev
     - mesa-common-dev
     - ninja-build
     - ocl-icd-libopencl1

.azuredevops/components/origami.yml

@@ -39,7 +39,6 @@ parameters:
     - python3
     - python3-dev
     - python3-pip
-    - python3-venv
     - libgtest-dev
     - libboost-filesystem-dev
     - libboost-program-options-dev
@@ -47,8 +46,6 @@ parameters:
   type: object
   default:
     - nanobind>=2.0.0
-    - pytest
-    - pytest-cov
 - name: rocmDependencies
   type: object
   default:
@@ -75,10 +72,8 @@ parameters:
       - { os: ubuntu2204, packageManager: apt }
       - { os: almalinux8, packageManager: dnf }
     testJobs:
+      - { os: ubuntu2204, packageManager: apt, target: gfx942 }
       - { os: ubuntu2204, packageManager: apt, target: gfx90a }
-      # - { os: ubuntu2204, packageManager: apt, target: gfx1100 }
-      # - { os: ubuntu2204, packageManager: apt, target: gfx1151 }
-      # - { os: ubuntu2204, packageManager: apt, target: gfx1201 }
 - name: downstreamComponentMatrix
   type: object
   default:
@@ -121,11 +116,6 @@ jobs:
       parameters:
         dependencyList:
           - gtest
-    - ${{ if ne(job.os, 'almalinux8') }}:
-      - template: ${{ variables.CI_TEMPLATE_PATH }}/steps/dependencies-vendor.yml
-        parameters:
-          dependencyList:
-            - catch2
     - template: ${{ variables.CI_TEMPLATE_PATH }}/steps/checkout.yml
       parameters:
         checkoutRepo: ${{ parameters.checkoutRepo }}
@@ -147,7 +137,6 @@ jobs:
           -DORIGAMI_BUILD_SHARED_LIBS=ON
           -DORIGAMI_ENABLE_PYTHON=ON
           -DORIGAMI_BUILD_TESTING=ON
-          -DORIGAMI_ENABLE_FETCH=ON
           -GNinja
     - ${{ if ne(job.os, 'almalinux8') }}:
       - task: PublishPipelineArtifact@1
@@ -180,6 +169,7 @@ jobs:
       dependsOn: origami_build_${{ job.os }}
       condition:
         and(succeeded(),
+          eq(variables['ENABLE_${{ upper(job.target) }}_TESTS'], 'true'),
           not(containsValue(split(variables['DISABLED_${{ upper(job.target) }}_TESTS'], ','), '${{ parameters.componentName }}')),
           eq(${{ parameters.aggregatePipeline }}, False)
         )
@@ -190,30 +180,30 @@ jobs:
       workspace:
         clean: all
       steps:
+      - template: ${{ variables.CI_TEMPLATE_PATH }}/steps/checkout.yml
+        parameters:
+          checkoutRepo: ${{ parameters.checkoutRepo }}
+          sparseCheckoutDir: ${{ parameters.sparseCheckoutDir }}
       - template: ${{ variables.CI_TEMPLATE_PATH }}/steps/dependencies-other.yml
         parameters:
           aptPackages: ${{ parameters.aptPackages }}
           pipModules: ${{ parameters.pipModules }}
           packageManager: ${{ job.packageManager }}
-      - template: ${{ variables.CI_TEMPLATE_PATH }}/steps/dependencies-cmake-custom.yml
       - template: ${{ variables.CI_TEMPLATE_PATH }}/steps/preamble.yml
-      - template: ${{ variables.CI_TEMPLATE_PATH }}/steps/checkout.yml
-        parameters:
-          checkoutRepo: ${{ parameters.checkoutRepo }}
-          sparseCheckoutDir: ${{ parameters.sparseCheckoutDir }}
-      - template: ${{ variables.CI_TEMPLATE_PATH }}/steps/dependencies-vendor.yml
-        parameters:
-          dependencyList:
-            - gtest
-      - ${{ if ne(job.os, 'almalinux8') }}:
-        - template: ${{ variables.CI_TEMPLATE_PATH }}/steps/dependencies-vendor.yml
-          parameters:
-            dependencyList:
-              - catch2
       - template: ${{ variables.CI_TEMPLATE_PATH }}/steps/local-artifact-download.yml
         parameters:
           preTargetFilter: ${{ parameters.componentName }}
           os: ${{ job.os }}
+      - task: DownloadPipelineArtifact@2
+        displayName: 'Download Build Directory Artifact'
+        inputs:
+          artifact: '${{ parameters.componentName }}_${{ job.os }}_build_dir'
+          path: '$(Agent.BuildDirectory)/s/build'
+      - task: DownloadPipelineArtifact@2
+        displayName: 'Download Python Source Artifact'
+        inputs:
+          artifact: '${{ parameters.componentName }}_${{ job.os }}_python_src'
+          path: '$(Agent.BuildDirectory)/s/python'
       - template: ${{ variables.CI_TEMPLATE_PATH }}/steps/dependencies-rocm.yml
         parameters:
           checkoutRef: ${{ parameters.checkoutRef }}
@@ -222,72 +212,25 @@ jobs:
           gpuTarget: ${{ job.target }}
           ${{ if parameters.triggerDownstreamJobs }}:
             downstreamAggregateNames: ${{ parameters.downstreamAggregateNames }}
-      - task: CMake@1
-        displayName: 'Origami Test CMake Configuration'
-        inputs:
-          cmakeArgs: >-
-            -DCMAKE_PREFIX_PATH=$(Agent.BuildDirectory)/rocm;$(Agent.BuildDirectory)/vendor
-            -DCMAKE_CXX_COMPILER=$(Agent.BuildDirectory)/rocm/llvm/bin/amdclang++
-            -DORIGAMI_BUILD_SHARED_LIBS=ON
-            -DORIGAMI_ENABLE_PYTHON=ON
-            -DORIGAMI_BUILD_TESTING=ON
-            -GNinja
-            $(Agent.BuildDirectory)/s
-      - task: Bash@3
-        displayName: 'Build Origami Tests and Python Bindings'
-        inputs:
-          targetType: inline
-          workingDirectory: build
-          script: |
-            cmake --build . --target origami-tests origami_python -- -j$(nproc)
       - template: ${{ variables.CI_TEMPLATE_PATH }}/steps/gpu-diagnostics.yml
-      # Run tests using CTest (discovers and runs both C++ and Python tests)
       - template: ${{ variables.CI_TEMPLATE_PATH }}/steps/test.yml
         parameters:
           componentName: ${{ parameters.componentName }}
           os: ${{ job.os }}
-          testDir: 'build'
-          testParameters: '--output-on-failure --force-new-ctest-process --output-junit test_output.xml'
-      # Test pip install workflow
-      # - task: Bash@3
-      #   displayName: 'Test Pip Install'
-      #   inputs:
-      #     targetType: inline
-      #     script: |
-      #       set -e
-            
-      #       echo "==================================================================="
-      #       echo "Testing pip install workflow (pip install -e .)"
-      #       echo "==================================================================="
-            
-      #       # Set environment variables for pip install CMake build
-      #       export ROCM_PATH=$(Agent.BuildDirectory)/rocm
-      #       export CMAKE_PREFIX_PATH=$(Agent.BuildDirectory)/rocm:$(Agent.BuildDirectory)/vendor
-      #       export CMAKE_CXX_COMPILER=$(Agent.BuildDirectory)/rocm/llvm/bin/amdclang++
-            
-      #       echo "ROCM_PATH: $ROCM_PATH"
-      #       echo "CMAKE_PREFIX_PATH: $CMAKE_PREFIX_PATH"
-      #       echo "CMAKE_CXX_COMPILER: $CMAKE_CXX_COMPILER"
-      #       echo ""
-            
-      #       # Install from source directory
-      #       cd "$(Agent.BuildDirectory)/s/python"
-      #       pip install -e .
-            
-      #       # Verify import works
-      #       echo ""
-      #       echo "Verifying origami can be imported..."
-      #       python3 -c "import origami; print('✓ Successfully imported origami')"
-            
-      #       # Run pytest on installed package
-      #       echo ""
-      #       echo "Running pytest tests..."
-      #       python3 -m pytest tests/ -v -m "not slow" --tb=short
-            
-      #       echo ""
-      #       echo "==================================================================="
-      #       echo "Pip install test completed successfully"
-      #       echo "==================================================================="
+          testDir: '$(Agent.BuildDirectory)/rocm/bin'
+          testExecutable: './origami-tests'
+          testParameters: '--yaml origami-tests.yaml --gtest_output=xml:./test_output.xml --gtest_color=yes'
+      - script: |
+          set -e
+          export PYTHONPATH=$(Agent.BuildDirectory)/s/build/python:$PYTHONPATH
+
+          echo "--- Running origami_test.py ---"
+          python3 $(Agent.BuildDirectory)/s/python/origami_test.py
+          
+          echo "--- Running origami_grid_test.py ---"
+          python3 $(Agent.BuildDirectory)/s/python/origami_grid_test.py
+        displayName: 'Run Python Binding Tests'
+        condition: succeeded()
       - template: ${{ variables.CI_TEMPLATE_PATH }}/steps/docker-container.yml
         parameters:
           aptPackages: ${{ parameters.aptPackages }}

.azuredevops/components/rocm-examples.yml

@@ -30,7 +30,6 @@ parameters:
     - python3-pip
     - protobuf-compiler
     - libprotoc-dev
-    - libopencv-dev
 - name: pipModules
   type: object
   default:
@@ -65,7 +64,6 @@ parameters:
     - MIVisionX
     - rocm_smi_lib
     - rccl
-    - rocAL
     - rocALUTION
     - rocBLAS
     - rocDecode
@@ -105,7 +103,6 @@ parameters:
     - MIVisionX
     - rocm_smi_lib
     - rccl
-    - rocAL
     - rocALUTION
     - rocBLAS
     - rocDecode

.wordlist.txt

@@ -36,6 +36,7 @@ Andrej
 Arb
 Autocast
 autograd
+Backported
 BARs
 BatchNorm
 BLAS
@@ -203,9 +204,11 @@ GenAI
 GenZ
 GitHub
 Gitpod
+hardcoded
 HBM
 HCA
 HGX
+HLO
 HIPCC
 hipDataType
 HIPExtension
@@ -333,6 +336,7 @@ MoEs
 Mooncake
 Mpops
 Multicore
+multihost
 Multithreaded
 mx
 MXFP
@@ -1027,6 +1031,7 @@ uncacheable
 uncorrectable
 underoptimized
 unhandled
+unfused
 uninstallation
 unmapped
 unsqueeze

RELEASE.md

@@ -270,26 +270,26 @@ The [ROCm examples repository](https://github.com/ROCm/rocm-examples) has been e
 :margin: auto 0 auto auto
 :::{grid}
 :margin: auto 0 auto auto
-* [hipBLASLt](https://rocm.docs.amd.com/projects/hipBLASLt/en/latest/)
-* [hipSPARSE](https://rocm.docs.amd.com/projects/hipSPARSE/en/latest/)
-* [hipSPARSELt](https://rocm.docs.amd.com/projects/hipSPARSELt/en/latest/)
-* [hipTensor](https://rocm.docs.amd.com/projects/hipTensor/en/latest/)
+* [hipBLASLt](https://github.com/ROCm/rocm-examples/tree/amd-staging/Libraries/hipBLASLt)
+* [hipSPARSE](https://github.com/ROCm/rocm-examples/tree/amd-staging/Libraries/hipSPARSE)
+* [hipSPARSELt](https://github.com/ROCm/rocm-examples/tree/amd-staging/Libraries/hipSPARSELt)
+* [hipTensor](https://github.com/ROCm/rocm-examples/tree/amd-staging/Libraries/hipTensor)
 :::
 :::{grid}
 :margin: auto 0 auto auto
-* [rocALUTION](https://rocm.docs.amd.com/projects/rocALUTION/en/latest/)
-* [ROCprofiler-SDK](https://rocm.docs.amd.com/projects/rocprofiler-sdk/en/latest/)
-* [rocWMMA](https://rocm.docs.amd.com/projects/rocWMMA/en/latest/)
+* [rocALUTION](https://github.com/ROCm/rocm-examples/tree/amd-staging/Libraries/rocALUTION)
+* [ROCprofiler-SDK](https://github.com/ROCm/rocm-examples/tree/amd-staging/Libraries/rocProfiler-SDK)
+* [rocWMMA](https://github.com/ROCm/rocm-examples/tree/amd-staging/Libraries/rocWMMA)
 :::
 ::::
 
 Usage examples are now available for the following performance analysis tools:
 
-* [ROCm Compute Profiler](https://rocm.docs.amd.com/projects/rocprofiler-compute/en/latest/index.html)
-* [ROCm Systems Profiler](https://rocm.docs.amd.com/projects/rocprofiler-systems/en/latest/index.html)
-* [rocprofv3](https://rocm.docs.amd.com/projects/rocprofiler-sdk/en/latest/how-to/using-rocprofv3.html)
+* [ROCm Compute Profiler](https://github.com/ROCm/rocm-examples/tree/amd-staging/Tools/rocprof-compute)
+* [ROCm Systems Profiler](https://github.com/ROCm/rocm-examples/tree/amd-staging/Tools/rocprof-systems)
+* [rocprofv3](https://github.com/ROCm/rocm-examples/tree/amd-staging/Tools/rocprofv3)
 
-The complete source code for the [HIP Graph Tutorial](https://rocm.docs.amd.com/projects/HIP/en/latest/tutorial/graph_api.html) is also available as part of the ROCm examples.
+The complete source code for the [HIP Graph Tutorial](https://github.com/ROCm/rocm-examples/tree/amd-staging/HIP-Doc/Tutorials/graph_api) is also available as part of the ROCm examples.
 
 ### ROCm documentation updates
 

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

@@ -269,6 +269,33 @@ For a complete and up-to-date list of JAX public modules (for example, ``jax.num
   JAX API modules are maintained by the JAX project and is subject to change.
   Refer to the official Jax documentation for the most up-to-date information.
 
+Key features and enhancements for ROCm 7.1
+===============================================================================
+
+- Enabled compilation of multihost HLO runner Python bindings.
+
+  - Backported multihost HLO runner bindings and some related changes to
+    :code:`FunctionalHloRunner`.
+
+  - Added :code:`requirements_lock_3_12` to enable building for Python 3.12.
+
+- Removed hardcoded NHWC convolution layout for ``fp16`` precision to address the performance drops for ``fp16`` precision on gfx12xx GPUs.
+
+
+- ROCprofiler-SDK integration:
+
+  - Integrated ROCprofiler-SDK (v3) to XLA to improve profiling of GPU events,
+    support both time-based and step-based profiling.
+
+  - Added unit tests for :code:`rocm_collector` and :code:`rocm_tracer`.
+
+- Added Triton unsupported conversion from ``f8E4M3FNUZ`` to ``fp16`` with
+  rounding mode.
+
+- Introduced :code:`CudnnFusedConvDecomposer` to revert fused convolutions
+  when :code:`ConvAlgorithmPicker` fails to find a fused algorithm, and removed
+  unfused fallback paths from :code:`RocmFusedConvRunner`.
+
 Key features and enhancements for ROCm 7.0
 ===============================================================================
 

docs/conf.py

@@ -268,6 +268,3 @@ html_context = {
     "granularity_type" : [('Coarse-grained', 'coarse-grained'), ('Fine-grained', 'fine-grained')],
     "scope_type" : [('Device', 'device'), ('System', 'system')]
 }
-
-# Disable figure and table numbering
-numfig = False

docs/how-to/rocm-for-ai/fine-tuning/multi-gpu-fine-tuning-and-inference.rst

@@ -44,7 +44,7 @@ Setting up the base implementation environment
 
    .. code-block:: shell
 
-      rocm-smi --showproductname
+      amd-smi static --board
 
 #. Check that your GPUs are available to PyTorch.
 
@@ -65,8 +65,8 @@ Setting up the base implementation environment
 
 .. tip::
 
-   During training and inference, you can check the memory usage by running the ``rocm-smi`` command in your terminal.
-   This tool helps you see shows which GPUs are involved.
+   During training and inference, you can check the memory usage by running the ``amd-smi`` command in your terminal.
+   This tool helps you see which GPUs are involved.
 
 
 .. _fine-tuning-llms-multi-gpu-hugging-face-accelerate:
@@ -91,10 +91,10 @@ Now, it's important to adjust how you load the model. Add the ``device_map`` par
 
    ...
    base_model_name = "meta-llama/Llama-2-7b-chat-hf"
-   
+
    # Load base model to GPU memory
    base_model = AutoModelForCausalLM.from_pretrained(
-           base_model_name, 
+           base_model_name,
            device_map = "auto",
            trust_remote_code = True)
    ...
@@ -130,7 +130,7 @@ After loading the model in this way, the model is fully ready to use the resourc
 torchtune for fine-tuning and inference
 =============================================
 
-`torchtune <https://meta-pytorch.org/torchtune/main/>`_ is a PyTorch-native library for easy single and multi-GPU 
+`torchtune <https://pytorch.org/torchtune/main/>`_ is a PyTorch-native library for easy single and multi-GPU
 model fine-tuning and inference with LLMs.
 
 #. Install torchtune using pip.
@@ -139,7 +139,7 @@ model fine-tuning and inference with LLMs.
 
       # Install torchtune with PyTorch release 2.2.2+
       pip install torchtune
-      
+
       # To confirm that the package is installed correctly
       tune --help
 
@@ -148,12 +148,12 @@ model fine-tuning and inference with LLMs.
    .. code-block:: shell
 
       usage: tune [-h] {download,ls,cp,run,validate} ...
-      
+
       Welcome to the TorchTune CLI!
-      
+
       options:
         -h, --help            show this help message and exit
-      
+
       subcommands:
         {download,ls,cp,run,validate}
 
@@ -194,11 +194,11 @@ model fine-tuning and inference with LLMs.
         apply_lora_to_output: False
         lora_rank: 8
         lora_alpha: 16
-      
+
       tokenizer:
         _component_: torchtune.models.llama2.llama2_tokenizer
         path: /tmp/Llama-2-7b-hf/tokenizer.model
-      
+
       # Dataset and sampler
       dataset:
         _component_: torchtune.datasets.alpaca_cleaned_dataset

docs/how-to/rocm-for-ai/fine-tuning/single-gpu-fine-tuning-and-inference.rst

@@ -44,20 +44,19 @@ Setting up the base implementation environment
 
    .. code-block:: shell
 
-      rocm-smi --showproductname
+      amd-smi static --board
 
    Your output should look like this:
 
    .. code-block:: shell
 
-      ============================ ROCm System Management Interface ============================
-      ====================================== Product Info ======================================
-      GPU[0]          : Card Series:          AMD Instinct MI300X OAM
-      GPU[0]          : Card model:           0x74a1
-      GPU[0]          : Card vendor:          Advanced Micro Devices, Inc. [AMD/ATI]
-      GPU[0]          : Card SKU:             MI3SRIOV
-      ==========================================================================================
-      ================================== End of ROCm SMI Log ===================================
+      GPU: 0
+         BOARD:
+            MODEL_NUMBER: 102-G39203-0B
+            PRODUCT_SERIAL: PCB079220-1150
+            FRU_ID: 113-AMDG392030B04-100-300000097H
+            PRODUCT_NAME: AMD Instinct MI325 OAM
+            MANUFACTURER_NAME: AMD
 
 #. Check that your GPUs are available to PyTorch.
 
@@ -94,13 +93,13 @@ Setting up the base implementation environment
       pip install -r requirements-dev.txt
       cmake -DBNB_ROCM_ARCH="gfx942" -DCOMPUTE_BACKEND=hip -S .
       python setup.py install
-      
+
       # To leverage the SFTTrainer in TRL for model fine-tuning.
       pip install trl
-      
+
       # To leverage PEFT for efficiently adapting pre-trained language models .
       pip install peft
-      
+
       # Install the other dependencies.
       pip install transformers datasets huggingface-hub scipy
 
@@ -132,7 +131,7 @@ Download the base model and fine-tuning dataset
 
    .. note::
 
-      You can also use the `NousResearch Llama-2-7b-chat-hf <https://huggingface.co/NousResearch/Llama-2-7b-chat-hf>`_ 
+      You can also use the `NousResearch Llama-2-7b-chat-hf <https://huggingface.co/NousResearch/Llama-2-7b-chat-hf>`_
       as a substitute. It has the same model weights as the original.
 
 #. Run the following code to load the base model and tokenizer.
@@ -141,14 +140,14 @@ Download the base model and fine-tuning dataset
 
       # Base model and tokenizer names.
       base_model_name = "meta-llama/Llama-2-7b-chat-hf"
-      
+
       # Load base model to GPU memory.
       device = "cuda:0"
       base_model = AutoModelForCausalLM.from_pretrained(base_model_name, trust_remote_code = True).to(device)
-      
+
       # Load tokenizer.
       tokenizer = AutoTokenizer.from_pretrained(
-              base_model_name, 
+              base_model_name,
               trust_remote_code = True)
       tokenizer.pad_token = tokenizer.eos_token
       tokenizer.padding_side = "right"
@@ -162,10 +161,10 @@ Download the base model and fine-tuning dataset
       # Dataset for fine-tuning.
       training_dataset_name = "mlabonne/guanaco-llama2-1k"
       training_dataset = load_dataset(training_dataset_name, split = "train")
-      
+
       # Check the data.
       print(training_dataset)
-      
+
       # Dataset 11 is a QA sample in English.
       print(training_dataset[11])
 
@@ -252,8 +251,8 @@ Compare the number of trainable parameters and training time under the two diffe
                     dataset_text_field = "text",
                     tokenizer = tokenizer,
                     args = training_arguments
-            ) 
-            
+            )
+
             # Run the trainer.
             sft_trainer.train()
 
@@ -286,7 +285,7 @@ Compare the number of trainable parameters and training time under the two diffe
                     if param.requires_grad:
                         trainable_params += param.numel()
                 print(f"trainable params: {trainable_params} || all params: {all_param} || trainable%: {100 * trainable_params / all_param:.2f}")
-            
+
             sft_trainer.peft_config = None
             print_trainable_parameters(sft_trainer.model)
 
@@ -309,8 +308,8 @@ Compare the number of trainable parameters and training time under the two diffe
                     dataset_text_field = "text",
                     tokenizer = tokenizer,
                     args = training_arguments
-            ) 
-            
+            )
+
             # Training.
             trainer_full.train()
 
@@ -349,7 +348,7 @@ store, and load.
 
          # PEFT adapter name.
          adapter_name = "llama-2-7b-enhanced-adapter"
-         
+
          # Save PEFT adapter.
          sft_trainer.model.save_pretrained(adapter_name)
 
@@ -359,21 +358,21 @@ store, and load.
 
          # Access adapter directory.
          cd llama-2-7b-enhanced-adapter
-         
+
          # List all adapter files.
          README.md  adapter_config.json  adapter_model.safetensors
 
    .. tab-item:: Saving a fully fine-tuned model
       :sync: without
 
-      If you're not using LoRA and PEFT so there is no PEFT LoRA configuration used for training, use the following code 
+      If you're not using LoRA and PEFT so there is no PEFT LoRA configuration used for training, use the following code
       to save your fine-tuned model to your system.
 
       .. code-block:: python
 
          # Fully fine-tuned model name.
          new_model_name = "llama-2-7b-enhanced"
-         
+
          # Save the fully fine-tuned model.
          full_trainer.model.save_pretrained(new_model_name)
 
@@ -383,7 +382,7 @@ store, and load.
 
          # Access new model directory.
          cd llama-2-7b-enhanced
-         
+
          # List all model files.
          config.json                       model-00002-of-00006.safetensors  model-00005-of-00006.safetensors
          generation_config.json            model-00003-of-00006.safetensors  model-00006-of-00006.safetensors
@@ -412,26 +411,26 @@ Let's look at achieving model inference using these types of models.
 
    .. tab-item:: Inference using PEFT adapters
 
-      To use PEFT adapters like a normal transformer model, you can run the generation by loading a base model along with PEFT 
+      To use PEFT adapters like a normal transformer model, you can run the generation by loading a base model along with PEFT
       adapters as follows.
 
       .. code-block:: python
 
          from peft import PeftModel
          from transformers import AutoModelForCausalLM
-         
+
          # Set the path of the model or the name on Hugging face hub
          base_model_name = "meta-llama/Llama-2-7b-chat-hf"
-         
+
          # Set the path of the adapter
          adapter_name = "Llama-2-7b-enhanced-adpater"
-         
-         # Load base model 
+
+         # Load base model
          base_model = AutoModelForCausalLM.from_pretrained(base_model_name)
-         
-         # Adapt the base model with the adapter 
+
+         # Adapt the base model with the adapter
          new_model = PeftModel.from_pretrained(base_model, adapter_name)
-         
+
          # Then, run generation as the same with a normal model outlined in 2.1
 
       The PEFT library provides a ``merge_and_unload`` method, which merges the adapter layers into the base model. This is
@@ -439,13 +438,13 @@ Let's look at achieving model inference using these types of models.
 
       .. code-block:: python
 
-         # Load base model 
+         # Load base model
          base_model = AutoModelForCausalLM.from_pretrained(base_model_name)
-         
-         # Adapt the base model with the adapter 
+
+         # Adapt the base model with the adapter
          new_model = PeftModel.from_pretrained(base_model, adapter_name)
-         
-         # Merge adapter 
+
+         # Merge adapter
          model = model.merge_and_unload()
 
          # Save the merged model into local
@@ -461,25 +460,25 @@ Let's look at achieving model inference using these types of models.
 
          # Import relevant class for loading model and tokenizer
          from transformers import AutoTokenizer, AutoModelForCausalLM
-         
+
          # Set the pre-trained model name on Hugging face hub
          model_name = "meta-llama/Llama-2-7b-chat-hf"
-         
-         # Set device type 
+
+         # Set device type
          device = "cuda:0"
-         
-         # Load model and tokenizer 
+
+         # Load model and tokenizer
          model = AutoModelForCausalLM.from_pretrained(model_name).to(device)
          tokenizer = AutoTokenizer.from_pretrained(model_name)
-         
-         # Input prompt encoding 
+
+         # Input prompt encoding
          query = "What is a large language model?"
          inputs = tokenizer.encode(query, return_tensors="pt").to(device)
-         
-         # Token generation  
-         outputs = model.generate(inputs) 
-         
-         # Outputs decoding 
+
+         # Token generation
+         outputs = model.generate(inputs)
+
+         # Outputs decoding
          print(tokenizer.decode(outputs[0]))
 
       In addition, pipelines from Transformers offer simple APIs to use pre-trained models for different tasks, including
@@ -490,14 +489,14 @@ Let's look at achieving model inference using these types of models.
 
          # Import relevant class for loading model and tokenizer
          from transformers import pipeline
-         
+
          # Set the path of your model or the name on Hugging face hub
          model_name_or_path = "meta-llama/Llama-2-7b-chat-hf"
-         
-         # Set pipeline 
+
+         # Set pipeline
          # A positive device value will run the model on associated CUDA device id
          pipe = pipeline("text-generation", model=model_name_or_path, device=0)
-         
+
          # Token generation
          print(pipe("What is a large language model?")[0]["generated_text"])
 

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

@@ -31,16 +31,16 @@ in the Instinct documentation for more information.
 Hardware verification with ROCm
 -------------------------------
 
-Use the command ``rocm-smi --setperfdeterminism 1900`` to set the max clock speed up to 1900 MHz
+Use the command ``amd-smi set --perf-determinism 1900`` to set the max clock speed up to 1900 MHz
 instead of the default 2100 MHz. This can reduce the chance of a PCC event lowering the attainable
 GPU clocks. This setting will not be required for new IFWI releases with the production PRC feature.
-You can restore this setting to its default value with the ``rocm-smi -r`` command.
+You can restore this setting to its default value with the ``amd-smi reset --clocks`` command.
 
 Run the command:
 
 .. code-block:: shell
 
-   rocm-smi --setperfdeterminism 1900
+   amd-smi set --perf-determinism 1900
 
 See `Hardware verfication for ROCm <https://instinct.docs.amd.com/projects/amdgpu-docs/en/latest/system-optimization/mi300x.html#hardware-verification-with-rocm>`_
 in the Instinct documentation for more information.

docs/how-to/rocm-for-ai/training/benchmark-docker/previous-versions/megatron-lm-v24.12-dev.rst

@@ -108,16 +108,16 @@ for more information.
 Hardware verification with ROCm
 -------------------------------
 
-Use the command ``rocm-smi --setperfdeterminism 1900`` to set the max clock speed up to 1900 MHz
+Use the command ``amd-smi set --perf-determinism 1900`` to set the max clock speed up to 1900 MHz
 instead of the default 2100 MHz. This can reduce the chance of a PCC event lowering the attainable
 GPU clocks. This setting will not be required for new IFWI releases with the production PRC feature.
-You can restore this setting to its default value with the ``rocm-smi -r`` command.
+You can restore this setting to its default value with the ``amd-smi reset --clocks`` command.
 
 Run the command:
 
 .. code-block:: shell
 
-   rocm-smi --setperfdeterminism 1900
+   amd-smi set --perf-determinism 1900
 
 See `Hardware verification with ROCm <https://instinct.docs.amd.com/projects/amdgpu-docs/en/latest/system-optimization/mi300x.html#hardware-verification-with-rocm>`_ for more information.
 
@@ -248,7 +248,7 @@ Download the Docker image and required packages
       Checking out this specific commit is recommended for a stable and reproducible environment.
 
       .. code-block:: shell
-         
+
          git checkout bb93ccbfeae6363c67b361a97a27c74ab86e7e92
 
 Prepare training datasets

docs/sphinx/requirements.in

@@ -1,4 +1,4 @@
-rocm-docs-core==1.31.1
+rocm-docs-core==1.30.0
 sphinx-reredirects
 sphinx-sitemap
 sphinxcontrib.datatemplates==0.11.0

docs/sphinx/requirements.txt

@@ -132,7 +132,6 @@ nest-asyncio==1.6.0
 packaging==25.0
     # via
     #   ipykernel
-    #   pydata-sphinx-theme
     #   sphinx
 parso==0.8.5
     # via jedi
@@ -150,7 +149,7 @@ pure-eval==0.2.3
     # via stack-data
 pycparser==2.23
     # via cffi
-pydata-sphinx-theme==0.15.4
+pydata-sphinx-theme==0.16.1
     # via
     #   rocm-docs-core
     #   sphinx-book-theme
@@ -164,7 +163,7 @@ pygments==2.19.2
     #   sphinx
 pyjwt[crypto]==2.10.1
     # via pygithub
-pynacl==1.6.2
+pynacl==1.6.1
     # via pygithub
 python-dateutil==2.9.0.post0
     # via jupyter-client
@@ -188,7 +187,7 @@ requests==2.32.5
     # via
     #   pygithub
     #   sphinx
-rocm-docs-core==1.31.1
+rocm-docs-core==1.30.0
     # via -r requirements.in
 rpds-py==0.29.0
     # via
@@ -218,7 +217,7 @@ sphinx==8.1.3
     #   sphinx-reredirects
     #   sphinxcontrib-datatemplates
     #   sphinxcontrib-runcmd
-sphinx-book-theme==1.1.4
+sphinx-book-theme==1.1.3
     # via rocm-docs-core
 sphinx-copybutton==0.5.2
     # via rocm-docs-core
@@ -282,7 +281,7 @@ typing-extensions==4.15.0
     #   pygithub
     #   referencing
     #   sqlalchemy
-urllib3==2.6.3
+urllib3==2.5.0
     # via
     #   pygithub
     #   requests

docs/what-is-rocm.rst

@@ -123,7 +123,8 @@ Performance
 
 .. note::
 
-  `ROCprof Compute Viewer <https://rocm.docs.amd.com/projects/rocprof-compute-viewer/en/amd-mainline/>`_ is a tool for visualizing and analyzing GPU thread trace data collected using :doc:`rocprofv3 <rocprofiler-sdk:index>`. Note that `ROCprof Compute Viewer <https://rocm.docs.amd.com/projects/rocprof-compute-viewer/en/amd-mainline/>`_ is in an early access state. Running production workloads is not recommended.
+  `ROCprof Compute Viewer <https://rocm.docs.amd.com/projects/rocprof-compute-viewer/en/amd-mainline/>`_ is a tool for visualizing and analyzing GPU thread trace data collected using :doc:`rocprofv3 <rocprofiler-sdk:index>`.
+  Note that `ROCprof Compute Viewer <https://rocm.docs.amd.com/projects/rocprof-compute-viewer/en/amd-mainline/>`_ is in an early access state. Running production workloads is not recommended.
 
 Development
 ^^^^^^^^^^^