Skip to content

Commit f4b5bf2

Browse files
ggerganovggerganov
authored
ci : re-enable mac workflows (ggml-org#21894)
* ci : re-enable mac workflows * vulkan : fix compile warning
1 parent aa0f189 commit f4b5bf2

3 files changed

Lines changed: 93 additions & 94 deletions

File tree

‎.github/workflows/build-self-hosted.yml‎

Lines changed: 53 additions & 55 deletions
Original file line numberDiff line numberDiff line change
@@ -141,61 +141,59 @@ jobs:
141141
# amd-smi static
142142
# GG_BUILD_ROCM=1 GG_BUILD_AMDGPU_TARGETS="gfx1101" bash ./ci/run.sh ~/results/llama.cpp /mnt/llama.cpp
143143

144-
# TODO: sandbox Mac runners
145-
# ggml-ci-mac-metal:
146-
# runs-on: [self-hosted, macOS, ARM64]
147-
#
148-
# steps:
149-
# - name: Clone
150-
# id: checkout
151-
# uses: actions/checkout@v6
152-
#
153-
# - name: Test
154-
# id: ggml-ci
155-
# run: |
156-
# GG_BUILD_METAL=1 bash ./ci/run.sh ~/results/llama.cpp ~/mnt/llama.cpp
157-
#
158-
# ggml-ci-mac-webgpu:
159-
# runs-on: [self-hosted, macOS, ARM64]
160-
#
161-
# steps:
162-
# - name: Clone
163-
# id: checkout
164-
# uses: actions/checkout@v6
165-
#
166-
# - name: Dawn Dependency
167-
# id: dawn-depends
168-
# run: |
169-
# DAWN_VERSION="v2.0.0"
170-
# DAWN_OWNER="reeselevine"
171-
# DAWN_REPO="dawn"
172-
# DAWN_ASSET_NAME="Dawn-5e9a4865b1635796ccc77dd30057f2b4002a1355-macos-latest-Release"
173-
# echo "Fetching release asset from https://github.com/${DAWN_OWNER}/${DAWN_REPO}/releases/download/${DAWN_VERSION}/${DAWN_ASSET_NAME}.zip"
174-
# curl -L -o artifact.zip \
175-
# "https://github.com/${DAWN_OWNER}/${DAWN_REPO}/releases/download/${DAWN_VERSION}/${DAWN_ASSET_NAME}.zip"
176-
# mkdir dawn
177-
# unzip artifact.zip
178-
# tar -xvf ${DAWN_ASSET_NAME}.tar.gz -C dawn --strip-components=1
179-
#
180-
# - name: Test
181-
# id: ggml-ci
182-
# run: |
183-
# GG_BUILD_WEBGPU=1 GG_BUILD_WEBGPU_DAWN_PREFIX="$GITHUB_WORKSPACE/dawn" \
184-
# bash ./ci/run.sh ~/results/llama.cpp ~/mnt/llama.cpp
185-
#
186-
# ggml-ci-mac-vulkan:
187-
# runs-on: [self-hosted, macOS, ARM64]
188-
#
189-
# steps:
190-
# - name: Clone
191-
# id: checkout
192-
# uses: actions/checkout@v6
193-
#
194-
# - name: Test
195-
# id: ggml-ci
196-
# run: |
197-
# vulkaninfo --summary
198-
# GG_BUILD_VULKAN=1 bash ./ci/run.sh ~/results/llama.cpp ~/mnt/llama.cpp
144+
ggml-ci-mac-metal:
145+
runs-on: [self-hosted, macOS, ARM64]
146+
147+
steps:
148+
- name: Clone
149+
id: checkout
150+
uses: actions/checkout@v6
151+
152+
- name: Test
153+
id: ggml-ci
154+
run: |
155+
GG_BUILD_METAL=1 bash ./ci/run.sh ~/results/llama.cpp ~/mnt/llama.cpp
156+
157+
ggml-ci-mac-webgpu:
158+
runs-on: [self-hosted, macOS, ARM64]
159+
160+
steps:
161+
- name: Clone
162+
id: checkout
163+
uses: actions/checkout@v6
164+
165+
- name: Dawn Dependency
166+
id: dawn-depends
167+
run: |
168+
DAWN_VERSION="v20260317.182325"
169+
DAWN_OWNER="google"
170+
DAWN_REPO="dawn"
171+
DAWN_ASSET_NAME="Dawn-18eb229ef5f707c1464cc581252e7603c73a3ef0-macos-latest-Release"
172+
echo "Fetching release asset from https://github.com/google/dawn/releases/download/${DAWN_VERSION}/${DAWN_ASSET_NAME}.tar.gz"
173+
curl -L -o artifact.tar.gz \
174+
"https://github.com/google/dawn/releases/download/${DAWN_VERSION}/${DAWN_ASSET_NAME}.tar.gz"
175+
mkdir dawn
176+
tar -xvf artifact.tar.gz -C dawn --strip-components=1
177+
178+
- name: Test
179+
id: ggml-ci
180+
run: |
181+
GG_BUILD_WEBGPU=1 GG_BUILD_WEBGPU_DAWN_PREFIX="$GITHUB_WORKSPACE/dawn" \
182+
bash ./ci/run.sh ~/results/llama.cpp ~/mnt/llama.cpp
183+
184+
ggml-ci-mac-vulkan:
185+
runs-on: [self-hosted, macOS, ARM64]
186+
187+
steps:
188+
- name: Clone
189+
id: checkout
190+
uses: actions/checkout@v6
191+
192+
- name: Test
193+
id: ggml-ci
194+
run: |
195+
vulkaninfo --summary
196+
GG_BUILD_VULKAN=1 bash ./ci/run.sh ~/results/llama.cpp ~/mnt/llama.cpp
199197
200198
ggml-ci-linux-intel-vulkan:
201199
runs-on: [self-hosted, Linux, Intel]

‎.github/workflows/server-self-hosted.yml‎

Lines changed: 39 additions & 38 deletions
Original file line numberDiff line numberDiff line change
@@ -84,41 +84,42 @@ jobs:
8484
export ${{ matrix.extra_args }}
8585
pytest -v -x -m "not slow"
8686
87-
server-cuda:
88-
runs-on: [self-hosted, llama-server, Linux, NVIDIA]
89-
90-
name: server-cuda (${{ matrix.wf_name }})
91-
strategy:
92-
matrix:
93-
build_type: [Release]
94-
wf_name: ["GPUx1"]
95-
include:
96-
- build_type: Release
97-
extra_args: "LLAMA_ARG_BACKEND_SAMPLING=1"
98-
wf_name: "GPUx1, backend-sampling"
99-
fail-fast: false
100-
101-
steps:
102-
- name: Clone
103-
id: checkout
104-
uses: actions/checkout@v6
105-
with:
106-
fetch-depth: 0
107-
ref: ${{ github.event.inputs.sha || github.event.pull_request.head.sha || github.sha || github.head_ref || github.ref_name }}
108-
109-
- name: Build
110-
id: cmake_build
111-
run: |
112-
cmake -B build -DGGML_SCHED_NO_REALLOC=ON
113-
cmake --build build --config ${{ matrix.build_type }} -j $(sysctl -n hw.logicalcpu) --target llama-server
114-
115-
- name: Tests
116-
id: server_integration_tests
117-
if: ${{ (!matrix.disabled_on_pr || !github.event.pull_request) }}
118-
run: |
119-
cd tools/server/tests
120-
python3 -m venv venv
121-
source venv/bin/activate
122-
pip install -r requirements.txt
123-
export ${{ matrix.extra_args }}
124-
pytest -v -x -m "not slow"
87+
# TODO: provision CUDA runner
88+
# server-cuda:
89+
# runs-on: [self-hosted, llama-server, Linux, NVIDIA]
90+
#
91+
# name: server-cuda (${{ matrix.wf_name }})
92+
# strategy:
93+
# matrix:
94+
# build_type: [Release]
95+
# wf_name: ["GPUx1"]
96+
# include:
97+
# - build_type: Release
98+
# extra_args: "LLAMA_ARG_BACKEND_SAMPLING=1"
99+
# wf_name: "GPUx1, backend-sampling"
100+
# fail-fast: false
101+
#
102+
# steps:
103+
# - name: Clone
104+
# id: checkout
105+
# uses: actions/checkout@v6
106+
# with:
107+
# fetch-depth: 0
108+
# ref: ${{ github.event.inputs.sha || github.event.pull_request.head.sha || github.sha || github.head_ref || github.ref_name }}
109+
#
110+
# - name: Build
111+
# id: cmake_build
112+
# run: |
113+
# cmake -B build -DGGML_SCHED_NO_REALLOC=ON
114+
# cmake --build build --config ${{ matrix.build_type }} -j $(sysctl -n hw.logicalcpu) --target llama-server
115+
#
116+
# - name: Tests
117+
# id: server_integration_tests
118+
# if: ${{ (!matrix.disabled_on_pr || !github.event.pull_request) }}
119+
# run: |
120+
# cd tools/server/tests
121+
# python3 -m venv venv
122+
# source venv/bin/activate
123+
# pip install -r requirements.txt
124+
# export ${{ matrix.extra_args }}
125+
# pytest -v -x -m "not slow"

‎ggml/src/ggml-webgpu/ggml-webgpu.cpp‎

Lines changed: 1 addition & 1 deletion
Original file line numberDiff line numberDiff line change
@@ -3485,7 +3485,7 @@ static bool create_webgpu_device(ggml_backend_webgpu_reg_context * ctx) {
34853485
dev_desc.requiredFeatureCount = required_features.size();
34863486
dev_desc.SetDeviceLostCallback(
34873487
wgpu::CallbackMode::AllowSpontaneous,
3488-
[ctx](const wgpu::Device & device, wgpu::DeviceLostReason reason, wgpu::StringView message) {
3488+
[](const wgpu::Device & device, wgpu::DeviceLostReason reason, wgpu::StringView message) {
34893489
if (reason == wgpu::DeviceLostReason::Destroyed) {
34903490
return;
34913491
}

0 commit comments

Comments
 (0)