Skip to content

Commit

Permalink
revert: "feat: support musa"
Browse files Browse the repository at this point in the history 
This reverts commit 0907d66.

Signed-off-by: thxCode <thxcode0824@gmail.com>
  • Loading branch information
@thxCode
thxCode committed Jul 29, 2024
1 parent 0907d66 commit 7756804
Show file tree
Hide file tree
Showing 2 changed files with 14 additions and 103 deletions.
87 changes: 0 additions & 87 deletions .github/workflows/ci.yml
View file
Original file line number Diff line number Diff line change
Expand Up @@ -475,92 +475,6 @@ jobs:
path: ${{ github.workspace }}/out/*.zip
name: llama-box-linux-${{ matrix.arch }}-cann-${{ matrix.version }}

linux-musa:
strategy:
fail-fast: false
matrix:
# see https://mcconline.mthreads.com/repo/s3000-musa-toolkit?repoName=s3000-musa-toolkit&repoNamespace=mcconline&displayName=S3000_MUSA_Toolkit_%E5%9F%BA%E7%A1%80%E9%95%9C%E5%83%8F.
# 1.5 ==> musa_rtm_v1.5.2
# 2.0 ==> musa_rc_v2.0.0
arch: [ amd64 ]
version: [ '1.5', '2.0' ]
runs-on: ubuntu-22.04
steps:
- name: Maximize Space
# see https://github.com/easimon/maximize-build-space/blob/master/action.yml.
run: |
sudo rm -rf /usr/share/dotnet
sudo rm -rf /usr/local/lib/android
sudo rm -rf /opt/ghc
sudo rm -rf /opt/hostedtoolcache/CodeQL
sudo docker image prune --all --force
- name: Clone
uses: actions/checkout@v4
with:
fetch-depth: 0
submodules: 'recursive'
- name: Setup Cache
timeout-minutes: 5
uses: actions/cache@v3
with:
key: cache-linux-musa-${{ matrix.arch }}-${{ matrix.version }}
path: |
${{ github.workspace }}/.cache
- name: Setup QEMU
if: ${{ matrix.arch == 'arm64' }}
uses: docker/setup-qemu-action@v3
with:
image: tonistiigi/binfmt:qemu-v7.0.0
platforms: "arm64"
- name: Build
# disable OpenMP,
# see https://github.com/ggerganov/llama.cpp/issues/7743#issuecomment-2148342691,
# https://github.com/ggerganov/llama.cpp/issues/7719#issuecomment-2147631216.
env:
CCACHE_DIR: "${{ github.workspace }}/.cache/ccache"
run: |
echo "===== SCRIPT ====="
cat <<EOF > /tmp/entrypoint.sh
#!/bin/bash
apt-get update && apt-get install -y build-essential git cmake ccache
git config --system --add safe.directory '*'
mkdir -p ${{ github.workspace }}/.cache
cmake -S ${{ github.workspace }} -B ${{ github.workspace }}/build -DCMAKE_BUILD_TYPE=Release \
-DGGML_MUSA=on \
${{ matrix.arch == 'amd64' && '-DGGML_NATIVE=off' || '-DGGML_NATIVE=on' }} \
-DGGML_OPENMP=off
cmake --build ${{ github.workspace }}/build --target llama-box --config Release -- -j $(nproc)
echo "===== RESULT ====="
if [ -f ${{ github.workspace }}/build/bin/llama-box ]; then
ldd ${{ github.workspace }}/build/bin/llama-box
else
exit 1
fi
EOF
chmod +x /tmp/entrypoint.sh
cat /tmp/entrypoint.sh
echo "===== BUILD ====="
docker run \
--rm \
--privileged \
--platform linux/${{ matrix.arch }} \
--volume ${{ github.workspace }}:${{ github.workspace }} \
--workdir ${{ github.workspace }} \
--env CCACHE_DIR \
--volume /tmp/entrypoint.sh:/entrypoint.sh \
--entrypoint /entrypoint.sh \
registry.mthreads.com/mcconline/s3000-musa-toolkit:${{ matrix.version == '1.5' && 'musa_rtm_v1.5.2' || 'musa_rc_v2.0.0' }}
echo "===== PACKAGE ====="
mkdir -p ${{ github.workspace }}/out
zip -j ${{ github.workspace }}/out/llama-box-linux-${{ matrix.arch }}-musa-${{ matrix.version }}.zip ${{ github.workspace }}/build/bin/*
- name: Upload Artifact
uses: actions/upload-artifact@v4
with:
path: ${{ github.workspace }}/out/*.zip
name: llama-box-linux-${{ matrix.arch }}-musa-${{ matrix.version }}

windows-hip:
continue-on-error: ${{ !startsWith(github.ref, 'refs/tags/') }}
strategy:
Expand Down Expand Up @@ -832,7 +746,6 @@ jobs:
- linux-cuda
- linux-oneapi
- linux-cann
- linux-musa
- windows-hip
- windows-cuda
- windows-oneapi
Expand Down
30 changes: 14 additions & 16 deletions README.md
View file
Original file line number Diff line number Diff line change
Expand Up @@ -12,22 +12,20 @@ than [llama-server](https://github.com/ggerganov/llama.cpp/blob/master/examples/
Download LLaMA Box from [the latest release](https://github.com/gpustack/llama-box/releases/latest) page please, now
LLaMA Box supports the following platforms.

| Backend | OS/Arch | Supported Devices |
|----------------------------|-------------------------------|-------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------|
| Apple Metal 3 | `darwin/amd64` `darwin/arm64` | Install macOS Ventura or later, see https://support.apple.com/en-sg/102894. |
| NVIDIA CUDA 11.8-s | `linux/amd64` `windows/amd64` | Compute capability is `8.0`, `8.6` or `8.9`, see https://developer.nvidia.com/cuda-gpus. |
| NVIDIA CUDA 12.5-s | `linux/amd64` `windows/amd64` | Compute capability is `8.0`, `8.6` or `8.9`, see https://developer.nvidia.com/cuda-gpus. |
| NVIDIA CUDA 12.5-l | `linux/amd64` `windows/amd64` | Compute capability is `6.0`, `6.1`, `7.0`, `7.5` ,`8.0`, `8.6` or `8.9`, see https://developer.nvidia.com/cuda-gpus. |
| AMD ROCm/HIP 5.5-s | `windows/amd64` | LLVM target is `gfx1030`, `gfx1100`, `gfx1101` or `gfx1102`, see https://rocm.docs.amd.com/en/docs-5.5.1/release/windows_support.html. |
| AMD ROCm/HIP 5.7-s | `linux/amd64` `windows/amd64` | LLVM target is `gfx1030`, `gfx1100`, `gfx1101` or `gfx1102`, see https://rocm.docs.amd.com/en/docs-5.7.1/release/gpu_os_support.html, https://rocm.docs.amd.com/en/docs-5.7.1/release/windows_support.html. |
| AMD ROCm/HIP 5.7-l | `windows/amd64` | LLVM target is `gfx900`, `gfx906`,`gfx908`, `gfx90a`, `gfx940`, `gfx1030`, `gfx1100`, `gfx1101` or `gfx1102`, see https://rocm.docs.amd.com/en/docs-5.7.1/release/windows_support.html. |
| AMD ROCm/HIP 6.1-s | `linux/amd64` | LLVM target is `gfx1030`, `gfx1100`, `gfx1101` or `gfx1102`, see https://rocm.docs.amd.com/en/docs-5.5.1/release/windows_support.html. |
| AMD ROCm/HIP 6.1-l | `linux/amd64` | LLVM target is `gfx900`, `gfx906`,`gfx908`, `gfx90a`, `gfx940`, `gfx1030`, `gfx1100`, `gfx1101` or `gfx1102`, see https://rocm.docs.amd.com/projects/install-on-linux/en/docs-6.1.2/reference/system-requirements.html. |
| Intel oneAPI 2024.1 | `linux/amd64` `windows/amd64` | See Linux/Windows distributions with oneAPI 2024.1 from https://www.intel.com/content/www/us/en/developer/articles/system-requirements/intel-oneapi-base-toolkit-system-requirements.html. |
| Intel oneAPI 2024.2 | `linux/amd64` `windows/amd64` | See Linux/Windows distributions with oneAPI 2024.2 from https://www.intel.com/content/www/us/en/developer/articles/system-requirements/intel-oneapi-base-toolkit-system-requirements.html. |
| Ascend CANN 8.0 | `linux/amd64` `linux/arm64` | See https://www.hiascend.com/document/detail/zh/CANNCommunityEdition/80RC3alpha001/quickstart/quickstart/quickstart_18_0002.html. | |
| Moore Threads MUSA 1.5.2 | `linux/amd64` | `MTT S80`, `MTT S3000` or `MTT S4000`, see https://developer.mthreads.com/sdk/download/musa?equipment=&os=Ubuntu&driverVersion=&version=rc2.1.0. | |
| Moore Threads MUSA rc2.0.0 | `linux/amd64` | `MTT S80`, `MTT S3000` or `MTT S4000`, see https://developer.mthreads.com/sdk/download/musa?equipment=&os=Ubuntu&driverVersion=&version=1.5.2. | |
| Backend | OS/Arch | Supported Devices |
|---------------------|-------------------------------|-------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------|
| Apple Metal 3 | `darwin/amd64` `darwin/arm64` | Install macOS Ventura or later, see https://support.apple.com/en-sg/102894. |
| NVIDIA CUDA 11.8-s | `linux/amd64` `windows/amd64` | Compute capability is `8.0`, `8.6` or `8.9`, see https://developer.nvidia.com/cuda-gpus. |
| NVIDIA CUDA 12.5-s | `linux/amd64` `windows/amd64` | Compute capability is `8.0`, `8.6` or `8.9`, see https://developer.nvidia.com/cuda-gpus. |
| NVIDIA CUDA 12.5-l | `linux/amd64` `windows/amd64` | Compute capability is `6.0`, `6.1`, `7.0`, `7.5` ,`8.0`, `8.6` or `8.9`, see https://developer.nvidia.com/cuda-gpus. |
| AMD ROCm/HIP 5.5-s | `windows/amd64` | LLVM target is `gfx1030`, `gfx1100`, `gfx1101` or `gfx1102`, see https://rocm.docs.amd.com/en/docs-5.5.1/release/windows_support.html. |
| AMD ROCm/HIP 5.7-s | `linux/amd64` `windows/amd64` | LLVM target is `gfx1030`, `gfx1100`, `gfx1101` or `gfx1102`, see https://rocm.docs.amd.com/en/docs-5.7.1/release/gpu_os_support.html, https://rocm.docs.amd.com/en/docs-5.7.1/release/windows_support.html. |
| AMD ROCm/HIP 5.7-l | `windows/amd64` | LLVM target is `gfx900`, `gfx906`,`gfx908`, `gfx90a`, `gfx940`, `gfx1030`, `gfx1100`, `gfx1101` or `gfx1102`, see https://rocm.docs.amd.com/en/docs-5.7.1/release/windows_support.html. |
| AMD ROCm/HIP 6.1-s | `linux/amd64` | LLVM target is `gfx1030`, `gfx1100`, `gfx1101` or `gfx1102`, see https://rocm.docs.amd.com/en/docs-5.5.1/release/windows_support.html. |
| AMD ROCm/HIP 6.1-l | `linux/amd64` | LLVM target is `gfx900`, `gfx906`,`gfx908`, `gfx90a`, `gfx940`, `gfx1030`, `gfx1100`, `gfx1101` or `gfx1102`, see https://rocm.docs.amd.com/projects/install-on-linux/en/docs-6.1.2/reference/system-requirements.html. |
| Intel oneAPI 2024.1 | `linux/amd64` `windows/amd64` | See Linux/Windows distributions with oneAPI 2024.1 from https://www.intel.com/content/www/us/en/developer/articles/system-requirements/intel-oneapi-base-toolkit-system-requirements.html. |
| Intel oneAPI 2024.2 | `linux/amd64` `windows/amd64` | See Linux/Windows distributions with oneAPI 2024.2 from https://www.intel.com/content/www/us/en/developer/articles/system-requirements/intel-oneapi-base-toolkit-system-requirements.html. |
| Ascend CANN 8.0 | `linux/amd64` `linux/arm64` | See https://www.hiascend.com/document/detail/zh/CANNCommunityEdition/80RC3alpha001/quickstart/quickstart/quickstart_18_0002.html | |

## Examples

Expand Down

0 comments on commit 7756804

Please sign in to comment.