From 2392641d70f8704be19bfca888b2b5afdf349c06 Mon Sep 17 00:00:00 2001
From: "copilot-swe-agent[bot]" <198982749+Copilot@users.noreply.github.com>
Date: Fri, 22 Aug 2025 01:52:53 +0000
Subject: [PATCH 01/12] Initial plan


From 70833d1e32bc00c116a5246e9e68ba2810d8f5ee Mon Sep 17 00:00:00 2001
From: "copilot-swe-agent[bot]" <198982749+Copilot@users.noreply.github.com>
Date: Fri, 22 Aug 2025 01:59:37 +0000
Subject: [PATCH 02/12] Consolidate test matrices from workflows into
 ci/test-matrix.json

Co-authored-by: leofang <5534781+leofang@users.noreply.github.com>
---
 .github/workflows/test-wheel-linux.yml   | 80 +++++++-----------------
 .github/workflows/test-wheel-windows.yml | 34 ++++------
 ci/test-matrix.json                      | 65 +++++++++++++++++++
 3 files changed, 99 insertions(+), 80 deletions(-)
 create mode 100644 ci/test-matrix.json

diff --git a/.github/workflows/test-wheel-linux.yml b/.github/workflows/test-wheel-linux.yml
index b775670a9b..9e99a22a79 100644
--- a/.github/workflows/test-wheel-linux.yml
+++ b/.github/workflows/test-wheel-linux.yml
@@ -34,12 +34,18 @@ jobs:
     outputs:
       MATRIX: ${{ steps.compute-matrix.outputs.MATRIX }}
     steps:
+      - name: Checkout ${{ github.event.repository.name }}
+        uses: actions/checkout@08c6903cd8c0fde910a37f88322edcfb5dd907a8  # v5.0.0
+        with:
+          fetch-depth: 0
+
       - name: Validate Test Type
         run: |
           if [[ "$BUILD_TYPE" != "pull-request" ]] && [[ "$BUILD_TYPE" != "nightly" ]] && [[ "$BUILD_TYPE" != "branch" ]]; then
               echo "Invalid build type! Must be one of 'nightly', 'pull-request', or 'branch'."
               exit 1
           fi
+
       - name: Compute Python Test Matrix
         id: compute-matrix
         run: |
@@ -48,72 +54,30 @@ jobs:
           if [[ "${ARCH}" == "arm64" ]]; then
             gpu="a100"
           fi
-          # Add a special entry for the H100 runner on amd64.
-          special_runner=""
-          if [[ "${ARCH}" == "amd64" ]]; then
-            special_runner="- { ARCH: ${ARCH}, PY_VER: '3.13', CUDA_VER: '13.0.0', LOCAL_CTK: '1', GPU: 'H100', DRIVER: 'latest' }"
-          fi
-
-          # Please keep the matrices sorted in ascending order by the following:
-          #
-          #     [PY_VER, CUDA_VER, LOCAL_CTK, GPU, DRIVER]
-          #
-          # Note that DRIVER: `earliest` does not work with CUDA 12.9.0 and LOCAL_CTK: 0 does not work with CUDA 12.0.1.
-          #
-          export MATRICES="
-            pull-request:
-              - { ARCH: ${ARCH}, PY_VER: '3.9',  CUDA_VER: '12.9.0', LOCAL_CTK: '0', GPU: ${gpu}, DRIVER: 'latest' }
-              - { ARCH: ${ARCH}, PY_VER: '3.9',  CUDA_VER: '13.0.0', LOCAL_CTK: '1', GPU: ${gpu}, DRIVER: 'latest' }
-              - { ARCH: ${ARCH}, PY_VER: '3.10', CUDA_VER: '12.9.0', LOCAL_CTK: '1', GPU: ${gpu}, DRIVER: 'latest' }
-              - { ARCH: ${ARCH}, PY_VER: '3.10', CUDA_VER: '13.0.0', LOCAL_CTK: '0', GPU: ${gpu}, DRIVER: 'latest' }
-              - { ARCH: ${ARCH}, PY_VER: '3.11', CUDA_VER: '12.9.0', LOCAL_CTK: '0', GPU: ${gpu}, DRIVER: 'latest' }
-              - { ARCH: ${ARCH}, PY_VER: '3.11', CUDA_VER: '13.0.0', LOCAL_CTK: '1', GPU: ${gpu}, DRIVER: 'latest' }
-              - { ARCH: ${ARCH}, PY_VER: '3.12', CUDA_VER: '12.9.0', LOCAL_CTK: '1', GPU: ${gpu}, DRIVER: 'latest' }
-              - { ARCH: ${ARCH}, PY_VER: '3.12', CUDA_VER: '13.0.0', LOCAL_CTK: '0', GPU: ${gpu}, DRIVER: 'latest' }
-              - { ARCH: ${ARCH}, PY_VER: '3.13', CUDA_VER: '12.9.0', LOCAL_CTK: '0', GPU: ${gpu}, DRIVER: 'latest' }
-              - { ARCH: ${ARCH}, PY_VER: '3.13', CUDA_VER: '13.0.0', LOCAL_CTK: '1', GPU: ${gpu}, DRIVER: 'latest' }
-              ${special_runner}
-            nightly:
-              - { ARCH: ${ARCH}, PY_VER: '3.9',  CUDA_VER: '11.8.0', LOCAL_CTK: '0', GPU: ${gpu}, DRIVER: 'earliest' }
-              - { ARCH: ${ARCH}, PY_VER: '3.9',  CUDA_VER: '11.8.0', LOCAL_CTK: '1', GPU: ${gpu}, DRIVER: 'latest' }
-              - { ARCH: ${ARCH}, PY_VER: '3.9',  CUDA_VER: '12.0.1', LOCAL_CTK: '1', GPU: ${gpu}, DRIVER: 'latest' }
-              - { ARCH: ${ARCH}, PY_VER: '3.9',  CUDA_VER: '12.9.0', LOCAL_CTK: '0', GPU: ${gpu}, DRIVER: 'latest' }
-              - { ARCH: ${ARCH}, PY_VER: '3.9',  CUDA_VER: '12.9.0', LOCAL_CTK: '1', GPU: ${gpu}, DRIVER: 'latest' }
-              - { ARCH: ${ARCH}, PY_VER: '3.10', CUDA_VER: '11.8.0', LOCAL_CTK: '0', GPU: ${gpu}, DRIVER: 'earliest' }
-              - { ARCH: ${ARCH}, PY_VER: '3.10', CUDA_VER: '11.8.0', LOCAL_CTK: '1', GPU: ${gpu}, DRIVER: 'latest' }
-              - { ARCH: ${ARCH}, PY_VER: '3.10', CUDA_VER: '12.0.1', LOCAL_CTK: '1', GPU: ${gpu}, DRIVER: 'latest' }
-              - { ARCH: ${ARCH}, PY_VER: '3.10', CUDA_VER: '12.9.0', LOCAL_CTK: '0', GPU: ${gpu}, DRIVER: 'latest' }
-              - { ARCH: ${ARCH}, PY_VER: '3.10', CUDA_VER: '12.9.0', LOCAL_CTK: '1', GPU: ${gpu}, DRIVER: 'latest' }
-              - { ARCH: ${ARCH}, PY_VER: '3.11', CUDA_VER: '11.8.0', LOCAL_CTK: '0', GPU: ${gpu}, DRIVER: 'earliest' }
-              - { ARCH: ${ARCH}, PY_VER: '3.11', CUDA_VER: '11.8.0', LOCAL_CTK: '1', GPU: ${gpu}, DRIVER: 'latest' }
-              - { ARCH: ${ARCH}, PY_VER: '3.11', CUDA_VER: '12.0.1', LOCAL_CTK: '1', GPU: ${gpu}, DRIVER: 'latest' }
-              - { ARCH: ${ARCH}, PY_VER: '3.11', CUDA_VER: '12.9.0', LOCAL_CTK: '0', GPU: ${gpu}, DRIVER: 'latest' }
-              - { ARCH: ${ARCH}, PY_VER: '3.11', CUDA_VER: '12.9.0', LOCAL_CTK: '1', GPU: ${gpu}, DRIVER: 'latest' }
-              - { ARCH: ${ARCH}, PY_VER: '3.12', CUDA_VER: '11.8.0', LOCAL_CTK: '0', GPU: ${gpu}, DRIVER: 'earliest' }
-              - { ARCH: ${ARCH}, PY_VER: '3.12', CUDA_VER: '11.8.0', LOCAL_CTK: '1', GPU: ${gpu}, DRIVER: 'latest' }
-              - { ARCH: ${ARCH}, PY_VER: '3.12', CUDA_VER: '12.0.1', LOCAL_CTK: '1', GPU: ${gpu}, DRIVER: 'latest' }
-              - { ARCH: ${ARCH}, PY_VER: '3.12', CUDA_VER: '12.9.0', LOCAL_CTK: '0', GPU: ${gpu}, DRIVER: 'latest' }
-              - { ARCH: ${ARCH}, PY_VER: '3.12', CUDA_VER: '12.9.0', LOCAL_CTK: '1', GPU: ${gpu}, DRIVER: 'latest' }
-              - { ARCH: ${ARCH}, PY_VER: '3.13', CUDA_VER: '11.8.0', LOCAL_CTK: '0', GPU: ${gpu}, DRIVER: 'earliest' }
-              - { ARCH: ${ARCH}, PY_VER: '3.13', CUDA_VER: '11.8.0', LOCAL_CTK: '1', GPU: ${gpu}, DRIVER: 'latest' }
-              - { ARCH: ${ARCH}, PY_VER: '3.13', CUDA_VER: '12.0.1', LOCAL_CTK: '1', GPU: ${gpu}, DRIVER: 'latest' }
-              - { ARCH: ${ARCH}, PY_VER: '3.13', CUDA_VER: '12.9.0', LOCAL_CTK: '0', GPU: ${gpu}, DRIVER: 'latest' }
-              - { ARCH: ${ARCH}, PY_VER: '3.13', CUDA_VER: '12.9.0', LOCAL_CTK: '1', GPU: ${gpu}, DRIVER: 'latest' }
-              ${special_runner}
-          "
 
           # Use the nightly matrix for branch tests
           MATRIX_TYPE="${BUILD_TYPE}"
           if [[ "${MATRIX_TYPE}" == "branch" ]]; then
             MATRIX_TYPE="nightly"
           fi
-          export MATRIX_TYPE
-          TEST_MATRIX=$(yq -n 'env(MATRICES) | .[strenv(MATRIX_TYPE)]')
-          export TEST_MATRIX
+
+          # Read base matrix from JSON file and substitute GPU placeholder
+          TEST_MATRIX=$(jq --arg gpu "$gpu" --arg arch "$ARCH" '
+            .linux[env.MATRIX_TYPE] | 
+            map(. + {ARCH: $arch} | if .GPU == "{gpu}" then . + {GPU: $gpu} else . end)
+          ' ci/test-matrix.json)
+
+          # Add special runner for amd64 if applicable
+          if [[ "${ARCH}" == "amd64" ]]; then
+            SPECIAL_RUNNERS=$(jq --arg arch "$ARCH" '
+              .linux.special_runners.amd64 | 
+              map(. + {ARCH: $arch})
+            ' ci/test-matrix.json)
+            TEST_MATRIX=$(jq --argjson special "$SPECIAL_RUNNERS" '. + $special' <<< "$TEST_MATRIX")
+          fi
 
           MATRIX="$(
-            yq -n -o json 'env(TEST_MATRIX)' | \
-            jq -c '${{ inputs.matrix_filter }} | if (. | length) > 0 then {include: .} else "Error: Empty matrix\n" | halt_error(1) end'
+            jq -c '${{ inputs.matrix_filter }} | if (. | length) > 0 then {include: .} else "Error: Empty matrix\n" | halt_error(1) end' <<< "$TEST_MATRIX"
           )"
 
           echo "MATRIX=${MATRIX}" | tee --append "${GITHUB_OUTPUT}"
diff --git a/.github/workflows/test-wheel-windows.yml b/.github/workflows/test-wheel-windows.yml
index 18ddbcb45a..cdc9dfa6b8 100644
--- a/.github/workflows/test-wheel-windows.yml
+++ b/.github/workflows/test-wheel-windows.yml
@@ -32,6 +32,11 @@ jobs:
     outputs:
       MATRIX: ${{ steps.compute-matrix.outputs.MATRIX }}
     steps:
+      - name: Checkout ${{ github.event.repository.name }}
+        uses: actions/checkout@08c6903cd8c0fde910a37f88322edcfb5dd907a8  # v5.0.0
+        with:
+          fetch-depth: 0
+
       - name: Validate Test Type
         run: |
           if [[ "$BUILD_TYPE" != "pull-request" ]] && [[ "$BUILD_TYPE" != "nightly" ]] && [[ "$BUILD_TYPE" != "branch" ]]; then
@@ -41,35 +46,20 @@ jobs:
       - name: Compute Python Test Matrix
         id: compute-matrix
         run: |
-          # Please keep the matrices sorted in ascending order by the following:
-          #
-          #     [PY_VER, CUDA_VER, LOCAL_CTK]
-          #
-          export MATRICES="
-            pull-request:
-              - { ARCH: ${ARCH}, PY_VER: '3.12',  CUDA_VER: '12.9.0', LOCAL_CTK: '0' }
-              - { ARCH: ${ARCH}, PY_VER: '3.12',  CUDA_VER: '12.9.0', LOCAL_CTK: '1' }
-              - { ARCH: ${ARCH}, PY_VER: '3.13',  CUDA_VER: '13.0.0', LOCAL_CTK: '0' }
-              - { ARCH: ${ARCH}, PY_VER: '3.13',  CUDA_VER: '13.0.0', LOCAL_CTK: '1' }
-            nightly:
-              - { ARCH: ${ARCH}, PY_VER: '3.12',  CUDA_VER: '11.8.0', LOCAL_CTK: '0' }
-              - { ARCH: ${ARCH}, PY_VER: '3.12',  CUDA_VER: '11.8.0', LOCAL_CTK: '1' }
-              - { ARCH: ${ARCH}, PY_VER: '3.12',  CUDA_VER: '12.9.0', LOCAL_CTK: '0' }
-              - { ARCH: ${ARCH}, PY_VER: '3.12',  CUDA_VER: '12.9.0', LOCAL_CTK: '1' }
-          "
-
           # Use the nightly matrix for branch tests
           MATRIX_TYPE="${BUILD_TYPE}"
           if [[ "${MATRIX_TYPE}" == "branch" ]]; then
             MATRIX_TYPE="nightly"
           fi
-          export MATRIX_TYPE
-          TEST_MATRIX=$(yq -n 'env(MATRICES) | .[strenv(MATRIX_TYPE)]')
-          export TEST_MATRIX
+
+          # Read base matrix from JSON file and add ARCH field
+          TEST_MATRIX=$(jq --arg arch "$ARCH" '
+            .windows[env.MATRIX_TYPE] | 
+            map(. + {ARCH: $arch})
+          ' ci/test-matrix.json)
 
           MATRIX="$(
-            yq -n -o json 'env(TEST_MATRIX)' | \
-            jq -c '${{ inputs.matrix_filter }} | if (. | length) > 0 then {include: .} else "Error: Empty matrix\n" | halt_error(1) end'
+            jq -c '${{ inputs.matrix_filter }} | if (. | length) > 0 then {include: .} else "Error: Empty matrix\n" | halt_error(1) end' <<< "$TEST_MATRIX"
           )"
 
           echo "MATRIX=${MATRIX}" | tee --append "${GITHUB_OUTPUT}"
diff --git a/ci/test-matrix.json b/ci/test-matrix.json
new file mode 100644
index 0000000000..2491e18f37
--- /dev/null
+++ b/ci/test-matrix.json
@@ -0,0 +1,65 @@
+{
+  "_description": "Test matrix configurations for CUDA Python CI workflows. This file consolidates the test matrices that were previously hardcoded in the workflow files. The {gpu} placeholder in linux configurations is dynamically replaced based on architecture: l4 for amd64, a100 for arm64.",
+  "_sorted_by": "Please keep matrices sorted in ascending order by [PY_VER, CUDA_VER, LOCAL_CTK, GPU, DRIVER]",
+  "_notes": "DRIVER: 'earliest' does not work with CUDA 12.9.0 and LOCAL_CTK: 0 does not work with CUDA 12.0.1",
+  "linux": {
+    "pull-request": [
+      { "PY_VER": "3.9",  "CUDA_VER": "12.9.0", "LOCAL_CTK": "0", "GPU": "{gpu}", "DRIVER": "latest" },
+      { "PY_VER": "3.9",  "CUDA_VER": "13.0.0", "LOCAL_CTK": "1", "GPU": "{gpu}", "DRIVER": "latest" },
+      { "PY_VER": "3.10", "CUDA_VER": "12.9.0", "LOCAL_CTK": "1", "GPU": "{gpu}", "DRIVER": "latest" },
+      { "PY_VER": "3.10", "CUDA_VER": "13.0.0", "LOCAL_CTK": "0", "GPU": "{gpu}", "DRIVER": "latest" },
+      { "PY_VER": "3.11", "CUDA_VER": "12.9.0", "LOCAL_CTK": "0", "GPU": "{gpu}", "DRIVER": "latest" },
+      { "PY_VER": "3.11", "CUDA_VER": "13.0.0", "LOCAL_CTK": "1", "GPU": "{gpu}", "DRIVER": "latest" },
+      { "PY_VER": "3.12", "CUDA_VER": "12.9.0", "LOCAL_CTK": "1", "GPU": "{gpu}", "DRIVER": "latest" },
+      { "PY_VER": "3.12", "CUDA_VER": "13.0.0", "LOCAL_CTK": "0", "GPU": "{gpu}", "DRIVER": "latest" },
+      { "PY_VER": "3.13", "CUDA_VER": "12.9.0", "LOCAL_CTK": "0", "GPU": "{gpu}", "DRIVER": "latest" },
+      { "PY_VER": "3.13", "CUDA_VER": "13.0.0", "LOCAL_CTK": "1", "GPU": "{gpu}", "DRIVER": "latest" }
+    ],
+    "nightly": [
+      { "PY_VER": "3.9",  "CUDA_VER": "11.8.0", "LOCAL_CTK": "0", "GPU": "{gpu}", "DRIVER": "earliest" },
+      { "PY_VER": "3.9",  "CUDA_VER": "11.8.0", "LOCAL_CTK": "1", "GPU": "{gpu}", "DRIVER": "latest" },
+      { "PY_VER": "3.9",  "CUDA_VER": "12.0.1", "LOCAL_CTK": "1", "GPU": "{gpu}", "DRIVER": "latest" },
+      { "PY_VER": "3.9",  "CUDA_VER": "12.9.0", "LOCAL_CTK": "0", "GPU": "{gpu}", "DRIVER": "latest" },
+      { "PY_VER": "3.9",  "CUDA_VER": "12.9.0", "LOCAL_CTK": "1", "GPU": "{gpu}", "DRIVER": "latest" },
+      { "PY_VER": "3.10", "CUDA_VER": "11.8.0", "LOCAL_CTK": "0", "GPU": "{gpu}", "DRIVER": "earliest" },
+      { "PY_VER": "3.10", "CUDA_VER": "11.8.0", "LOCAL_CTK": "1", "GPU": "{gpu}", "DRIVER": "latest" },
+      { "PY_VER": "3.10", "CUDA_VER": "12.0.1", "LOCAL_CTK": "1", "GPU": "{gpu}", "DRIVER": "latest" },
+      { "PY_VER": "3.10", "CUDA_VER": "12.9.0", "LOCAL_CTK": "0", "GPU": "{gpu}", "DRIVER": "latest" },
+      { "PY_VER": "3.10", "CUDA_VER": "12.9.0", "LOCAL_CTK": "1", "GPU": "{gpu}", "DRIVER": "latest" },
+      { "PY_VER": "3.11", "CUDA_VER": "11.8.0", "LOCAL_CTK": "0", "GPU": "{gpu}", "DRIVER": "earliest" },
+      { "PY_VER": "3.11", "CUDA_VER": "11.8.0", "LOCAL_CTK": "1", "GPU": "{gpu}", "DRIVER": "latest" },
+      { "PY_VER": "3.11", "CUDA_VER": "12.0.1", "LOCAL_CTK": "1", "GPU": "{gpu}", "DRIVER": "latest" },
+      { "PY_VER": "3.11", "CUDA_VER": "12.9.0", "LOCAL_CTK": "0", "GPU": "{gpu}", "DRIVER": "latest" },
+      { "PY_VER": "3.11", "CUDA_VER": "12.9.0", "LOCAL_CTK": "1", "GPU": "{gpu}", "DRIVER": "latest" },
+      { "PY_VER": "3.12", "CUDA_VER": "11.8.0", "LOCAL_CTK": "0", "GPU": "{gpu}", "DRIVER": "earliest" },
+      { "PY_VER": "3.12", "CUDA_VER": "11.8.0", "LOCAL_CTK": "1", "GPU": "{gpu}", "DRIVER": "latest" },
+      { "PY_VER": "3.12", "CUDA_VER": "12.0.1", "LOCAL_CTK": "1", "GPU": "{gpu}", "DRIVER": "latest" },
+      { "PY_VER": "3.12", "CUDA_VER": "12.9.0", "LOCAL_CTK": "0", "GPU": "{gpu}", "DRIVER": "latest" },
+      { "PY_VER": "3.12", "CUDA_VER": "12.9.0", "LOCAL_CTK": "1", "GPU": "{gpu}", "DRIVER": "latest" },
+      { "PY_VER": "3.13", "CUDA_VER": "11.8.0", "LOCAL_CTK": "0", "GPU": "{gpu}", "DRIVER": "earliest" },
+      { "PY_VER": "3.13", "CUDA_VER": "11.8.0", "LOCAL_CTK": "1", "GPU": "{gpu}", "DRIVER": "latest" },
+      { "PY_VER": "3.13", "CUDA_VER": "12.0.1", "LOCAL_CTK": "1", "GPU": "{gpu}", "DRIVER": "latest" },
+      { "PY_VER": "3.13", "CUDA_VER": "12.9.0", "LOCAL_CTK": "0", "GPU": "{gpu}", "DRIVER": "latest" },
+      { "PY_VER": "3.13", "CUDA_VER": "12.9.0", "LOCAL_CTK": "1", "GPU": "{gpu}", "DRIVER": "latest" }
+    ],
+    "special_runners": {
+      "amd64": [
+        { "PY_VER": "3.13", "CUDA_VER": "13.0.0", "LOCAL_CTK": "1", "GPU": "H100", "DRIVER": "latest" }
+      ]
+    }
+  },
+  "windows": {
+    "pull-request": [
+      { "PY_VER": "3.12", "CUDA_VER": "12.9.0", "LOCAL_CTK": "0" },
+      { "PY_VER": "3.12", "CUDA_VER": "12.9.0", "LOCAL_CTK": "1" },
+      { "PY_VER": "3.13", "CUDA_VER": "13.0.0", "LOCAL_CTK": "0" },
+      { "PY_VER": "3.13", "CUDA_VER": "13.0.0", "LOCAL_CTK": "1" }
+    ],
+    "nightly": [
+      { "PY_VER": "3.12", "CUDA_VER": "11.8.0", "LOCAL_CTK": "0" },
+      { "PY_VER": "3.12", "CUDA_VER": "11.8.0", "LOCAL_CTK": "1" },
+      { "PY_VER": "3.12", "CUDA_VER": "12.9.0", "LOCAL_CTK": "0" },
+      { "PY_VER": "3.12", "CUDA_VER": "12.9.0", "LOCAL_CTK": "1" }
+    ]
+  }
+}
\ No newline at end of file

From 27232d6966d25294cca6cf4827b1276e78210800 Mon Sep 17 00:00:00 2001
From: "copilot-swe-agent[bot]" <198982749+Copilot@users.noreply.github.com>
Date: Fri, 22 Aug 2025 02:12:11 +0000
Subject: [PATCH 03/12] Hard-code all GPU and ARCH values in test-matrix.json
 with 6 fields per entry

Co-authored-by: leofang <5534781+leofang@users.noreply.github.com>
---
 .github/workflows/test-wheel-linux.yml   |  19 ++--
 .github/workflows/test-wheel-windows.yml |   8 +-
 ci/test-matrix.json                      | 127 +++++++++++++++--------
 3 files changed, 91 insertions(+), 63 deletions(-)

diff --git a/.github/workflows/test-wheel-linux.yml b/.github/workflows/test-wheel-linux.yml
index 9e99a22a79..60840749c5 100644
--- a/.github/workflows/test-wheel-linux.yml
+++ b/.github/workflows/test-wheel-linux.yml
@@ -49,29 +49,22 @@ jobs:
       - name: Compute Python Test Matrix
         id: compute-matrix
         run: |
-          # Set a default GPU based upon architecture.
-          gpu="l4"
-          if [[ "${ARCH}" == "arm64" ]]; then
-            gpu="a100"
-          fi
-
           # Use the nightly matrix for branch tests
           MATRIX_TYPE="${BUILD_TYPE}"
           if [[ "${MATRIX_TYPE}" == "branch" ]]; then
             MATRIX_TYPE="nightly"
           fi
 
-          # Read base matrix from JSON file and substitute GPU placeholder
-          TEST_MATRIX=$(jq --arg gpu "$gpu" --arg arch "$ARCH" '
-            .linux[env.MATRIX_TYPE] | 
-            map(. + {ARCH: $arch} | if .GPU == "{gpu}" then . + {GPU: $gpu} else . end)
+          # Read base matrix from JSON file for the specific architecture
+          TEST_MATRIX=$(jq --arg arch "$ARCH" --arg matrix_type "$MATRIX_TYPE" '
+            .linux[$matrix_type] | 
+            map(select(.ARCH == $arch))
           ' ci/test-matrix.json)
 
           # Add special runner for amd64 if applicable
           if [[ "${ARCH}" == "amd64" ]]; then
-            SPECIAL_RUNNERS=$(jq --arg arch "$ARCH" '
-              .linux.special_runners.amd64 | 
-              map(. + {ARCH: $arch})
+            SPECIAL_RUNNERS=$(jq '
+              .linux.special_runners.amd64
             ' ci/test-matrix.json)
             TEST_MATRIX=$(jq --argjson special "$SPECIAL_RUNNERS" '. + $special' <<< "$TEST_MATRIX")
           fi
diff --git a/.github/workflows/test-wheel-windows.yml b/.github/workflows/test-wheel-windows.yml
index cdc9dfa6b8..632c8752b1 100644
--- a/.github/workflows/test-wheel-windows.yml
+++ b/.github/workflows/test-wheel-windows.yml
@@ -52,10 +52,10 @@ jobs:
             MATRIX_TYPE="nightly"
           fi
 
-          # Read base matrix from JSON file and add ARCH field
-          TEST_MATRIX=$(jq --arg arch "$ARCH" '
-            .windows[env.MATRIX_TYPE] | 
-            map(. + {ARCH: $arch})
+          # Read base matrix from JSON file for the specific architecture
+          TEST_MATRIX=$(jq --arg arch "$ARCH" --arg matrix_type "$MATRIX_TYPE" '
+            .windows[$matrix_type] | 
+            map(select(.ARCH == $arch))
           ' ci/test-matrix.json)
 
           MATRIX="$(
diff --git a/ci/test-matrix.json b/ci/test-matrix.json
index 2491e18f37..568767668b 100644
--- a/ci/test-matrix.json
+++ b/ci/test-matrix.json
@@ -1,65 +1,100 @@
 {
-  "_description": "Test matrix configurations for CUDA Python CI workflows. This file consolidates the test matrices that were previously hardcoded in the workflow files. The {gpu} placeholder in linux configurations is dynamically replaced based on architecture: l4 for amd64, a100 for arm64.",
-  "_sorted_by": "Please keep matrices sorted in ascending order by [PY_VER, CUDA_VER, LOCAL_CTK, GPU, DRIVER]",
+  "_description": "Test matrix configurations for CUDA Python CI workflows. This file consolidates the test matrices that were previously hardcoded in the workflow files. All GPU and ARCH values are hard-coded for each architecture: l4 GPU for amd64, a100 GPU for arm64.",
+  "_sorted_by": "Please keep matrices sorted in ascending order by [ARCH, PY_VER, CUDA_VER, LOCAL_CTK, GPU, DRIVER]",
   "_notes": "DRIVER: 'earliest' does not work with CUDA 12.9.0 and LOCAL_CTK: 0 does not work with CUDA 12.0.1",
   "linux": {
     "pull-request": [
-      { "PY_VER": "3.9",  "CUDA_VER": "12.9.0", "LOCAL_CTK": "0", "GPU": "{gpu}", "DRIVER": "latest" },
-      { "PY_VER": "3.9",  "CUDA_VER": "13.0.0", "LOCAL_CTK": "1", "GPU": "{gpu}", "DRIVER": "latest" },
-      { "PY_VER": "3.10", "CUDA_VER": "12.9.0", "LOCAL_CTK": "1", "GPU": "{gpu}", "DRIVER": "latest" },
-      { "PY_VER": "3.10", "CUDA_VER": "13.0.0", "LOCAL_CTK": "0", "GPU": "{gpu}", "DRIVER": "latest" },
-      { "PY_VER": "3.11", "CUDA_VER": "12.9.0", "LOCAL_CTK": "0", "GPU": "{gpu}", "DRIVER": "latest" },
-      { "PY_VER": "3.11", "CUDA_VER": "13.0.0", "LOCAL_CTK": "1", "GPU": "{gpu}", "DRIVER": "latest" },
-      { "PY_VER": "3.12", "CUDA_VER": "12.9.0", "LOCAL_CTK": "1", "GPU": "{gpu}", "DRIVER": "latest" },
-      { "PY_VER": "3.12", "CUDA_VER": "13.0.0", "LOCAL_CTK": "0", "GPU": "{gpu}", "DRIVER": "latest" },
-      { "PY_VER": "3.13", "CUDA_VER": "12.9.0", "LOCAL_CTK": "0", "GPU": "{gpu}", "DRIVER": "latest" },
-      { "PY_VER": "3.13", "CUDA_VER": "13.0.0", "LOCAL_CTK": "1", "GPU": "{gpu}", "DRIVER": "latest" }
+      { "ARCH": "amd64", "PY_VER": "3.9",  "CUDA_VER": "12.9.0", "LOCAL_CTK": "0", "GPU": "l4", "DRIVER": "latest" },
+      { "ARCH": "amd64", "PY_VER": "3.9",  "CUDA_VER": "13.0.0", "LOCAL_CTK": "1", "GPU": "l4", "DRIVER": "latest" },
+      { "ARCH": "amd64", "PY_VER": "3.10", "CUDA_VER": "12.9.0", "LOCAL_CTK": "1", "GPU": "l4", "DRIVER": "latest" },
+      { "ARCH": "amd64", "PY_VER": "3.10", "CUDA_VER": "13.0.0", "LOCAL_CTK": "0", "GPU": "l4", "DRIVER": "latest" },
+      { "ARCH": "amd64", "PY_VER": "3.11", "CUDA_VER": "12.9.0", "LOCAL_CTK": "0", "GPU": "l4", "DRIVER": "latest" },
+      { "ARCH": "amd64", "PY_VER": "3.11", "CUDA_VER": "13.0.0", "LOCAL_CTK": "1", "GPU": "l4", "DRIVER": "latest" },
+      { "ARCH": "amd64", "PY_VER": "3.12", "CUDA_VER": "12.9.0", "LOCAL_CTK": "1", "GPU": "l4", "DRIVER": "latest" },
+      { "ARCH": "amd64", "PY_VER": "3.12", "CUDA_VER": "13.0.0", "LOCAL_CTK": "0", "GPU": "l4", "DRIVER": "latest" },
+      { "ARCH": "amd64", "PY_VER": "3.13", "CUDA_VER": "12.9.0", "LOCAL_CTK": "0", "GPU": "l4", "DRIVER": "latest" },
+      { "ARCH": "amd64", "PY_VER": "3.13", "CUDA_VER": "13.0.0", "LOCAL_CTK": "1", "GPU": "l4", "DRIVER": "latest" },
+      { "ARCH": "arm64", "PY_VER": "3.9",  "CUDA_VER": "12.9.0", "LOCAL_CTK": "0", "GPU": "a100", "DRIVER": "latest" },
+      { "ARCH": "arm64", "PY_VER": "3.9",  "CUDA_VER": "13.0.0", "LOCAL_CTK": "1", "GPU": "a100", "DRIVER": "latest" },
+      { "ARCH": "arm64", "PY_VER": "3.10", "CUDA_VER": "12.9.0", "LOCAL_CTK": "1", "GPU": "a100", "DRIVER": "latest" },
+      { "ARCH": "arm64", "PY_VER": "3.10", "CUDA_VER": "13.0.0", "LOCAL_CTK": "0", "GPU": "a100", "DRIVER": "latest" },
+      { "ARCH": "arm64", "PY_VER": "3.11", "CUDA_VER": "12.9.0", "LOCAL_CTK": "0", "GPU": "a100", "DRIVER": "latest" },
+      { "ARCH": "arm64", "PY_VER": "3.11", "CUDA_VER": "13.0.0", "LOCAL_CTK": "1", "GPU": "a100", "DRIVER": "latest" },
+      { "ARCH": "arm64", "PY_VER": "3.12", "CUDA_VER": "12.9.0", "LOCAL_CTK": "1", "GPU": "a100", "DRIVER": "latest" },
+      { "ARCH": "arm64", "PY_VER": "3.12", "CUDA_VER": "13.0.0", "LOCAL_CTK": "0", "GPU": "a100", "DRIVER": "latest" },
+      { "ARCH": "arm64", "PY_VER": "3.13", "CUDA_VER": "12.9.0", "LOCAL_CTK": "0", "GPU": "a100", "DRIVER": "latest" },
+      { "ARCH": "arm64", "PY_VER": "3.13", "CUDA_VER": "13.0.0", "LOCAL_CTK": "1", "GPU": "a100", "DRIVER": "latest" }
     ],
     "nightly": [
-      { "PY_VER": "3.9",  "CUDA_VER": "11.8.0", "LOCAL_CTK": "0", "GPU": "{gpu}", "DRIVER": "earliest" },
-      { "PY_VER": "3.9",  "CUDA_VER": "11.8.0", "LOCAL_CTK": "1", "GPU": "{gpu}", "DRIVER": "latest" },
-      { "PY_VER": "3.9",  "CUDA_VER": "12.0.1", "LOCAL_CTK": "1", "GPU": "{gpu}", "DRIVER": "latest" },
-      { "PY_VER": "3.9",  "CUDA_VER": "12.9.0", "LOCAL_CTK": "0", "GPU": "{gpu}", "DRIVER": "latest" },
-      { "PY_VER": "3.9",  "CUDA_VER": "12.9.0", "LOCAL_CTK": "1", "GPU": "{gpu}", "DRIVER": "latest" },
-      { "PY_VER": "3.10", "CUDA_VER": "11.8.0", "LOCAL_CTK": "0", "GPU": "{gpu}", "DRIVER": "earliest" },
-      { "PY_VER": "3.10", "CUDA_VER": "11.8.0", "LOCAL_CTK": "1", "GPU": "{gpu}", "DRIVER": "latest" },
-      { "PY_VER": "3.10", "CUDA_VER": "12.0.1", "LOCAL_CTK": "1", "GPU": "{gpu}", "DRIVER": "latest" },
-      { "PY_VER": "3.10", "CUDA_VER": "12.9.0", "LOCAL_CTK": "0", "GPU": "{gpu}", "DRIVER": "latest" },
-      { "PY_VER": "3.10", "CUDA_VER": "12.9.0", "LOCAL_CTK": "1", "GPU": "{gpu}", "DRIVER": "latest" },
-      { "PY_VER": "3.11", "CUDA_VER": "11.8.0", "LOCAL_CTK": "0", "GPU": "{gpu}", "DRIVER": "earliest" },
-      { "PY_VER": "3.11", "CUDA_VER": "11.8.0", "LOCAL_CTK": "1", "GPU": "{gpu}", "DRIVER": "latest" },
-      { "PY_VER": "3.11", "CUDA_VER": "12.0.1", "LOCAL_CTK": "1", "GPU": "{gpu}", "DRIVER": "latest" },
-      { "PY_VER": "3.11", "CUDA_VER": "12.9.0", "LOCAL_CTK": "0", "GPU": "{gpu}", "DRIVER": "latest" },
-      { "PY_VER": "3.11", "CUDA_VER": "12.9.0", "LOCAL_CTK": "1", "GPU": "{gpu}", "DRIVER": "latest" },
-      { "PY_VER": "3.12", "CUDA_VER": "11.8.0", "LOCAL_CTK": "0", "GPU": "{gpu}", "DRIVER": "earliest" },
-      { "PY_VER": "3.12", "CUDA_VER": "11.8.0", "LOCAL_CTK": "1", "GPU": "{gpu}", "DRIVER": "latest" },
-      { "PY_VER": "3.12", "CUDA_VER": "12.0.1", "LOCAL_CTK": "1", "GPU": "{gpu}", "DRIVER": "latest" },
-      { "PY_VER": "3.12", "CUDA_VER": "12.9.0", "LOCAL_CTK": "0", "GPU": "{gpu}", "DRIVER": "latest" },
-      { "PY_VER": "3.12", "CUDA_VER": "12.9.0", "LOCAL_CTK": "1", "GPU": "{gpu}", "DRIVER": "latest" },
-      { "PY_VER": "3.13", "CUDA_VER": "11.8.0", "LOCAL_CTK": "0", "GPU": "{gpu}", "DRIVER": "earliest" },
-      { "PY_VER": "3.13", "CUDA_VER": "11.8.0", "LOCAL_CTK": "1", "GPU": "{gpu}", "DRIVER": "latest" },
-      { "PY_VER": "3.13", "CUDA_VER": "12.0.1", "LOCAL_CTK": "1", "GPU": "{gpu}", "DRIVER": "latest" },
-      { "PY_VER": "3.13", "CUDA_VER": "12.9.0", "LOCAL_CTK": "0", "GPU": "{gpu}", "DRIVER": "latest" },
-      { "PY_VER": "3.13", "CUDA_VER": "12.9.0", "LOCAL_CTK": "1", "GPU": "{gpu}", "DRIVER": "latest" }
+      { "ARCH": "amd64", "PY_VER": "3.9",  "CUDA_VER": "11.8.0", "LOCAL_CTK": "0", "GPU": "l4", "DRIVER": "earliest" },
+      { "ARCH": "amd64", "PY_VER": "3.9",  "CUDA_VER": "11.8.0", "LOCAL_CTK": "1", "GPU": "l4", "DRIVER": "latest" },
+      { "ARCH": "amd64", "PY_VER": "3.9",  "CUDA_VER": "12.0.1", "LOCAL_CTK": "1", "GPU": "l4", "DRIVER": "latest" },
+      { "ARCH": "amd64", "PY_VER": "3.9",  "CUDA_VER": "12.9.0", "LOCAL_CTK": "0", "GPU": "l4", "DRIVER": "latest" },
+      { "ARCH": "amd64", "PY_VER": "3.9",  "CUDA_VER": "12.9.0", "LOCAL_CTK": "1", "GPU": "l4", "DRIVER": "latest" },
+      { "ARCH": "amd64", "PY_VER": "3.10", "CUDA_VER": "11.8.0", "LOCAL_CTK": "0", "GPU": "l4", "DRIVER": "earliest" },
+      { "ARCH": "amd64", "PY_VER": "3.10", "CUDA_VER": "11.8.0", "LOCAL_CTK": "1", "GPU": "l4", "DRIVER": "latest" },
+      { "ARCH": "amd64", "PY_VER": "3.10", "CUDA_VER": "12.0.1", "LOCAL_CTK": "1", "GPU": "l4", "DRIVER": "latest" },
+      { "ARCH": "amd64", "PY_VER": "3.10", "CUDA_VER": "12.9.0", "LOCAL_CTK": "0", "GPU": "l4", "DRIVER": "latest" },
+      { "ARCH": "amd64", "PY_VER": "3.10", "CUDA_VER": "12.9.0", "LOCAL_CTK": "1", "GPU": "l4", "DRIVER": "latest" },
+      { "ARCH": "amd64", "PY_VER": "3.11", "CUDA_VER": "11.8.0", "LOCAL_CTK": "0", "GPU": "l4", "DRIVER": "earliest" },
+      { "ARCH": "amd64", "PY_VER": "3.11", "CUDA_VER": "11.8.0", "LOCAL_CTK": "1", "GPU": "l4", "DRIVER": "latest" },
+      { "ARCH": "amd64", "PY_VER": "3.11", "CUDA_VER": "12.0.1", "LOCAL_CTK": "1", "GPU": "l4", "DRIVER": "latest" },
+      { "ARCH": "amd64", "PY_VER": "3.11", "CUDA_VER": "12.9.0", "LOCAL_CTK": "0", "GPU": "l4", "DRIVER": "latest" },
+      { "ARCH": "amd64", "PY_VER": "3.11", "CUDA_VER": "12.9.0", "LOCAL_CTK": "1", "GPU": "l4", "DRIVER": "latest" },
+      { "ARCH": "amd64", "PY_VER": "3.12", "CUDA_VER": "11.8.0", "LOCAL_CTK": "0", "GPU": "l4", "DRIVER": "earliest" },
+      { "ARCH": "amd64", "PY_VER": "3.12", "CUDA_VER": "11.8.0", "LOCAL_CTK": "1", "GPU": "l4", "DRIVER": "latest" },
+      { "ARCH": "amd64", "PY_VER": "3.12", "CUDA_VER": "12.0.1", "LOCAL_CTK": "1", "GPU": "l4", "DRIVER": "latest" },
+      { "ARCH": "amd64", "PY_VER": "3.12", "CUDA_VER": "12.9.0", "LOCAL_CTK": "0", "GPU": "l4", "DRIVER": "latest" },
+      { "ARCH": "amd64", "PY_VER": "3.12", "CUDA_VER": "12.9.0", "LOCAL_CTK": "1", "GPU": "l4", "DRIVER": "latest" },
+      { "ARCH": "amd64", "PY_VER": "3.13", "CUDA_VER": "11.8.0", "LOCAL_CTK": "0", "GPU": "l4", "DRIVER": "earliest" },
+      { "ARCH": "amd64", "PY_VER": "3.13", "CUDA_VER": "11.8.0", "LOCAL_CTK": "1", "GPU": "l4", "DRIVER": "latest" },
+      { "ARCH": "amd64", "PY_VER": "3.13", "CUDA_VER": "12.0.1", "LOCAL_CTK": "1", "GPU": "l4", "DRIVER": "latest" },
+      { "ARCH": "amd64", "PY_VER": "3.13", "CUDA_VER": "12.9.0", "LOCAL_CTK": "0", "GPU": "l4", "DRIVER": "latest" },
+      { "ARCH": "amd64", "PY_VER": "3.13", "CUDA_VER": "12.9.0", "LOCAL_CTK": "1", "GPU": "l4", "DRIVER": "latest" },
+      { "ARCH": "arm64", "PY_VER": "3.9",  "CUDA_VER": "11.8.0", "LOCAL_CTK": "0", "GPU": "a100", "DRIVER": "earliest" },
+      { "ARCH": "arm64", "PY_VER": "3.9",  "CUDA_VER": "11.8.0", "LOCAL_CTK": "1", "GPU": "a100", "DRIVER": "latest" },
+      { "ARCH": "arm64", "PY_VER": "3.9",  "CUDA_VER": "12.0.1", "LOCAL_CTK": "1", "GPU": "a100", "DRIVER": "latest" },
+      { "ARCH": "arm64", "PY_VER": "3.9",  "CUDA_VER": "12.9.0", "LOCAL_CTK": "0", "GPU": "a100", "DRIVER": "latest" },
+      { "ARCH": "arm64", "PY_VER": "3.9",  "CUDA_VER": "12.9.0", "LOCAL_CTK": "1", "GPU": "a100", "DRIVER": "latest" },
+      { "ARCH": "arm64", "PY_VER": "3.10", "CUDA_VER": "11.8.0", "LOCAL_CTK": "0", "GPU": "a100", "DRIVER": "earliest" },
+      { "ARCH": "arm64", "PY_VER": "3.10", "CUDA_VER": "11.8.0", "LOCAL_CTK": "1", "GPU": "a100", "DRIVER": "latest" },
+      { "ARCH": "arm64", "PY_VER": "3.10", "CUDA_VER": "12.0.1", "LOCAL_CTK": "1", "GPU": "a100", "DRIVER": "latest" },
+      { "ARCH": "arm64", "PY_VER": "3.10", "CUDA_VER": "12.9.0", "LOCAL_CTK": "0", "GPU": "a100", "DRIVER": "latest" },
+      { "ARCH": "arm64", "PY_VER": "3.10", "CUDA_VER": "12.9.0", "LOCAL_CTK": "1", "GPU": "a100", "DRIVER": "latest" },
+      { "ARCH": "arm64", "PY_VER": "3.11", "CUDA_VER": "11.8.0", "LOCAL_CTK": "0", "GPU": "a100", "DRIVER": "earliest" },
+      { "ARCH": "arm64", "PY_VER": "3.11", "CUDA_VER": "11.8.0", "LOCAL_CTK": "1", "GPU": "a100", "DRIVER": "latest" },
+      { "ARCH": "arm64", "PY_VER": "3.11", "CUDA_VER": "12.0.1", "LOCAL_CTK": "1", "GPU": "a100", "DRIVER": "latest" },
+      { "ARCH": "arm64", "PY_VER": "3.11", "CUDA_VER": "12.9.0", "LOCAL_CTK": "0", "GPU": "a100", "DRIVER": "latest" },
+      { "ARCH": "arm64", "PY_VER": "3.11", "CUDA_VER": "12.9.0", "LOCAL_CTK": "1", "GPU": "a100", "DRIVER": "latest" },
+      { "ARCH": "arm64", "PY_VER": "3.12", "CUDA_VER": "11.8.0", "LOCAL_CTK": "0", "GPU": "a100", "DRIVER": "earliest" },
+      { "ARCH": "arm64", "PY_VER": "3.12", "CUDA_VER": "11.8.0", "LOCAL_CTK": "1", "GPU": "a100", "DRIVER": "latest" },
+      { "ARCH": "arm64", "PY_VER": "3.12", "CUDA_VER": "12.0.1", "LOCAL_CTK": "1", "GPU": "a100", "DRIVER": "latest" },
+      { "ARCH": "arm64", "PY_VER": "3.12", "CUDA_VER": "12.9.0", "LOCAL_CTK": "0", "GPU": "a100", "DRIVER": "latest" },
+      { "ARCH": "arm64", "PY_VER": "3.12", "CUDA_VER": "12.9.0", "LOCAL_CTK": "1", "GPU": "a100", "DRIVER": "latest" },
+      { "ARCH": "arm64", "PY_VER": "3.13", "CUDA_VER": "11.8.0", "LOCAL_CTK": "0", "GPU": "a100", "DRIVER": "earliest" },
+      { "ARCH": "arm64", "PY_VER": "3.13", "CUDA_VER": "11.8.0", "LOCAL_CTK": "1", "GPU": "a100", "DRIVER": "latest" },
+      { "ARCH": "arm64", "PY_VER": "3.13", "CUDA_VER": "12.0.1", "LOCAL_CTK": "1", "GPU": "a100", "DRIVER": "latest" },
+      { "ARCH": "arm64", "PY_VER": "3.13", "CUDA_VER": "12.9.0", "LOCAL_CTK": "0", "GPU": "a100", "DRIVER": "latest" },
+      { "ARCH": "arm64", "PY_VER": "3.13", "CUDA_VER": "12.9.0", "LOCAL_CTK": "1", "GPU": "a100", "DRIVER": "latest" }
     ],
     "special_runners": {
       "amd64": [
-        { "PY_VER": "3.13", "CUDA_VER": "13.0.0", "LOCAL_CTK": "1", "GPU": "H100", "DRIVER": "latest" }
+        { "ARCH": "amd64", "PY_VER": "3.13", "CUDA_VER": "13.0.0", "LOCAL_CTK": "1", "GPU": "H100", "DRIVER": "latest" }
       ]
     }
   },
   "windows": {
     "pull-request": [
-      { "PY_VER": "3.12", "CUDA_VER": "12.9.0", "LOCAL_CTK": "0" },
-      { "PY_VER": "3.12", "CUDA_VER": "12.9.0", "LOCAL_CTK": "1" },
-      { "PY_VER": "3.13", "CUDA_VER": "13.0.0", "LOCAL_CTK": "0" },
-      { "PY_VER": "3.13", "CUDA_VER": "13.0.0", "LOCAL_CTK": "1" }
+      { "ARCH": "amd64", "PY_VER": "3.12", "CUDA_VER": "12.9.0", "LOCAL_CTK": "0", "GPU": "gpu", "DRIVER": "latest" },
+      { "ARCH": "amd64", "PY_VER": "3.12", "CUDA_VER": "12.9.0", "LOCAL_CTK": "1", "GPU": "gpu", "DRIVER": "latest" },
+      { "ARCH": "amd64", "PY_VER": "3.13", "CUDA_VER": "13.0.0", "LOCAL_CTK": "0", "GPU": "gpu", "DRIVER": "latest" },
+      { "ARCH": "amd64", "PY_VER": "3.13", "CUDA_VER": "13.0.0", "LOCAL_CTK": "1", "GPU": "gpu", "DRIVER": "latest" }
     ],
     "nightly": [
-      { "PY_VER": "3.12", "CUDA_VER": "11.8.0", "LOCAL_CTK": "0" },
-      { "PY_VER": "3.12", "CUDA_VER": "11.8.0", "LOCAL_CTK": "1" },
-      { "PY_VER": "3.12", "CUDA_VER": "12.9.0", "LOCAL_CTK": "0" },
-      { "PY_VER": "3.12", "CUDA_VER": "12.9.0", "LOCAL_CTK": "1" }
+      { "ARCH": "amd64", "PY_VER": "3.12", "CUDA_VER": "11.8.0", "LOCAL_CTK": "0", "GPU": "gpu", "DRIVER": "latest" },
+      { "ARCH": "amd64", "PY_VER": "3.12", "CUDA_VER": "11.8.0", "LOCAL_CTK": "1", "GPU": "gpu", "DRIVER": "latest" },
+      { "ARCH": "amd64", "PY_VER": "3.12", "CUDA_VER": "12.9.0", "LOCAL_CTK": "0", "GPU": "gpu", "DRIVER": "latest" },
+      { "ARCH": "amd64", "PY_VER": "3.12", "CUDA_VER": "12.9.0", "LOCAL_CTK": "1", "GPU": "gpu", "DRIVER": "latest" }
     ]
   }
 }
\ No newline at end of file

From 111875a0a9fb702d4165cbd3f4299c383b4ac197 Mon Sep 17 00:00:00 2001
From: "copilot-swe-agent[bot]" <198982749+Copilot@users.noreply.github.com>
Date: Sun, 24 Aug 2025 03:54:45 +0000
Subject: [PATCH 04/12] Update Windows test matrix with a100 GPU and latest-1
 driver, configure self-hosted runners

Co-authored-by: leofang <5534781+leofang@users.noreply.github.com>
---
 .github/workflows/test-wheel-windows.yml |  4 ++--
 ci/test-matrix.json                      | 16 ++++++++--------
 2 files changed, 10 insertions(+), 10 deletions(-)

diff --git a/.github/workflows/test-wheel-windows.yml b/.github/workflows/test-wheel-windows.yml
index 632c8752b1..ebd1a72ef0 100644
--- a/.github/workflows/test-wheel-windows.yml
+++ b/.github/workflows/test-wheel-windows.yml
@@ -65,14 +65,14 @@ jobs:
           echo "MATRIX=${MATRIX}" | tee --append "${GITHUB_OUTPUT}"
 
   test:
-    name: py${{ matrix.PY_VER }}, ${{ matrix.CUDA_VER }}, ${{ (matrix.LOCAL_CTK == '1' && 'local') || 'wheels' }}
+    name: py${{ matrix.PY_VER }}, ${{ matrix.CUDA_VER }}, ${{ (matrix.LOCAL_CTK == '1' && 'local') || 'wheels' }}, GPU ${{ matrix.GPU }}
     # The build stage could fail but we want the CI to keep moving.
     needs: compute-matrix
     strategy:
       fail-fast: false
       matrix: ${{ fromJSON(needs.compute-matrix.outputs.MATRIX) }}
     if: ${{ github.repository_owner == 'nvidia' && !cancelled() }}
-    runs-on: 'cuda-python-windows-gpu-github'
+    runs-on: "windows-${{ matrix.ARCH }}-gpu-${{ matrix.GPU }}-${{ matrix.DRIVER }}-1"
     steps:
       - name: Checkout ${{ github.event.repository.name }}
         uses: actions/checkout@08c6903cd8c0fde910a37f88322edcfb5dd907a8  # v5.0.0
diff --git a/ci/test-matrix.json b/ci/test-matrix.json
index 568767668b..3c302b93d7 100644
--- a/ci/test-matrix.json
+++ b/ci/test-matrix.json
@@ -85,16 +85,16 @@
   },
   "windows": {
     "pull-request": [
-      { "ARCH": "amd64", "PY_VER": "3.12", "CUDA_VER": "12.9.0", "LOCAL_CTK": "0", "GPU": "gpu", "DRIVER": "latest" },
-      { "ARCH": "amd64", "PY_VER": "3.12", "CUDA_VER": "12.9.0", "LOCAL_CTK": "1", "GPU": "gpu", "DRIVER": "latest" },
-      { "ARCH": "amd64", "PY_VER": "3.13", "CUDA_VER": "13.0.0", "LOCAL_CTK": "0", "GPU": "gpu", "DRIVER": "latest" },
-      { "ARCH": "amd64", "PY_VER": "3.13", "CUDA_VER": "13.0.0", "LOCAL_CTK": "1", "GPU": "gpu", "DRIVER": "latest" }
+      { "ARCH": "amd64", "PY_VER": "3.12", "CUDA_VER": "12.9.0", "LOCAL_CTK": "0", "GPU": "a100", "DRIVER": "latest-1" },
+      { "ARCH": "amd64", "PY_VER": "3.12", "CUDA_VER": "12.9.0", "LOCAL_CTK": "1", "GPU": "a100", "DRIVER": "latest-1" },
+      { "ARCH": "amd64", "PY_VER": "3.13", "CUDA_VER": "13.0.0", "LOCAL_CTK": "0", "GPU": "a100", "DRIVER": "latest-1" },
+      { "ARCH": "amd64", "PY_VER": "3.13", "CUDA_VER": "13.0.0", "LOCAL_CTK": "1", "GPU": "a100", "DRIVER": "latest-1" }
     ],
     "nightly": [
-      { "ARCH": "amd64", "PY_VER": "3.12", "CUDA_VER": "11.8.0", "LOCAL_CTK": "0", "GPU": "gpu", "DRIVER": "latest" },
-      { "ARCH": "amd64", "PY_VER": "3.12", "CUDA_VER": "11.8.0", "LOCAL_CTK": "1", "GPU": "gpu", "DRIVER": "latest" },
-      { "ARCH": "amd64", "PY_VER": "3.12", "CUDA_VER": "12.9.0", "LOCAL_CTK": "0", "GPU": "gpu", "DRIVER": "latest" },
-      { "ARCH": "amd64", "PY_VER": "3.12", "CUDA_VER": "12.9.0", "LOCAL_CTK": "1", "GPU": "gpu", "DRIVER": "latest" }
+      { "ARCH": "amd64", "PY_VER": "3.12", "CUDA_VER": "11.8.0", "LOCAL_CTK": "0", "GPU": "a100", "DRIVER": "latest-1" },
+      { "ARCH": "amd64", "PY_VER": "3.12", "CUDA_VER": "11.8.0", "LOCAL_CTK": "1", "GPU": "a100", "DRIVER": "latest-1" },
+      { "ARCH": "amd64", "PY_VER": "3.12", "CUDA_VER": "12.9.0", "LOCAL_CTK": "0", "GPU": "a100", "DRIVER": "latest-1" },
+      { "ARCH": "amd64", "PY_VER": "3.12", "CUDA_VER": "12.9.0", "LOCAL_CTK": "1", "GPU": "a100", "DRIVER": "latest-1" }
     ]
   }
 }
\ No newline at end of file

From eed0b71957abb6d22a4fbebb96b1cb1141e99ec3 Mon Sep 17 00:00:00 2001
From: Leo Fang <leof@nvidia.com>
Date: Sat, 23 Aug 2025 23:58:35 -0400
Subject: [PATCH 05/12] fix

---
 .github/workflows/test-wheel-windows.yml | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/.github/workflows/test-wheel-windows.yml b/.github/workflows/test-wheel-windows.yml
index ebd1a72ef0..5b52279fe5 100644
--- a/.github/workflows/test-wheel-windows.yml
+++ b/.github/workflows/test-wheel-windows.yml
@@ -72,7 +72,7 @@ jobs:
       fail-fast: false
       matrix: ${{ fromJSON(needs.compute-matrix.outputs.MATRIX) }}
     if: ${{ github.repository_owner == 'nvidia' && !cancelled() }}
-    runs-on: "windows-${{ matrix.ARCH }}-gpu-${{ matrix.GPU }}-${{ matrix.DRIVER }}-1"
+    runs-on: "windows-${{ matrix.ARCH }}-gpu-${{ matrix.GPU }}-${{ matrix.DRIVER }}"
     steps:
       - name: Checkout ${{ github.event.repository.name }}
         uses: actions/checkout@08c6903cd8c0fde910a37f88322edcfb5dd907a8  # v5.0.0

From 69875a3fbab8efc0c774d279819a91b964222258 Mon Sep 17 00:00:00 2001
From: "copilot-swe-agent[bot]" <198982749+Copilot@users.noreply.github.com>
Date: Sun, 24 Aug 2025 04:05:43 +0000
Subject: [PATCH 06/12] Revert eed0b71 and change Windows DRIVER from latest-1
 to latest

Co-authored-by: leofang <5534781+leofang@users.noreply.github.com>
---
 .github/workflows/test-wheel-windows.yml |  2 +-
 ci/test-matrix.json                      | 16 ++++++++--------
 2 files changed, 9 insertions(+), 9 deletions(-)

diff --git a/.github/workflows/test-wheel-windows.yml b/.github/workflows/test-wheel-windows.yml
index 5b52279fe5..ebd1a72ef0 100644
--- a/.github/workflows/test-wheel-windows.yml
+++ b/.github/workflows/test-wheel-windows.yml
@@ -72,7 +72,7 @@ jobs:
       fail-fast: false
       matrix: ${{ fromJSON(needs.compute-matrix.outputs.MATRIX) }}
     if: ${{ github.repository_owner == 'nvidia' && !cancelled() }}
-    runs-on: "windows-${{ matrix.ARCH }}-gpu-${{ matrix.GPU }}-${{ matrix.DRIVER }}"
+    runs-on: "windows-${{ matrix.ARCH }}-gpu-${{ matrix.GPU }}-${{ matrix.DRIVER }}-1"
     steps:
       - name: Checkout ${{ github.event.repository.name }}
         uses: actions/checkout@08c6903cd8c0fde910a37f88322edcfb5dd907a8  # v5.0.0
diff --git a/ci/test-matrix.json b/ci/test-matrix.json
index 3c302b93d7..5eb3fbe9fd 100644
--- a/ci/test-matrix.json
+++ b/ci/test-matrix.json
@@ -85,16 +85,16 @@
   },
   "windows": {
     "pull-request": [
-      { "ARCH": "amd64", "PY_VER": "3.12", "CUDA_VER": "12.9.0", "LOCAL_CTK": "0", "GPU": "a100", "DRIVER": "latest-1" },
-      { "ARCH": "amd64", "PY_VER": "3.12", "CUDA_VER": "12.9.0", "LOCAL_CTK": "1", "GPU": "a100", "DRIVER": "latest-1" },
-      { "ARCH": "amd64", "PY_VER": "3.13", "CUDA_VER": "13.0.0", "LOCAL_CTK": "0", "GPU": "a100", "DRIVER": "latest-1" },
-      { "ARCH": "amd64", "PY_VER": "3.13", "CUDA_VER": "13.0.0", "LOCAL_CTK": "1", "GPU": "a100", "DRIVER": "latest-1" }
+      { "ARCH": "amd64", "PY_VER": "3.12", "CUDA_VER": "12.9.0", "LOCAL_CTK": "0", "GPU": "a100", "DRIVER": "latest" },
+      { "ARCH": "amd64", "PY_VER": "3.12", "CUDA_VER": "12.9.0", "LOCAL_CTK": "1", "GPU": "a100", "DRIVER": "latest" },
+      { "ARCH": "amd64", "PY_VER": "3.13", "CUDA_VER": "13.0.0", "LOCAL_CTK": "0", "GPU": "a100", "DRIVER": "latest" },
+      { "ARCH": "amd64", "PY_VER": "3.13", "CUDA_VER": "13.0.0", "LOCAL_CTK": "1", "GPU": "a100", "DRIVER": "latest" }
     ],
     "nightly": [
-      { "ARCH": "amd64", "PY_VER": "3.12", "CUDA_VER": "11.8.0", "LOCAL_CTK": "0", "GPU": "a100", "DRIVER": "latest-1" },
-      { "ARCH": "amd64", "PY_VER": "3.12", "CUDA_VER": "11.8.0", "LOCAL_CTK": "1", "GPU": "a100", "DRIVER": "latest-1" },
-      { "ARCH": "amd64", "PY_VER": "3.12", "CUDA_VER": "12.9.0", "LOCAL_CTK": "0", "GPU": "a100", "DRIVER": "latest-1" },
-      { "ARCH": "amd64", "PY_VER": "3.12", "CUDA_VER": "12.9.0", "LOCAL_CTK": "1", "GPU": "a100", "DRIVER": "latest-1" }
+      { "ARCH": "amd64", "PY_VER": "3.12", "CUDA_VER": "11.8.0", "LOCAL_CTK": "0", "GPU": "a100", "DRIVER": "latest" },
+      { "ARCH": "amd64", "PY_VER": "3.12", "CUDA_VER": "11.8.0", "LOCAL_CTK": "1", "GPU": "a100", "DRIVER": "latest" },
+      { "ARCH": "amd64", "PY_VER": "3.12", "CUDA_VER": "12.9.0", "LOCAL_CTK": "0", "GPU": "a100", "DRIVER": "latest" },
+      { "ARCH": "amd64", "PY_VER": "3.12", "CUDA_VER": "12.9.0", "LOCAL_CTK": "1", "GPU": "a100", "DRIVER": "latest" }
     ]
   }
 }
\ No newline at end of file

From ae93e2bf70b0b26b893a82dcdd26407cf6300397 Mon Sep 17 00:00:00 2001
From: "copilot-swe-agent[bot]" <198982749+Copilot@users.noreply.github.com>
Date: Sun, 24 Aug 2025 04:50:18 +0000
Subject: [PATCH 07/12] Add proxy cache setup to Windows workflow for
 self-hosted runners

Co-authored-by: leofang <5534781+leofang@users.noreply.github.com>
---
 .github/workflows/test-wheel-windows.yml | 4 +++-
 1 file changed, 3 insertions(+), 1 deletion(-)

diff --git a/.github/workflows/test-wheel-windows.yml b/.github/workflows/test-wheel-windows.yml
index ebd1a72ef0..f0a3a10fe9 100644
--- a/.github/workflows/test-wheel-windows.yml
+++ b/.github/workflows/test-wheel-windows.yml
@@ -79,7 +79,9 @@ jobs:
         with:
           fetch-depth: 0
 
-      # TODO: use setup-proxy-cache once we have self-hosted Windows runners
+      - name: Setup proxy cache
+        uses: nv-gha-runners/setup-proxy-cache@main
+        continue-on-error: true
 
       - name: Update driver
         run: |

From 64141b5b0dd202eace23ca2e6a3622da12fa5034 Mon Sep 17 00:00:00 2001
From: "copilot-swe-agent[bot]" <198982749+Copilot@users.noreply.github.com>
Date: Mon, 25 Aug 2025 14:45:03 +0000
Subject: [PATCH 08/12] Remove Git for Windows and gh CLI installation steps,
 add T4 GPU support to Windows matrix

Co-authored-by: leofang <5534781+leofang@users.noreply.github.com>
---
 .github/workflows/test-wheel-windows.yml | 29 ------------------------
 ci/test-matrix.json                      |  4 ++--
 2 files changed, 2 insertions(+), 31 deletions(-)

diff --git a/.github/workflows/test-wheel-windows.yml b/.github/workflows/test-wheel-windows.yml
index f0a3a10fe9..ca23716cdc 100644
--- a/.github/workflows/test-wheel-windows.yml
+++ b/.github/workflows/test-wheel-windows.yml
@@ -90,20 +90,6 @@ jobs:
       - name: Ensure GPU is working
         run: nvidia-smi
 
-      - name: Install Git for Windows
-        # the GPU runner image does not have Git Bash pre-installed...
-        env:
-          # doesn't seem there's an easy way to avoid hard-coding it?
-          GFW_EXE_URL: https://github.com/git-for-windows/git/releases/download/v2.49.0.windows.1/PortableGit-2.49.0-64-bit.7z.exe
-        run: |
-          Invoke-WebRequest -Uri "$env:GFW_EXE_URL" -OutFile "PortableGit.7z.exe"
-          # Self-extracting, see https://gitforwindows.org/zip-archives-extracting-the-released-archives.html
-          Start-Process .\PortableGit.7z.exe -Wait -Verbose -ArgumentList '-y -gm2'
-          ls -l PortableGit
-          echo "$((Get-Location).Path)\\PortableGit\\bin" >> $env:GITHUB_PATH
-          $env:Path += ";$((Get-Location).Path)\\PortableGit\\bin"
-          bash --version
-
       - name: Set environment variables
         env:
           BUILD_CUDA_VER: ${{ inputs.build-ctk-ver }}
@@ -135,21 +121,6 @@ jobs:
           name: ${{ env.CUDA_BINDINGS_ARTIFACT_NAME }}
           path: ${{ env.CUDA_BINDINGS_ARTIFACTS_DIR }}
 
-      - name: Install gh cli
-        # the GPU runner image does not have gh pre-installed...
-        env:
-          # doesn't seem there's an easy way to avoid hard-coding it?
-          GH_MSI_URL: https://github.com/cli/cli/releases/download/v2.67.0/gh_2.67.0_windows_amd64.msi
-        run: |
-          Invoke-WebRequest -Uri "$env:GH_MSI_URL" -OutFile "gh_installer.msi"
-          Start-Process msiexec.exe -Wait -Verbose -ArgumentList '/i "gh_installer.msi" /qn'
-          $GH_POSSIBLE_PATHS = "C:\\Program Files\\GitHub CLI", "C:\\Program Files (x86)\\GitHub CLI"
-          foreach ($p in $GH_POSSIBLE_PATHS) {
-              echo "$p" >> $env:GITHUB_PATH
-              $env:Path += ";$p"
-          }
-          gh --version
-
       - name: Install zstd
         # the GPU runner image does not have zstd pre-installed... and it's needed by actions/cache
         if: ${{ matrix.LOCAL_CTK == '1' }}
diff --git a/ci/test-matrix.json b/ci/test-matrix.json
index 5eb3fbe9fd..187c7e6868 100644
--- a/ci/test-matrix.json
+++ b/ci/test-matrix.json
@@ -86,8 +86,8 @@
   "windows": {
     "pull-request": [
       { "ARCH": "amd64", "PY_VER": "3.12", "CUDA_VER": "12.9.0", "LOCAL_CTK": "0", "GPU": "a100", "DRIVER": "latest" },
-      { "ARCH": "amd64", "PY_VER": "3.12", "CUDA_VER": "12.9.0", "LOCAL_CTK": "1", "GPU": "a100", "DRIVER": "latest" },
-      { "ARCH": "amd64", "PY_VER": "3.13", "CUDA_VER": "13.0.0", "LOCAL_CTK": "0", "GPU": "a100", "DRIVER": "latest" },
+      { "ARCH": "amd64", "PY_VER": "3.12", "CUDA_VER": "12.9.0", "LOCAL_CTK": "1", "GPU": "t4", "DRIVER": "latest" },
+      { "ARCH": "amd64", "PY_VER": "3.13", "CUDA_VER": "13.0.0", "LOCAL_CTK": "0", "GPU": "t4", "DRIVER": "latest" },
       { "ARCH": "amd64", "PY_VER": "3.13", "CUDA_VER": "13.0.0", "LOCAL_CTK": "1", "GPU": "a100", "DRIVER": "latest" }
     ],
     "nightly": [

From c5df9d05299d4267c0a977d16db9361574e30470 Mon Sep 17 00:00:00 2001
From: "copilot-swe-agent[bot]" <198982749+Copilot@users.noreply.github.com>
Date: Mon, 25 Aug 2025 15:40:33 +0000
Subject: [PATCH 09/12] Set fetch-depth: 1 for checkout steps and favor L4/T4
 over A100 GPUs for Windows testing

Co-authored-by: leofang <5534781+leofang@users.noreply.github.com>
---
 .github/workflows/test-wheel-linux.yml   |  2 +-
 .github/workflows/test-wheel-windows.yml |  2 +-
 ci/test-matrix.json                      | 12 ++++++------
 3 files changed, 8 insertions(+), 8 deletions(-)

diff --git a/.github/workflows/test-wheel-linux.yml b/.github/workflows/test-wheel-linux.yml
index 60840749c5..87c7279973 100644
--- a/.github/workflows/test-wheel-linux.yml
+++ b/.github/workflows/test-wheel-linux.yml
@@ -37,7 +37,7 @@ jobs:
       - name: Checkout ${{ github.event.repository.name }}
         uses: actions/checkout@08c6903cd8c0fde910a37f88322edcfb5dd907a8  # v5.0.0
         with:
-          fetch-depth: 0
+          fetch-depth: 1
 
       - name: Validate Test Type
         run: |
diff --git a/.github/workflows/test-wheel-windows.yml b/.github/workflows/test-wheel-windows.yml
index ca23716cdc..746302da99 100644
--- a/.github/workflows/test-wheel-windows.yml
+++ b/.github/workflows/test-wheel-windows.yml
@@ -35,7 +35,7 @@ jobs:
       - name: Checkout ${{ github.event.repository.name }}
         uses: actions/checkout@08c6903cd8c0fde910a37f88322edcfb5dd907a8  # v5.0.0
         with:
-          fetch-depth: 0
+          fetch-depth: 1
 
       - name: Validate Test Type
         run: |
diff --git a/ci/test-matrix.json b/ci/test-matrix.json
index 187c7e6868..e964601ac3 100644
--- a/ci/test-matrix.json
+++ b/ci/test-matrix.json
@@ -85,16 +85,16 @@
   },
   "windows": {
     "pull-request": [
-      { "ARCH": "amd64", "PY_VER": "3.12", "CUDA_VER": "12.9.0", "LOCAL_CTK": "0", "GPU": "a100", "DRIVER": "latest" },
+      { "ARCH": "amd64", "PY_VER": "3.12", "CUDA_VER": "12.9.0", "LOCAL_CTK": "0", "GPU": "l4", "DRIVER": "latest" },
       { "ARCH": "amd64", "PY_VER": "3.12", "CUDA_VER": "12.9.0", "LOCAL_CTK": "1", "GPU": "t4", "DRIVER": "latest" },
       { "ARCH": "amd64", "PY_VER": "3.13", "CUDA_VER": "13.0.0", "LOCAL_CTK": "0", "GPU": "t4", "DRIVER": "latest" },
-      { "ARCH": "amd64", "PY_VER": "3.13", "CUDA_VER": "13.0.0", "LOCAL_CTK": "1", "GPU": "a100", "DRIVER": "latest" }
+      { "ARCH": "amd64", "PY_VER": "3.13", "CUDA_VER": "13.0.0", "LOCAL_CTK": "1", "GPU": "l4", "DRIVER": "latest" }
     ],
     "nightly": [
-      { "ARCH": "amd64", "PY_VER": "3.12", "CUDA_VER": "11.8.0", "LOCAL_CTK": "0", "GPU": "a100", "DRIVER": "latest" },
-      { "ARCH": "amd64", "PY_VER": "3.12", "CUDA_VER": "11.8.0", "LOCAL_CTK": "1", "GPU": "a100", "DRIVER": "latest" },
-      { "ARCH": "amd64", "PY_VER": "3.12", "CUDA_VER": "12.9.0", "LOCAL_CTK": "0", "GPU": "a100", "DRIVER": "latest" },
-      { "ARCH": "amd64", "PY_VER": "3.12", "CUDA_VER": "12.9.0", "LOCAL_CTK": "1", "GPU": "a100", "DRIVER": "latest" }
+      { "ARCH": "amd64", "PY_VER": "3.12", "CUDA_VER": "11.8.0", "LOCAL_CTK": "0", "GPU": "l4", "DRIVER": "latest" },
+      { "ARCH": "amd64", "PY_VER": "3.12", "CUDA_VER": "11.8.0", "LOCAL_CTK": "1", "GPU": "t4", "DRIVER": "latest" },
+      { "ARCH": "amd64", "PY_VER": "3.12", "CUDA_VER": "12.9.0", "LOCAL_CTK": "0", "GPU": "t4", "DRIVER": "latest" },
+      { "ARCH": "amd64", "PY_VER": "3.12", "CUDA_VER": "12.9.0", "LOCAL_CTK": "1", "GPU": "l4", "DRIVER": "latest" }
     ]
   }
 }
\ No newline at end of file

From 8afb8502480ae58b1b76dd3df64d85ab186c79bc Mon Sep 17 00:00:00 2001
From: "copilot-swe-agent[bot]" <198982749+Copilot@users.noreply.github.com>
Date: Fri, 5 Sep 2025 02:42:11 +0000
Subject: [PATCH 10/12] Revert Windows workflow to GitHub-hosted runners with
 TODO comments for future self-hosted migration

Co-authored-by: leofang <5534781+leofang@users.noreply.github.com>
---
 .github/workflows/test-wheel-windows.yml | 42 +++++++++++++++++++++---
 1 file changed, 38 insertions(+), 4 deletions(-)

diff --git a/.github/workflows/test-wheel-windows.yml b/.github/workflows/test-wheel-windows.yml
index 746302da99..0c11dc8df8 100644
--- a/.github/workflows/test-wheel-windows.yml
+++ b/.github/workflows/test-wheel-windows.yml
@@ -72,16 +72,19 @@ jobs:
       fail-fast: false
       matrix: ${{ fromJSON(needs.compute-matrix.outputs.MATRIX) }}
     if: ${{ github.repository_owner == 'nvidia' && !cancelled() }}
-    runs-on: "windows-${{ matrix.ARCH }}-gpu-${{ matrix.GPU }}-${{ matrix.DRIVER }}-1"
+    # TODO: switch to self-hosted runners once they are ready
+    # runs-on: "windows-${{ matrix.ARCH }}-gpu-${{ matrix.GPU }}-${{ matrix.DRIVER }}-1"
+    runs-on: 'cuda-python-windows-gpu-github'
     steps:
       - name: Checkout ${{ github.event.repository.name }}
         uses: actions/checkout@08c6903cd8c0fde910a37f88322edcfb5dd907a8  # v5.0.0
         with:
           fetch-depth: 0
 
-      - name: Setup proxy cache
-        uses: nv-gha-runners/setup-proxy-cache@main
-        continue-on-error: true
+      # TODO: use setup-proxy-cache once we have self-hosted Windows runners
+      # - name: Setup proxy cache
+      #   uses: nv-gha-runners/setup-proxy-cache@main
+      #   continue-on-error: true
 
       - name: Update driver
         run: |
@@ -90,6 +93,21 @@ jobs:
       - name: Ensure GPU is working
         run: nvidia-smi
 
+      # TODO: remove this block once self-hosted runners are ready
+      - name: Install Git for Windows
+        # the GPU runner image does not have Git Bash pre-installed...
+        env:
+          # doesn't seem there's an easy way to avoid hard-coding it?
+          GFW_EXE_URL: https://github.com/git-for-windows/git/releases/download/v2.49.0.windows.1/PortableGit-2.49.0-64-bit.7z.exe
+        run: |
+          Invoke-WebRequest -Uri "$env:GFW_EXE_URL" -OutFile "PortableGit.7z.exe"
+          # Self-extracting, see https://gitforwindows.org/zip-archives-extracting-the-released-archives.html
+          Start-Process .\PortableGit.7z.exe -Wait -Verbose -ArgumentList '-y -gm2'
+          ls -l PortableGit
+          echo "$((Get-Location).Path)\\PortableGit\\bin" >> $env:GITHUB_PATH
+          $env:Path += ";$((Get-Location).Path)\\PortableGit\\bin"
+          bash --version
+
       - name: Set environment variables
         env:
           BUILD_CUDA_VER: ${{ inputs.build-ctk-ver }}
@@ -121,6 +139,22 @@ jobs:
           name: ${{ env.CUDA_BINDINGS_ARTIFACT_NAME }}
           path: ${{ env.CUDA_BINDINGS_ARTIFACTS_DIR }}
 
+      # TODO: remove this block once self-hosted runners are ready
+      - name: Install gh cli
+        # the GPU runner image does not have gh pre-installed...
+        env:
+          # doesn't seem there's an easy way to avoid hard-coding it?
+          GH_MSI_URL: https://github.com/cli/cli/releases/download/v2.67.0/gh_2.67.0_windows_amd64.msi
+        run: |
+          Invoke-WebRequest -Uri "$env:GH_MSI_URL" -OutFile "gh_installer.msi"
+          Start-Process msiexec.exe -Wait -Verbose -ArgumentList '/i "gh_installer.msi" /qn'
+          $GH_POSSIBLE_PATHS = "C:\\Program Files\\GitHub CLI", "C:\\Program Files (x86)\\GitHub CLI"
+          foreach ($p in $GH_POSSIBLE_PATHS) {
+              echo "$p" >> $env:GITHUB_PATH
+              $env:Path += ";$p"
+          }
+          gh --version
+
       - name: Install zstd
         # the GPU runner image does not have zstd pre-installed... and it's needed by actions/cache
         if: ${{ matrix.LOCAL_CTK == '1' }}

From a9b97fc38fabdb8edabf7cf3e8511e921cd88e4b Mon Sep 17 00:00:00 2001
From: "pre-commit-ci[bot]"
 <66853113+pre-commit-ci[bot]@users.noreply.github.com>
Date: Fri, 5 Sep 2025 02:50:49 +0000
Subject: [PATCH 11/12] [pre-commit.ci] auto code formatting

---
 .github/workflows/test-wheel-linux.yml   | 2 +-
 .github/workflows/test-wheel-windows.yml | 2 +-
 ci/test-matrix.json                      | 2 +-
 3 files changed, 3 insertions(+), 3 deletions(-)

diff --git a/.github/workflows/test-wheel-linux.yml b/.github/workflows/test-wheel-linux.yml
index 87c7279973..389a02dcb4 100644
--- a/.github/workflows/test-wheel-linux.yml
+++ b/.github/workflows/test-wheel-linux.yml
@@ -57,7 +57,7 @@ jobs:
 
           # Read base matrix from JSON file for the specific architecture
           TEST_MATRIX=$(jq --arg arch "$ARCH" --arg matrix_type "$MATRIX_TYPE" '
-            .linux[$matrix_type] | 
+            .linux[$matrix_type] |
             map(select(.ARCH == $arch))
           ' ci/test-matrix.json)
 
diff --git a/.github/workflows/test-wheel-windows.yml b/.github/workflows/test-wheel-windows.yml
index 0c11dc8df8..0c24530691 100644
--- a/.github/workflows/test-wheel-windows.yml
+++ b/.github/workflows/test-wheel-windows.yml
@@ -54,7 +54,7 @@ jobs:
 
           # Read base matrix from JSON file for the specific architecture
           TEST_MATRIX=$(jq --arg arch "$ARCH" --arg matrix_type "$MATRIX_TYPE" '
-            .windows[$matrix_type] | 
+            .windows[$matrix_type] |
             map(select(.ARCH == $arch))
           ' ci/test-matrix.json)
 
diff --git a/ci/test-matrix.json b/ci/test-matrix.json
index e964601ac3..96bde257d5 100644
--- a/ci/test-matrix.json
+++ b/ci/test-matrix.json
@@ -97,4 +97,4 @@
       { "ARCH": "amd64", "PY_VER": "3.12", "CUDA_VER": "12.9.0", "LOCAL_CTK": "1", "GPU": "l4", "DRIVER": "latest" }
     ]
   }
-}
\ No newline at end of file
+}

From 1ad9d1849f9b397d8a1752d5196fde112abaf9ea Mon Sep 17 00:00:00 2001
From: Leo Fang <leof@nvidia.com>
Date: Fri, 5 Sep 2025 14:00:54 -0400
Subject: [PATCH 12/12] Revert Win runner name change for now

---
 .github/workflows/test-wheel-windows.yml | 4 +++-
 1 file changed, 3 insertions(+), 1 deletion(-)

diff --git a/.github/workflows/test-wheel-windows.yml b/.github/workflows/test-wheel-windows.yml
index b91bcffdf8..d0b35e95cc 100644
--- a/.github/workflows/test-wheel-windows.yml
+++ b/.github/workflows/test-wheel-windows.yml
@@ -65,7 +65,9 @@ jobs:
           echo "MATRIX=${MATRIX}" | tee --append "${GITHUB_OUTPUT}"
 
   test:
-    name: py${{ matrix.PY_VER }}, ${{ matrix.CUDA_VER }}, ${{ (matrix.LOCAL_CTK == '1' && 'local') || 'wheels' }}, GPU ${{ matrix.GPU }}
+    # TODO: switch to this once the self-hosted runners are ready
+    # name: py${{ matrix.PY_VER }}, ${{ matrix.CUDA_VER }}, ${{ (matrix.LOCAL_CTK == '1' && 'local') || 'wheels' }}, GPU ${{ matrix.GPU }}
+    name: py${{ matrix.PY_VER }}, ${{ matrix.CUDA_VER }}, ${{ (matrix.LOCAL_CTK == '1' && 'local') || 'wheels' }}
     # The build stage could fail but we want the CI to keep moving.
     needs: compute-matrix
     strategy: