[ROCm] upgrade nightly wheels to rocm6.4 (#151355)

Pull Request resolved: https://github.com/pytorch/pytorch/pull/151355
Approved by: https://github.com/jeffdaily

Co-authored-by: Jeff Daily <jeff.daily@amd.com>

.ci/docker/common/install_rocm_drm.sh

@@ -25,9 +25,7 @@ python3 -m pip install meson ninja
 ###########################
 ### clone repo
 ###########################
-# TEMPORARY FIX: https://gitlab.freedesktop.org/mesa/drm.git is down until 2025/03/22
+GIT_SSL_NO_VERIFY=true git clone https://gitlab.freedesktop.org/mesa/drm.git
-# GIT_SSL_NO_VERIFY=true git clone https://gitlab.freedesktop.org/mesa/drm.git
-GIT_SSL_NO_VERIFY=true git clone git://anongit.freedesktop.org/mesa/drm
 pushd drm
 
 ###########################

.github/scripts/generate_binary_build_matrix.py

@@ -30,7 +30,7 @@ CUDA_ARCHES_CUDNN_VERSION = {
 }
 
 # NOTE: Also update the ROCm sources in tools/nightly.py when changing this list
-ROCM_ARCHES = ["6.2.4", "6.3"]
+ROCM_ARCHES = ["6.3", "6.4"]
 
 XPU_ARCHES = ["xpu"]
 

.github/workflows/build-libtorch-images.yml

@@ -87,7 +87,7 @@ jobs:
     runs-on: "${{ needs.get-label-type.outputs.label-type }}linux.9xlarge.ephemeral"
     strategy:
       matrix:
-        rocm_version: ["6.2.4", "6.3", "6.4"]
+        rocm_version: ["6.3", "6.4"]
     env:
       GPU_ARCH_TYPE: rocm
       GPU_ARCH_VERSION: ${{ matrix.rocm_version }}

.github/workflows/build-magma-rocm-linux.yml

@@ -34,7 +34,7 @@ jobs:
       id-token: write
     strategy:
       matrix:
-        rocm_version: ["64", "63", "624"]
+        rocm_version: ["64", "63"]
     steps:
       - name: Checkout PyTorch
         uses: actions/checkout@v4

.github/workflows/build-manywheel-images.yml

@@ -133,7 +133,7 @@ jobs:
     runs-on: "${{ needs.get-label-type.outputs.label-type }}linux.9xlarge.ephemeral"
     strategy:
       matrix:
-        rocm_version: ["6.2.4", "6.3", "6.4"]
+        rocm_version: ["6.3", "6.4"]
     env:
       GPU_ARCH_TYPE: rocm-manylinux_2_28
       GPU_ARCH_VERSION: ${{ matrix.rocm_version }}

.github/workflows/build-triton-wheel.yml

@@ -54,7 +54,7 @@ jobs:
         docker-image: ["pytorch/manylinux2_28-builder:cpu"]
         include:
           - device: "rocm"
-            rocm_version: "6.3"
+            rocm_version: "6.4"
             runs_on: "${{ needs.get-label-type.outputs.label-type }}linux.4xlarge"
           - device: "cuda"
             rocm_version: ""

.github/workflows/generated-linux-binary-libtorch-nightly.yml

@@ -301,98 +301,6 @@ jobs:
       github-token: ${{ secrets.GITHUB_TOKEN }}
     uses: ./.github/workflows/_binary-upload.yml
 
-  libtorch-rocm6_2_4-shared-with-deps-release-build:
-    if: ${{ github.repository_owner == 'pytorch' }}
-    uses: ./.github/workflows/_binary-build-linux.yml
-    needs: get-label-type
-    with:
-      PYTORCH_ROOT: /pytorch
-      PACKAGE_TYPE: libtorch
-      # TODO: This is a legacy variable that we eventually want to get rid of in
-      #       favor of GPU_ARCH_VERSION
-      DESIRED_CUDA: rocm6.2.4
-      GPU_ARCH_VERSION: 6.2.4
-      GPU_ARCH_TYPE: rocm
-      DOCKER_IMAGE: pytorch/libtorch-cxx11-builder:rocm6.2.4-main
-      LIBTORCH_CONFIG: release
-      LIBTORCH_VARIANT: shared-with-deps
-      runner_prefix: "${{ needs.get-label-type.outputs.label-type }}"
-      build_name: libtorch-rocm6_2_4-shared-with-deps-release
-      build_environment: linux-binary-libtorch
-    secrets:
-      github-token: ${{ secrets.GITHUB_TOKEN }}
-  libtorch-rocm6_2_4-shared-with-deps-release-test:  # Testing
-    if: ${{ github.repository_owner == 'pytorch' }}
-    needs:
-      - libtorch-rocm6_2_4-shared-with-deps-release-build
-      - get-label-type
-    runs-on: linux.rocm.gpu
-    timeout-minutes: 240
-    env:
-      PYTORCH_ROOT: /pytorch
-      PACKAGE_TYPE: libtorch
-      # TODO: This is a legacy variable that we eventually want to get rid of in
-      #       favor of GPU_ARCH_VERSION
-      DESIRED_CUDA: rocm6.2.4
-      GPU_ARCH_VERSION: 6.2.4
-      GPU_ARCH_TYPE: rocm
-      SKIP_ALL_TESTS: 1
-      DOCKER_IMAGE: pytorch/libtorch-cxx11-builder:rocm6.2.4-main
-      LIBTORCH_CONFIG: release
-      LIBTORCH_VARIANT: shared-with-deps
-    steps:
-      - name: Setup ROCm
-        uses: ./.github/actions/setup-rocm
-      - uses: actions/download-artifact@v4.1.7
-        name: Download Build Artifacts
-        with:
-          name: libtorch-rocm6_2_4-shared-with-deps-release
-          path: "${{ runner.temp }}/artifacts/"
-      - name: Checkout PyTorch
-        uses: actions/checkout@v4
-        with:
-          ref: ${{ github.event_name == 'pull_request' && github.event.pull_request.head.sha || github.sha }}
-          submodules: recursive
-          path: pytorch
-          show-progress: false
-      - name: Clean PyTorch checkout
-        run: |
-          # Remove any artifacts from the previous checkouts
-          git clean -fxd
-        working-directory: pytorch
-      - name: ROCm set GPU_FLAG
-        run: |
-          echo "GPU_FLAG=--device=/dev/mem --device=/dev/kfd --device=/dev/dri --group-add video --group-add daemon" >> "${GITHUB_ENV}"
-      - name: Pull Docker image
-        uses: pytorch/test-infra/.github/actions/pull-docker-image@main
-        with:
-          docker-image: pytorch/libtorch-cxx11-builder:rocm6.2.4-main
-      - name: Test Pytorch binary
-        uses: ./pytorch/.github/actions/test-pytorch-binary
-      - name: Teardown ROCm
-        uses: ./.github/actions/teardown-rocm
-  libtorch-rocm6_2_4-shared-with-deps-release-upload:  # Uploading
-    if: ${{ github.repository_owner == 'pytorch' }}
-    permissions:
-      id-token: write
-      contents: read
-    needs: libtorch-rocm6_2_4-shared-with-deps-release-test
-    with:
-      PYTORCH_ROOT: /pytorch
-      PACKAGE_TYPE: libtorch
-      # TODO: This is a legacy variable that we eventually want to get rid of in
-      #       favor of GPU_ARCH_VERSION
-      DESIRED_CUDA: rocm6.2.4
-      GPU_ARCH_VERSION: 6.2.4
-      GPU_ARCH_TYPE: rocm
-      DOCKER_IMAGE: pytorch/libtorch-cxx11-builder:rocm6.2.4-main
-      LIBTORCH_CONFIG: release
-      LIBTORCH_VARIANT: shared-with-deps
-      build_name: libtorch-rocm6_2_4-shared-with-deps-release
-    secrets:
-      github-token: ${{ secrets.GITHUB_TOKEN }}
-    uses: ./.github/workflows/_binary-upload.yml
-
   libtorch-rocm6_3-shared-with-deps-release-build:
     if: ${{ github.repository_owner == 'pytorch' }}
     uses: ./.github/workflows/_binary-build-linux.yml
@@ -484,3 +392,95 @@ jobs:
     secrets:
       github-token: ${{ secrets.GITHUB_TOKEN }}
     uses: ./.github/workflows/_binary-upload.yml
+
+  libtorch-rocm6_4-shared-with-deps-release-build:
+    if: ${{ github.repository_owner == 'pytorch' }}
+    uses: ./.github/workflows/_binary-build-linux.yml
+    needs: get-label-type
+    with:
+      PYTORCH_ROOT: /pytorch
+      PACKAGE_TYPE: libtorch
+      # TODO: This is a legacy variable that we eventually want to get rid of in
+      #       favor of GPU_ARCH_VERSION
+      DESIRED_CUDA: rocm6.4
+      GPU_ARCH_VERSION: 6.4
+      GPU_ARCH_TYPE: rocm
+      DOCKER_IMAGE: pytorch/libtorch-cxx11-builder:rocm6.4-main
+      LIBTORCH_CONFIG: release
+      LIBTORCH_VARIANT: shared-with-deps
+      runner_prefix: "${{ needs.get-label-type.outputs.label-type }}"
+      build_name: libtorch-rocm6_4-shared-with-deps-release
+      build_environment: linux-binary-libtorch
+    secrets:
+      github-token: ${{ secrets.GITHUB_TOKEN }}
+  libtorch-rocm6_4-shared-with-deps-release-test:  # Testing
+    if: ${{ github.repository_owner == 'pytorch' }}
+    needs:
+      - libtorch-rocm6_4-shared-with-deps-release-build
+      - get-label-type
+    runs-on: linux.rocm.gpu
+    timeout-minutes: 240
+    env:
+      PYTORCH_ROOT: /pytorch
+      PACKAGE_TYPE: libtorch
+      # TODO: This is a legacy variable that we eventually want to get rid of in
+      #       favor of GPU_ARCH_VERSION
+      DESIRED_CUDA: rocm6.4
+      GPU_ARCH_VERSION: 6.4
+      GPU_ARCH_TYPE: rocm
+      SKIP_ALL_TESTS: 1
+      DOCKER_IMAGE: pytorch/libtorch-cxx11-builder:rocm6.4-main
+      LIBTORCH_CONFIG: release
+      LIBTORCH_VARIANT: shared-with-deps
+    steps:
+      - name: Setup ROCm
+        uses: ./.github/actions/setup-rocm
+      - uses: actions/download-artifact@v4.1.7
+        name: Download Build Artifacts
+        with:
+          name: libtorch-rocm6_4-shared-with-deps-release
+          path: "${{ runner.temp }}/artifacts/"
+      - name: Checkout PyTorch
+        uses: actions/checkout@v4
+        with:
+          ref: ${{ github.event_name == 'pull_request' && github.event.pull_request.head.sha || github.sha }}
+          submodules: recursive
+          path: pytorch
+          show-progress: false
+      - name: Clean PyTorch checkout
+        run: |
+          # Remove any artifacts from the previous checkouts
+          git clean -fxd
+        working-directory: pytorch
+      - name: ROCm set GPU_FLAG
+        run: |
+          echo "GPU_FLAG=--device=/dev/mem --device=/dev/kfd --device=/dev/dri --group-add video --group-add daemon" >> "${GITHUB_ENV}"
+      - name: Pull Docker image
+        uses: pytorch/test-infra/.github/actions/pull-docker-image@main
+        with:
+          docker-image: pytorch/libtorch-cxx11-builder:rocm6.4-main
+      - name: Test Pytorch binary
+        uses: ./pytorch/.github/actions/test-pytorch-binary
+      - name: Teardown ROCm
+        uses: ./.github/actions/teardown-rocm
+  libtorch-rocm6_4-shared-with-deps-release-upload:  # Uploading
+    if: ${{ github.repository_owner == 'pytorch' }}
+    permissions:
+      id-token: write
+      contents: read
+    needs: libtorch-rocm6_4-shared-with-deps-release-test
+    with:
+      PYTORCH_ROOT: /pytorch
+      PACKAGE_TYPE: libtorch
+      # TODO: This is a legacy variable that we eventually want to get rid of in
+      #       favor of GPU_ARCH_VERSION
+      DESIRED_CUDA: rocm6.4
+      GPU_ARCH_VERSION: 6.4
+      GPU_ARCH_TYPE: rocm
+      DOCKER_IMAGE: pytorch/libtorch-cxx11-builder:rocm6.4-main
+      LIBTORCH_CONFIG: release
+      LIBTORCH_VARIANT: shared-with-deps
+      build_name: libtorch-rocm6_4-shared-with-deps-release
+    secrets:
+      github-token: ${{ secrets.GITHUB_TOKEN }}
+    uses: ./.github/workflows/_binary-upload.yml

tools/nightly.py

@@ -141,9 +141,9 @@ PIP_SOURCES = {
         supported_platforms={"Linux", "Windows"},
         accelerator="cuda",
     ),
-    "rocm-6.2.4": PipSource(
+    "rocm-6.4": PipSource(
-        name="rocm-6.2.4",
+        name="rocm-6.4",
-        index_url=f"{PYTORCH_NIGHTLY_PIP_INDEX_URL}/rocm6.2.4",
+        index_url=f"{PYTORCH_NIGHTLY_PIP_INDEX_URL}/rocm6.4",
         supported_platforms={"Linux"},
         accelerator="rocm",
     ),