From 90a478bd4d97683cd18643cee35e0521fd42dc28 Mon Sep 17 00:00:00 2001
From: Olivier Chafik <ochafik@google.com>
Date: Wed, 8 Jan 2025 17:50:26 +0000
Subject: [PATCH 01/20] Package linux cuda releases for various caps

---
 .github/workflows/build.yml | 85 ++++++++++++++++++++++++++++++++++++-
 ggml/src/kompute            |  1 +
 2 files changed, 84 insertions(+), 2 deletions(-)
 create mode 160000 ggml/src/kompute

diff --git a/.github/workflows/build.yml b/.github/workflows/build.yml
index a377eff38..df9abf304 100644
--- a/.github/workflows/build.yml
+++ b/.github/workflows/build.yml
@@ -809,7 +809,52 @@ jobs:
 
   ubuntu-latest-cmake-cuda:
     runs-on: ubuntu-latest
-    container: nvidia/cuda:12.6.2-devel-ubuntu24.04
+
+    strategy:
+      matrix:
+        cuda:
+          # Colab and lightning.ai currently use CUDA 12.2 (test w/ `nvidia-smi | grep "CUDA Version: "`)
+          # Capabilities of GPUs are listed on https://developer.nvidia.com/cuda-gpus, can test w/ `nvidia-smi --query-gpu=compute_cap --format=csv`
+          # See available containers at https://hub.docker.com/r/nvidia/cuda/tags
+          - version: 12.2
+            container: nvidia/cuda:12.2.2-devel-ubuntu24.04
+            cap: 7.5
+            arch: 75-real
+            example: 'T4'
+            package: true
+          - version: 12.2
+            container: nvidia/cuda:12.2.2-devel-ubuntu24.04
+            cap: 8.0
+            arch: 80-real
+            example: 'A100'
+            package: true
+          - version: 12.2
+            container: nvidia/cuda:12.2.2-devel-ubuntu24.04
+            cap: 8.6
+            arch: 86-real
+            example: 'A10'
+            package: true
+          - version: 12.2
+            container: nvidia/cuda:12.2.2-devel-ubuntu24.04
+            cap: 8.9
+            arch: 89-real
+            example: 'L4, L40S'
+          - version: 12.2
+            container: nvidia/cuda:12.2.2-devel-ubuntu24.04
+            cap: 9.0
+            arch: 90-real
+            example: 'H100'
+            package: true
+          # Build only, don't package.
+          - version: 12.6
+            container: nvidia/cuda:12.6.2-devel-ubuntu24.04
+            cap: 8.9
+            arch: 89-real
+            package: false
+
+    container: nvidia/cuda:${{ matrix.cuda.version }}-devel-ubuntu24.04
+
+    name: Linux CUDA ${{ matrix.cuda.version }} Cap ${{ matrix.cuda.cap }} (e.g. ${{ matrix.cuda_cap.example_gpu }})
 
     steps:
         - name: Clone
@@ -825,9 +870,45 @@ jobs:
 
         - name: Build with CMake
           run: |
-            cmake -S . -B build -G Ninja -DCMAKE_BUILD_TYPE=Release -DGGML_NATIVE=OFF -DGGML_CUDA=ON -DCMAKE_CUDA_ARCHITECTURES=89-real -DCMAKE_EXE_LINKER_FLAGS=-Wl,--allow-shlib-undefined -DLLAMA_FATAL_WARNINGS=ON
+            cmake -S . -B build -G Ninja \
+              -DCMAKE_BUILD_TYPE=Release \
+              -DGGML_NATIVE=OFF \
+              -DGGML_CUDA=ON \
+              -DCMAKE_CUDA_ARCHITECTURES=${{ matrix.cuda.arch }} \
+              -DCMAKE_EXE_LINKER_FLAGS=-Wl,--allow-shlib-undefined \
+              -DLLAMA_CURL=ON \
+              -DLLAMA_FATAL_WARNINGS=ON
             cmake --build build
 
+        - name: Determine tag name
+          if: ${{ matrix.cuda.package }}
+          id: tag
+          shell: bash
+          run: |
+            BUILD_NUMBER="$(git rev-list --count HEAD)"
+            SHORT_HASH="$(git rev-parse --short=7 HEAD)"
+            if [[ "${{ env.BRANCH_NAME }}" == "master" ]]; then
+              echo "name=b${BUILD_NUMBER}" >> $GITHUB_OUTPUT
+            else
+              SAFE_NAME=$(echo "${{ env.BRANCH_NAME }}" | tr '/' '-')
+              echo "name=${SAFE_NAME}-b${BUILD_NUMBER}-${SHORT_HASH}" >> $GITHUB_OUTPUT
+            fi
+            echo "cuda_name=cuda-${{ matrix.cuda.short_version }}-cap-${{ matrix.cuda.cap }}" >> $GITHUB_OUTPUT
+
+        - name: Pack artifacts
+          id: pack_artifacts
+          if: ${{ matrix.cuda.package && (( github.event_name == 'push' && github.ref == 'refs/heads/master' ) || github.event.inputs.create_release == 'true') }}
+          run: |
+            cp LICENSE ./build/bin/
+            zip -r llama-${{ steps.tag.outputs.name }}-bin-ubuntu-x64-${{ steps.tag.outputs.cuda_name }}.zip ./build/bin/*
+
+        - name: Upload artifacts
+          if: ${{ matrix.cuda.package && (( github.event_name == 'push' && github.ref == 'refs/heads/master' ) || github.event.inputs.create_release == 'true') }}
+          uses: actions/upload-artifact@v4
+          with:
+            path: llama-${{ steps.tag.outputs.name }}-bin-ubuntu-x64-${{ steps.tag.outputs.cuda_name }}.zip
+            name: llama-bin-ubuntu-x64.zip
+
   windows-2019-cmake-cuda:
     runs-on: windows-2019
 
diff --git a/ggml/src/kompute b/ggml/src/kompute
new file mode 160000
index 000000000..4565194ed
--- /dev/null
+++ b/ggml/src/kompute
@@ -0,0 +1 @@
+Subproject commit 4565194ed7c32d1d2efa32ceab4d3c6cae006306

From 41b4b11340882f7d2ff8ce75f01012b059a4c80c Mon Sep 17 00:00:00 2001
From: ochafik <ochafik@google.com>
Date: Tue, 14 Jan 2025 02:19:41 +0000
Subject: [PATCH 02/20] Fix container tags + rename

---
 .github/workflows/build.yml | 14 +++++++-------
 1 file changed, 7 insertions(+), 7 deletions(-)

diff --git a/.github/workflows/build.yml b/.github/workflows/build.yml
index b26c476d7..02056fcc4 100644
--- a/.github/workflows/build.yml
+++ b/.github/workflows/build.yml
@@ -815,44 +815,44 @@ jobs:
           # Capabilities of GPUs are listed on https://developer.nvidia.com/cuda-gpus, can test w/ `nvidia-smi --query-gpu=compute_cap --format=csv`
           # See available containers at https://hub.docker.com/r/nvidia/cuda/tags
           - version: 12.2
-            container: nvidia/cuda:12.2.2-devel-ubuntu24.04
+            container: nvidia/cuda:12.2.2-devel-ubuntu22.04
             cap: 7.5
             arch: 75-real
             example: 'T4'
             package: true
           - version: 12.2
-            container: nvidia/cuda:12.2.2-devel-ubuntu24.04
+            container: nvidia/cuda:12.2.2-devel-ubuntu22.04
             cap: 8.0
             arch: 80-real
             example: 'A100'
             package: true
           - version: 12.2
-            container: nvidia/cuda:12.2.2-devel-ubuntu24.04
+            container: nvidia/cuda:12.2.2-devel-ubuntu22.04
             cap: 8.6
             arch: 86-real
             example: 'A10'
             package: true
           - version: 12.2
-            container: nvidia/cuda:12.2.2-devel-ubuntu24.04
+            container: nvidia/cuda:12.2.2-devel-ubuntu22.04
             cap: 8.9
             arch: 89-real
             example: 'L4, L40S'
           - version: 12.2
-            container: nvidia/cuda:12.2.2-devel-ubuntu24.04
+            container: nvidia/cuda:12.2.2-devel-ubuntu22.04
             cap: 9.0
             arch: 90-real
             example: 'H100'
             package: true
           # Build only, don't package.
           - version: 12.6
-            container: nvidia/cuda:12.6.2-devel-ubuntu24.04
+            container: nvidia/cuda:12.6.2-devel-ubuntu22.04
             cap: 8.9
             arch: 89-real
             package: false
 
     container: nvidia/cuda:${{ matrix.cuda.version }}-devel-ubuntu24.04
 
-    name: Linux CUDA ${{ matrix.cuda.version }} Cap ${{ matrix.cuda.cap }} (e.g. ${{ matrix.cuda_cap.example_gpu }})
+    name: ubuntu-22-cuda (${{ matrix.cuda.version }} cap ${{ matrix.cuda.cap }}, e.g. ${{ matrix.cuda.example }})
 
     steps:
         - name: Clone

From 66ec17eb533ce6c4ad52846f4617bb763fdff812 Mon Sep 17 00:00:00 2001
From: ochafik <ochafik@google.com>
Date: Tue, 14 Jan 2025 02:20:50 +0000
Subject: [PATCH 03/20] Update build.yml

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

diff --git a/.github/workflows/build.yml b/.github/workflows/build.yml
index 02056fcc4..e6db8b130 100644
--- a/.github/workflows/build.yml
+++ b/.github/workflows/build.yml
@@ -850,7 +850,7 @@ jobs:
             arch: 89-real
             package: false
 
-    container: nvidia/cuda:${{ matrix.cuda.version }}-devel-ubuntu24.04
+    container: ${{ matrix.cuda.container }}
 
     name: ubuntu-22-cuda (${{ matrix.cuda.version }} cap ${{ matrix.cuda.cap }}, e.g. ${{ matrix.cuda.example }})
 

From 60151daef8c3cd24cc7200ee35d334420f4fc5fa Mon Sep 17 00:00:00 2001
From: ochafik <ochafik@google.com>
Date: Sun, 19 Jan 2025 02:07:58 +0000
Subject: [PATCH 04/20] Use ccache in Docker CUDA build

---
 .devops/cuda.Dockerfile | 5 +++--
 1 file changed, 3 insertions(+), 2 deletions(-)

diff --git a/.devops/cuda.Dockerfile b/.devops/cuda.Dockerfile
index 974dd78a8..b1c14c17a 100644
--- a/.devops/cuda.Dockerfile
+++ b/.devops/cuda.Dockerfile
@@ -12,13 +12,14 @@ FROM ${BASE_CUDA_DEV_CONTAINER} AS build
 ARG CUDA_DOCKER_ARCH=default
 
 RUN apt-get update && \
-    apt-get install -y build-essential cmake python3 python3-pip git libcurl4-openssl-dev libgomp1
+    apt-get install -y build-essential cmake python3 python3-pip git libcurl4-openssl-dev libgomp1 ccache
 
 WORKDIR /app
 
 COPY . .
 
-RUN if [ "${CUDA_DOCKER_ARCH}" != "default" ]; then \
+RUN --mount=type=cache,target=/root/.ccache \
+    if [ "${CUDA_DOCKER_ARCH}" != "default" ]; then \
     export CMAKE_ARGS="-DCMAKE_CUDA_ARCHITECTURES=${CUDA_DOCKER_ARCH}"; \
     fi && \
     cmake -B build -DGGML_NATIVE=OFF -DGGML_CUDA=ON -DLLAMA_CURL=ON ${CMAKE_ARGS} -DCMAKE_EXE_LINKER_FLAGS=-Wl,--allow-shlib-undefined . && \

From a4aed1d3025c7540c924fafc41997fe563978fc6 Mon Sep 17 00:00:00 2001
From: ochafik <ochafik@google.com>
Date: Sun, 19 Jan 2025 12:33:56 +0000
Subject: [PATCH 05/20] Update cuda.Dockerfile

---
 .devops/cuda.Dockerfile | 2 ++
 1 file changed, 2 insertions(+)

diff --git a/.devops/cuda.Dockerfile b/.devops/cuda.Dockerfile
index b1c14c17a..3c5cb0251 100644
--- a/.devops/cuda.Dockerfile
+++ b/.devops/cuda.Dockerfile
@@ -19,6 +19,8 @@ WORKDIR /app
 COPY . .
 
 RUN --mount=type=cache,target=/root/.ccache \
+    --mount=type=cache,target=/var/lib/apt/lists \
+    --mount=type=cache,target=/var/cache/apt \
     if [ "${CUDA_DOCKER_ARCH}" != "default" ]; then \
     export CMAKE_ARGS="-DCMAKE_CUDA_ARCHITECTURES=${CUDA_DOCKER_ARCH}"; \
     fi && \

From 5eb87e9aa361b6a7d915f5f928901dd3252a7d9a Mon Sep 17 00:00:00 2001
From: ochafik <ochafik@google.com>
Date: Mon, 20 Jan 2025 10:51:26 +0000
Subject: [PATCH 06/20] cuda builds: add libcurl

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

diff --git a/.github/workflows/build.yml b/.github/workflows/build.yml
index 9a064a8a3..1f11c70b6 100644
--- a/.github/workflows/build.yml
+++ b/.github/workflows/build.yml
@@ -868,7 +868,7 @@ jobs:
             DEBIAN_FRONTEND: noninteractive
           run: |
               apt update
-              apt install -y cmake build-essential ninja-build libgomp1 git
+              apt install -y cmake build-essential ninja-build libgomp1 git libcurl4-openssl-dev
 
         - name: Build with CMake
           run: |

From 2984d3cade59dbbaf886eafb81f6bf288bd0bfd4 Mon Sep 17 00:00:00 2001
From: ochafik <ochafik@google.com>
Date: Mon, 20 Jan 2025 11:11:23 +0000
Subject: [PATCH 07/20] Align artefact names on existing ones

---
 .github/workflows/build.yml | 10 +++++-----
 1 file changed, 5 insertions(+), 5 deletions(-)

diff --git a/.github/workflows/build.yml b/.github/workflows/build.yml
index 1f11c70b6..c49210eaa 100644
--- a/.github/workflows/build.yml
+++ b/.github/workflows/build.yml
@@ -868,7 +868,7 @@ jobs:
             DEBIAN_FRONTEND: noninteractive
           run: |
               apt update
-              apt install -y cmake build-essential ninja-build libgomp1 git libcurl4-openssl-dev
+              apt install -y cmake build-essential ninja-build libgomp1 git
 
         - name: Build with CMake
           run: |
@@ -895,21 +895,21 @@ jobs:
               SAFE_NAME=$(echo "${{ env.BRANCH_NAME }}" | tr '/' '-')
               echo "name=${SAFE_NAME}-b${BUILD_NUMBER}-${SHORT_HASH}" >> $GITHUB_OUTPUT
             fi
-            echo "cuda_name=cuda-${{ matrix.cuda.short_version }}-cap-${{ matrix.cuda.cap }}" >> $GITHUB_OUTPUT
+            echo "cuda_name=cu${{ matrix.cuda.short_version }}-cap${{ matrix.cuda.cap }}" >> $GITHUB_OUTPUT
 
         - name: Pack artifacts
           id: pack_artifacts
           if: ${{ matrix.cuda.package && (( github.event_name == 'push' && github.ref == 'refs/heads/master' ) || github.event.inputs.create_release == 'true') }}
           run: |
             cp LICENSE ./build/bin/
-            zip -r llama-${{ steps.tag.outputs.name }}-bin-ubuntu-x64-${{ steps.tag.outputs.cuda_name }}.zip ./build/bin/*
+            zip -r llama-${{ steps.tag.outputs.name }}-bin-ubuntu-cuda-${{ steps.tag.outputs.cuda_name }}-x64.zip ./build/bin/*
 
         - name: Upload artifacts
           if: ${{ matrix.cuda.package && (( github.event_name == 'push' && github.ref == 'refs/heads/master' ) || github.event.inputs.create_release == 'true') }}
           uses: actions/upload-artifact@v4
           with:
-            path: llama-${{ steps.tag.outputs.name }}-bin-ubuntu-x64-${{ steps.tag.outputs.cuda_name }}.zip
-            name: llama-bin-ubuntu-x64.zip
+            path: llama-${{ steps.tag.outputs.name }}-bin-ubuntu-cuda-${{ steps.tag.outputs.cuda_name }}-x64.zip
+            name: llama-${{ steps.tag.outputs.name }}-bin-ubuntu-cuda-${{ steps.tag.outputs.cuda_name }}-x64.zip
 
   windows-2019-cmake-cuda:
     runs-on: windows-2019

From abd27fc7a2ac5e7cf9be8735940d50336dedf585 Mon Sep 17 00:00:00 2001
From: Olivier Chafik <ochafik@users.noreply.github.com>
Date: Mon, 20 Jan 2025 12:41:04 +0000
Subject: [PATCH 08/20] Update build.yml

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

diff --git a/.github/workflows/build.yml b/.github/workflows/build.yml
index c49210eaa..526173641 100644
--- a/.github/workflows/build.yml
+++ b/.github/workflows/build.yml
@@ -174,7 +174,7 @@ jobs:
         id: depends
         run: |
           sudo apt-get update
-          sudo apt-get install build-essential libcurl4-openssl-dev
+          sudo apt-get install build-essential curl libcurl4-openssl-dev
 
       - name: Build
         id: cmake_build

From b71c43c294c5b6a7a85ab02954b37f49fc551da4 Mon Sep 17 00:00:00 2001
From: Olivier Chafik <ochafik@users.noreply.github.com>
Date: Mon, 20 Jan 2025 14:52:40 +0000
Subject: [PATCH 09/20] Update build.yml

---
 .github/workflows/build.yml | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/.github/workflows/build.yml b/.github/workflows/build.yml
index 526173641..f9de72abd 100644
--- a/.github/workflows/build.yml
+++ b/.github/workflows/build.yml
@@ -174,7 +174,7 @@ jobs:
         id: depends
         run: |
           sudo apt-get update
-          sudo apt-get install build-essential curl libcurl4-openssl-dev
+          sudo apt-get install build-essential libcurl4-openssl-dev
 
       - name: Build
         id: cmake_build
@@ -868,7 +868,7 @@ jobs:
             DEBIAN_FRONTEND: noninteractive
           run: |
               apt update
-              apt install -y cmake build-essential ninja-build libgomp1 git
+              apt install -y cmake build-essential ninja-build libcurl4-openssl-dev libgomp1 git
 
         - name: Build with CMake
           run: |

From ac045e378e7b6ed584d1ca6d512d3774101f03e7 Mon Sep 17 00:00:00 2001
From: Olivier Chafik <ochafik@users.noreply.github.com>
Date: Mon, 20 Jan 2025 15:46:43 +0000
Subject: [PATCH 10/20] Update build.yml

---
 .github/workflows/build.yml | 2 ++
 1 file changed, 2 insertions(+)

diff --git a/.github/workflows/build.yml b/.github/workflows/build.yml
index f9de72abd..d1330b6c9 100644
--- a/.github/workflows/build.yml
+++ b/.github/workflows/build.yml
@@ -862,6 +862,8 @@ jobs:
         - name: Clone
           id: checkout
           uses: actions/checkout@v4
+          with:
+            fetch-depth: 0
 
         - name: Install dependencies
           env:

From 22ed6028afeadf8bc1d3bfe6f275c2450a181141 Mon Sep 17 00:00:00 2001
From: Olivier Chafik <ochafik@users.noreply.github.com>
Date: Mon, 20 Jan 2025 15:59:29 +0000
Subject: [PATCH 11/20] Temporarily upload artefacts in normal CI run to test
 artefacts

---
 .github/workflows/build.yml | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/.github/workflows/build.yml b/.github/workflows/build.yml
index d1330b6c9..b98147320 100644
--- a/.github/workflows/build.yml
+++ b/.github/workflows/build.yml
@@ -901,17 +901,17 @@ jobs:
 
         - name: Pack artifacts
           id: pack_artifacts
-          if: ${{ matrix.cuda.package && (( github.event_name == 'push' && github.ref == 'refs/heads/master' ) || github.event.inputs.create_release == 'true') }}
+          # if: ${{ matrix.cuda.package && (( github.event_name == 'push' && github.ref == 'refs/heads/master' ) || github.event.inputs.create_release == 'true') }}
           run: |
             cp LICENSE ./build/bin/
             zip -r llama-${{ steps.tag.outputs.name }}-bin-ubuntu-cuda-${{ steps.tag.outputs.cuda_name }}-x64.zip ./build/bin/*
 
         - name: Upload artifacts
-          if: ${{ matrix.cuda.package && (( github.event_name == 'push' && github.ref == 'refs/heads/master' ) || github.event.inputs.create_release == 'true') }}
+          # if: ${{ matrix.cuda.package && (( github.event_name == 'push' && github.ref == 'refs/heads/master' ) || github.event.inputs.create_release == 'true') }}
           uses: actions/upload-artifact@v4
           with:
             path: llama-${{ steps.tag.outputs.name }}-bin-ubuntu-cuda-${{ steps.tag.outputs.cuda_name }}-x64.zip
-            name: llama-${{ steps.tag.outputs.name }}-bin-ubuntu-cuda-${{ steps.tag.outputs.cuda_name }}-x64.zip
+            name: llama-bin-ubuntu-cuda-${{ steps.tag.outputs.cuda_name }}-x64.zip
 
   windows-2019-cmake-cuda:
     runs-on: windows-2019

From 4165293c38d869703b0e82fa4b069639d28a142a Mon Sep 17 00:00:00 2001
From: Olivier Chafik <ochafik@users.noreply.github.com>
Date: Mon, 20 Jan 2025 18:03:30 +0000
Subject: [PATCH 12/20] Attempt to fix weird git error by installing deps
 before clone

---
 .github/workflows/build.yml | 15 ++++++++-------
 1 file changed, 8 insertions(+), 7 deletions(-)

diff --git a/.github/workflows/build.yml b/.github/workflows/build.yml
index b98147320..39ea90d32 100644
--- a/.github/workflows/build.yml
+++ b/.github/workflows/build.yml
@@ -859,19 +859,20 @@ jobs:
     name: ubuntu-22-cuda (${{ matrix.cuda.version }} cap ${{ matrix.cuda.cap }}, e.g. ${{ matrix.cuda.example }})
 
     steps:
+        - name: Dependencies
+          id: depends
+          env:
+            DEBIAN_FRONTEND: noninteractive
+          run: |
+            apt-get update
+            apt install -y cmake build-essential ninja-build libcurl4-openssl-dev libgomp1 git
+
         - name: Clone
           id: checkout
           uses: actions/checkout@v4
           with:
             fetch-depth: 0
 
-        - name: Install dependencies
-          env:
-            DEBIAN_FRONTEND: noninteractive
-          run: |
-              apt update
-              apt install -y cmake build-essential ninja-build libcurl4-openssl-dev libgomp1 git
-
         - name: Build with CMake
           run: |
             cmake -S . -B build -G Ninja \

From c92ae476376609885b7337f38c0ed420b1172d00 Mon Sep 17 00:00:00 2001
From: ochafik <ochafik@google.com>
Date: Tue, 21 Jan 2025 01:59:41 +0000
Subject: [PATCH 13/20] ci: attempt to fix safe directory issue

---
 .github/workflows/build.yml | 1 +
 1 file changed, 1 insertion(+)

diff --git a/.github/workflows/build.yml b/.github/workflows/build.yml
index 39ea90d32..133157933 100644
--- a/.github/workflows/build.yml
+++ b/.github/workflows/build.yml
@@ -872,6 +872,7 @@ jobs:
           uses: actions/checkout@v4
           with:
             fetch-depth: 0
+            set-safe-directory: .
 
         - name: Build with CMake
           run: |

From b7b264cac0241cf26ede2c8ebc6086e1b2231ef0 Mon Sep 17 00:00:00 2001
From: ochafik <ochafik@google.com>
Date: Tue, 21 Jan 2025 01:59:47 +0000
Subject: [PATCH 14/20] ci: setup ccache

---
 .github/workflows/build.yml | 5 +++++
 1 file changed, 5 insertions(+)

diff --git a/.github/workflows/build.yml b/.github/workflows/build.yml
index 133157933..95e3d8a2f 100644
--- a/.github/workflows/build.yml
+++ b/.github/workflows/build.yml
@@ -859,6 +859,11 @@ jobs:
     name: ubuntu-22-cuda (${{ matrix.cuda.version }} cap ${{ matrix.cuda.cap }}, e.g. ${{ matrix.cuda.example }})
 
     steps:
+        - name: ccache
+          uses: hendrikmuhs/ccache-action@v1.2.11
+          with:
+            key: ubuntu-22-cuda-${{ matrix.cuda.version }}-${{ matrix.cuda.cap }}
+
         - name: Dependencies
           id: depends
           env:

From 7a5b18e195821133e4d7dd15b13f2d2b85568acb Mon Sep 17 00:00:00 2001
From: ochafik <ochafik@google.com>
Date: Tue, 21 Jan 2025 02:31:05 +0000
Subject: [PATCH 15/20] ditch ccache action + require cuda in release

---
 .github/workflows/build.yml | 6 +-----
 1 file changed, 1 insertion(+), 5 deletions(-)

diff --git a/.github/workflows/build.yml b/.github/workflows/build.yml
index 95e3d8a2f..02381a284 100644
--- a/.github/workflows/build.yml
+++ b/.github/workflows/build.yml
@@ -859,11 +859,6 @@ jobs:
     name: ubuntu-22-cuda (${{ matrix.cuda.version }} cap ${{ matrix.cuda.cap }}, e.g. ${{ matrix.cuda.example }})
 
     steps:
-        - name: ccache
-          uses: hendrikmuhs/ccache-action@v1.2.11
-          with:
-            key: ubuntu-22-cuda-${{ matrix.cuda.version }}-${{ matrix.cuda.cap }}
-
         - name: Dependencies
           id: depends
           env:
@@ -1292,6 +1287,7 @@ jobs:
 
     needs:
       - ubuntu-latest-cmake
+      - ubuntu-latest-cmake-cuda
       - macOS-latest-cmake
       - windows-latest-cmake
       - windows-2019-cmake-cuda

From f60e148bffcfd3b9bdc28966af7433b637742fb2 Mon Sep 17 00:00:00 2001
From: ochafik <ochafik@google.com>
Date: Tue, 21 Jan 2025 03:28:56 +0000
Subject: [PATCH 16/20] shuffle actions back to original order

---
 .github/workflows/build.yml | 13 ++++++-------
 1 file changed, 6 insertions(+), 7 deletions(-)

diff --git a/.github/workflows/build.yml b/.github/workflows/build.yml
index 02381a284..0125fd5b3 100644
--- a/.github/workflows/build.yml
+++ b/.github/workflows/build.yml
@@ -859,6 +859,12 @@ jobs:
     name: ubuntu-22-cuda (${{ matrix.cuda.version }} cap ${{ matrix.cuda.cap }}, e.g. ${{ matrix.cuda.example }})
 
     steps:
+        - name: Clone
+          id: checkout
+          uses: actions/checkout@v4
+          with:
+            fetch-depth: 0
+
         - name: Dependencies
           id: depends
           env:
@@ -867,13 +873,6 @@ jobs:
             apt-get update
             apt install -y cmake build-essential ninja-build libcurl4-openssl-dev libgomp1 git
 
-        - name: Clone
-          id: checkout
-          uses: actions/checkout@v4
-          with:
-            fetch-depth: 0
-            set-safe-directory: .
-
         - name: Build with CMake
           run: |
             cmake -S . -B build -G Ninja \

From 1b8f9caa0512368fe8db7e96a382576fa4f79958 Mon Sep 17 00:00:00 2001
From: Olivier Chafik <ochafik@google.com>
Date: Fri, 31 Jan 2025 15:48:07 +0000
Subject: [PATCH 17/20] minimize diff

---
 .github/workflows/build.yml | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/.github/workflows/build.yml b/.github/workflows/build.yml
index 2cbabcb37..d5265f4b3 100644
--- a/.github/workflows/build.yml
+++ b/.github/workflows/build.yml
@@ -895,7 +895,7 @@ jobs:
           path: llama-${{ steps.tag.outputs.name }}-bin-win-${{ matrix.build }}.zip
           name: llama-bin-win-${{ matrix.build }}.zip
 
-  ubuntu-cpu-cmake-cuda:
+  ubuntu-latest-cmake-cuda:
     runs-on: ubuntu-latest
 
     strategy:
@@ -956,8 +956,8 @@ jobs:
           env:
             DEBIAN_FRONTEND: noninteractive
           run: |
-            apt-get update
-            apt install -y cmake build-essential ninja-build libcurl4-openssl-dev libgomp1 git
+              apt-get update
+              apt install -y cmake build-essential ninja-build libcurl4-openssl-dev libgomp1 git
 
         - name: ccache
           uses: hendrikmuhs/ccache-action@v1.2.16

From 89da8df649a09a1b5be48e7e94babc2c5d3fba81 Mon Sep 17 00:00:00 2001
From: Olivier Chafik <ochafik@google.com>
Date: Fri, 31 Jan 2025 16:05:29 +0000
Subject: [PATCH 18/20] fix typo

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

diff --git a/.github/workflows/build.yml b/.github/workflows/build.yml
index d5265f4b3..c55f2ecba 100644
--- a/.github/workflows/build.yml
+++ b/.github/workflows/build.yml
@@ -1412,7 +1412,7 @@ jobs:
 
     needs:
       - ubuntu-cpu-cmake
-      - ubuntu-cpu-cmake-cuda
+      - ubuntu-latest-cmake-cuda
       - windows-latest-cmake
       - windows-2019-cmake-cuda
       - windows-latest-cmake-hip-release

From ae175fe7001ed5be3410af471ecb4f01aaf7e40b Mon Sep 17 00:00:00 2001
From: Olivier Chafik <ochafik@google.com>
Date: Fri, 31 Jan 2025 16:55:58 +0000
Subject: [PATCH 19/20] ci + cuda: checkout w/ history when packaging needed

---
 .github/workflows/build.yml | 8 +++++---
 1 file changed, 5 insertions(+), 3 deletions(-)

diff --git a/.github/workflows/build.yml b/.github/workflows/build.yml
index c55f2ecba..d85321c69 100644
--- a/.github/workflows/build.yml
+++ b/.github/workflows/build.yml
@@ -945,9 +945,11 @@ jobs:
     name: ubuntu-22-cuda (${{ matrix.cuda.version }} cap ${{ matrix.cuda.cap }}, e.g. ${{ matrix.cuda.example }})
 
     steps:
-        - name: Clone
-          id: checkout
-          uses: actions/checkout@v4
+        - uses: actions/checkout@v4
+          if: ${{ matrix.cuda.package }}
+
+        - uses: actions/checkout@v4
+          if: ${{ !matrix.cuda.package }}
           with:
             fetch-depth: 0
 

From 167c50071651060de1fc98b5b4e80a43e960cba1 Mon Sep 17 00:00:00 2001
From: Olivier Chafik <ochafik@google.com>
Date: Fri, 31 Jan 2025 17:34:04 +0000
Subject: [PATCH 20/20] install zip in cuda container

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

diff --git a/.github/workflows/build.yml b/.github/workflows/build.yml
index d85321c69..e19d8032a 100644
--- a/.github/workflows/build.yml
+++ b/.github/workflows/build.yml
@@ -959,7 +959,7 @@ jobs:
             DEBIAN_FRONTEND: noninteractive
           run: |
               apt-get update
-              apt install -y cmake build-essential ninja-build libcurl4-openssl-dev libgomp1 git
+              apt install -y cmake build-essential ninja-build libcurl4-openssl-dev libgomp1 git zip
 
         - name: ccache
           uses: hendrikmuhs/ccache-action@v1.2.16