Skip to content

Commit 1f1c19d

Browse files
ilopezlunailopezluna
committed
implement build matrix for release workflow
1 parent 5a64bad commit 1f1c19d

File tree

1 file changed

+143
-126
lines changed

1 file changed

+143
-126
lines changed

‎.github/workflows/release.yml‎

Lines changed: 143 additions & 126 deletions
Original file line numberDiff line numberDiff line change
@@ -55,6 +55,52 @@ jobs:
5555
build:
5656
needs: test
5757
runs-on: ubuntu-latest
58+
strategy:
59+
fail-fast: false
60+
matrix:
61+
include:
62+
- name: cpu
63+
target: final-llamacpp
64+
platforms: "linux/amd64, linux/arm64"
65+
tag_suffix: ""
66+
variant: ""
67+
base_image: ""
68+
extra_build_args: ""
69+
70+
- name: cuda
71+
target: final-llamacpp
72+
platforms: "linux/amd64, linux/arm64"
73+
tag_suffix: "-cuda"
74+
variant: "cuda"
75+
base_image: "nvidia/cuda:12.9.0-runtime-ubuntu24.04"
76+
extra_build_args: ""
77+
78+
- name: vllm-cuda
79+
target: final-vllm
80+
platforms: "linux/amd64, linux/arm64"
81+
tag_suffix: "-vllm-cuda"
82+
variant: "cuda"
83+
base_image: "nvidia/cuda:13.0.2-runtime-ubuntu24.04"
84+
extra_build_args: |
85+
VLLM_CUDA_VERSION=cu130
86+
VLLM_PYTHON_TAG=cp38-abi3
87+
88+
- name: sglang-cuda
89+
target: final-sglang
90+
platforms: "linux/amd64"
91+
tag_suffix: "-sglang-cuda"
92+
variant: "cuda"
93+
base_image: "nvidia/cuda:12.9.0-runtime-ubuntu24.04"
94+
extra_build_args: ""
95+
96+
- name: rocm
97+
target: final-llamacpp
98+
platforms: "linux/amd64"
99+
tag_suffix: "-rocm"
100+
variant: "rocm"
101+
base_image: "rocm/dev-ubuntu-22.04"
102+
extra_build_args: ""
103+
58104
steps:
59105
- name: Checkout repo
60106
uses: actions/checkout@8e8c483db84b4bee98b60c0593521ed34d9990e8
@@ -63,46 +109,10 @@ jobs:
63109
id: tags
64110
shell: bash
65111
run: |
66-
echo "cpu<<EOF" >> "$GITHUB_OUTPUT"
67-
echo "docker/model-runner:${{ inputs.releaseTag }}" >> "$GITHUB_OUTPUT"
68-
if [ "${{ inputs.pushLatest }}" == "true" ]; then
69-
echo "docker/model-runner:latest" >> "$GITHUB_OUTPUT"
70-
fi
71-
echo 'EOF' >> "$GITHUB_OUTPUT"
72-
echo "cuda<<EOF" >> "$GITHUB_OUTPUT"
73-
echo "docker/model-runner:${{ inputs.releaseTag }}-cuda" >> "$GITHUB_OUTPUT"
74-
if [ "${{ inputs.pushLatest }}" == "true" ]; then
75-
echo "docker/model-runner:latest-cuda" >> "$GITHUB_OUTPUT"
76-
fi
77-
echo 'EOF' >> "$GITHUB_OUTPUT"
78-
echo "vllm-cuda<<EOF" >> "$GITHUB_OUTPUT"
79-
echo "docker/model-runner:${{ inputs.releaseTag }}-vllm-cuda" >> "$GITHUB_OUTPUT"
80-
if [ "${{ inputs.pushLatest }}" == "true" ]; then
81-
echo "docker/model-runner:latest-vllm-cuda" >> "$GITHUB_OUTPUT"
82-
fi
83-
echo 'EOF' >> "$GITHUB_OUTPUT"
84-
echo "sglang-cuda<<EOF" >> "$GITHUB_OUTPUT"
85-
echo "docker/model-runner:${{ inputs.releaseTag }}-sglang-cuda" >> "$GITHUB_OUTPUT"
86-
if [ "${{ inputs.pushLatest }}" == "true" ]; then
87-
echo "docker/model-runner:latest-sglang-cuda" >> "$GITHUB_OUTPUT"
88-
fi
89-
echo 'EOF' >> "$GITHUB_OUTPUT"
90-
echo "rocm<<EOF" >> "$GITHUB_OUTPUT"
91-
echo "docker/model-runner:${{ inputs.releaseTag }}-rocm" >> "$GITHUB_OUTPUT"
92-
if [ "${{ inputs.pushLatest }}" == "true" ]; then
93-
echo "docker/model-runner:latest-rocm" >> "$GITHUB_OUTPUT"
94-
fi
95-
echo 'EOF' >> "$GITHUB_OUTPUT"
96-
echo "musa<<EOF" >> "$GITHUB_OUTPUT"
97-
echo "docker/model-runner:${{ inputs.releaseTag }}-musa" >> "$GITHUB_OUTPUT"
98-
if [ "${{ inputs.pushLatest }}" == "true" ]; then
99-
echo "docker/model-runner:latest-musa" >> "$GITHUB_OUTPUT"
100-
fi
101-
echo 'EOF' >> "$GITHUB_OUTPUT"
102-
echo "cann<<EOF" >> "$GITHUB_OUTPUT"
103-
echo "docker/model-runner:${{ inputs.releaseTag }}-cann" >> "$GITHUB_OUTPUT"
112+
echo "tags<<EOF" >> "$GITHUB_OUTPUT"
113+
echo "docker/model-runner:${{ inputs.releaseTag }}${{ matrix.tag_suffix }}" >> "$GITHUB_OUTPUT"
104114
if [ "${{ inputs.pushLatest }}" == "true" ]; then
105-
echo "docker/model-runner:latest-cann" >> "$GITHUB_OUTPUT"
115+
echo "docker/model-runner:latest${{ matrix.tag_suffix }}" >> "$GITHUB_OUTPUT"
106116
fi
107117
echo 'EOF' >> "$GITHUB_OUTPUT"
108118
@@ -120,111 +130,118 @@ jobs:
120130
endpoint: "docker/make-product-smarter"
121131
install: true
122132

123-
- name: Build CPU image
124-
uses: docker/build-push-action@263435318d21b8e681c14492fe198d362a7d2c83
125-
with:
126-
file: Dockerfile
127-
target: final-llamacpp
128-
platforms: linux/amd64, linux/arm64
129-
build-args: |
130-
"LLAMA_SERVER_VERSION=${{ inputs.llamaServerVersion }}"
131-
push: true
132-
sbom: true
133-
provenance: mode=max
134-
tags: ${{ steps.tags.outputs.cpu }}
133+
- name: Prepare build args
134+
id: build_args
135+
shell: bash
136+
run: |
137+
ARGS="LLAMA_SERVER_VERSION=${{ inputs.llamaServerVersion }}"
138+
139+
if [ -n "${{ matrix.variant }}" ]; then
140+
ARGS="${ARGS}
141+
LLAMA_SERVER_VARIANT=${{ matrix.variant }}"
142+
fi
143+
144+
if [ -n "${{ matrix.base_image }}" ]; then
145+
ARGS="${ARGS}
146+
BASE_IMAGE=${{ matrix.base_image }}"
147+
fi
148+
149+
# Add vLLM version for vllm builds
150+
if [ "${{ matrix.name }}" == "vllm-cuda" ]; then
151+
ARGS="${ARGS}
152+
VLLM_VERSION=${{ inputs.vllmVersion }}"
153+
fi
154+
155+
# Add SGLang version for sglang builds
156+
if [ "${{ matrix.name }}" == "sglang-cuda" ]; then
157+
ARGS="${ARGS}
158+
SGLANG_VERSION=${{ inputs.sglangVersion }}"
159+
fi
160+
161+
# Add extra build args if present
162+
if [ -n "${{ matrix.extra_build_args }}" ]; then
163+
ARGS="${ARGS}
164+
${{ matrix.extra_build_args }}"
165+
fi
166+
167+
echo "args<<EOF" >> "$GITHUB_OUTPUT"
168+
echo "$ARGS" >> "$GITHUB_OUTPUT"
169+
echo "EOF" >> "$GITHUB_OUTPUT"
135170
136-
- name: Build CUDA image
171+
- name: Build and push ${{ matrix.name }} image
137172
uses: docker/build-push-action@263435318d21b8e681c14492fe198d362a7d2c83
138173
with:
139174
file: Dockerfile
140-
target: final-llamacpp
141-
platforms: linux/amd64, linux/arm64
142-
build-args: |
143-
"LLAMA_SERVER_VERSION=${{ inputs.llamaServerVersion }}"
144-
"LLAMA_SERVER_VARIANT=cuda"
145-
"BASE_IMAGE=nvidia/cuda:12.9.0-runtime-ubuntu24.04"
175+
target: ${{ matrix.target }}
176+
platforms: ${{ matrix.platforms }}
177+
build-args: ${{ steps.build_args.outputs.args }}
146178
push: true
147179
sbom: true
148180
provenance: mode=max
149-
tags: ${{ steps.tags.outputs.cuda }}
181+
tags: ${{ steps.tags.outputs.tags }}
150182

151-
- name: Build vLLM CUDA image
152-
uses: docker/build-push-action@263435318d21b8e681c14492fe198d362a7d2c83
153-
with:
154-
file: Dockerfile
155-
target: final-vllm
156-
platforms: linux/amd64, linux/arm64
157-
build-args: |
158-
"LLAMA_SERVER_VERSION=${{ inputs.llamaServerVersion }}"
159-
"LLAMA_SERVER_VARIANT=cuda"
160-
"BASE_IMAGE=nvidia/cuda:13.0.2-runtime-ubuntu24.04"
161-
"VLLM_VERSION=${{ inputs.vllmVersion }}"
162-
"VLLM_CUDA_VERSION=cu130"
163-
"VLLM_PYTHON_TAG=cp38-abi3"
164-
push: true
165-
sbom: true
166-
provenance: mode=max
167-
tags: ${{ steps.tags.outputs.vllm-cuda }}
183+
build-musa-cann:
184+
needs: test
185+
if: ${{ inputs.buildMusaCann }}
186+
runs-on: ubuntu-latest
187+
strategy:
188+
fail-fast: false
189+
matrix:
190+
include:
191+
- name: musa
192+
target: final-llamacpp
193+
platforms: "linux/amd64"
194+
tag_suffix: "-musa"
195+
variant: "musa"
196+
base_image: "mthreads/musa:rc4.3.0-runtime-ubuntu22.04-amd64"
168197

169-
- name: Build SGLang CUDA image
170-
uses: docker/build-push-action@263435318d21b8e681c14492fe198d362a7d2c83
171-
with:
172-
file: Dockerfile
173-
target: final-sglang
174-
platforms: linux/amd64
175-
build-args: |
176-
"LLAMA_SERVER_VERSION=${{ inputs.llamaServerVersion }}"
177-
"LLAMA_SERVER_VARIANT=cuda"
178-
"BASE_IMAGE=nvidia/cuda:12.9.0-runtime-ubuntu24.04"
179-
"SGLANG_VERSION=${{ inputs.sglangVersion }}"
180-
push: true
181-
sbom: true
182-
provenance: mode=max
183-
tags: ${{ steps.tags.outputs.sglang-cuda }}
198+
- name: cann
199+
target: final-llamacpp
200+
platforms: "linux/arm64, linux/amd64"
201+
tag_suffix: "-cann"
202+
variant: "cann"
203+
base_image: "ascendai/cann:8.2.rc2-910b-ubuntu22.04-py3.11"
184204

185-
- name: Build ROCm image
186-
uses: docker/build-push-action@263435318d21b8e681c14492fe198d362a7d2c83
205+
steps:
206+
- name: Checkout repo
207+
uses: actions/checkout@8e8c483db84b4bee98b60c0593521ed34d9990e8
208+
209+
- name: Format tags
210+
id: tags
211+
shell: bash
212+
run: |
213+
echo "tags<<EOF" >> "$GITHUB_OUTPUT"
214+
echo "docker/model-runner:${{ inputs.releaseTag }}${{ matrix.tag_suffix }}" >> "$GITHUB_OUTPUT"
215+
if [ "${{ inputs.pushLatest }}" == "true" ]; then
216+
echo "docker/model-runner:latest${{ matrix.tag_suffix }}" >> "$GITHUB_OUTPUT"
217+
fi
218+
echo 'EOF' >> "$GITHUB_OUTPUT"
219+
220+
- name: Log in to DockerHub
221+
uses: docker/login-action@5e57cd118135c172c3672efd75eb46360885c0ef
187222
with:
188-
file: Dockerfile
189-
target: final-llamacpp
190-
platforms: linux/amd64
191-
build-args: |
192-
"LLAMA_SERVER_VERSION=${{ inputs.llamaServerVersion }}"
193-
"LLAMA_SERVER_VARIANT=rocm"
194-
"BASE_IMAGE=rocm/dev-ubuntu-22.04"
195-
push: true
196-
sbom: true
197-
provenance: mode=max
198-
tags: ${{ steps.tags.outputs.rocm }}
223+
username: "docker"
224+
password: ${{ secrets.ORG_ACCESS_TOKEN }}
199225

200-
- name: Build MUSA image
201-
if: ${{ inputs.buildMusaCann }}
202-
uses: docker/build-push-action@263435318d21b8e681c14492fe198d362a7d2c83
226+
- name: Set up Buildx
227+
uses: docker/setup-buildx-action@e468171a9de216ec08956ac3ada2f0791b6bd435
203228
with:
204-
file: Dockerfile
205-
target: final-llamacpp
206-
platforms: linux/amd64
207-
build-args: |
208-
"LLAMA_SERVER_VERSION=${{ inputs.llamaServerVersion }}"
209-
"LLAMA_SERVER_VARIANT=musa"
210-
"BASE_IMAGE=mthreads/musa:rc4.3.0-runtime-ubuntu22.04-amd64"
211-
push: true
212-
sbom: true
213-
provenance: mode=max
214-
tags: ${{ steps.tags.outputs.musa }}
229+
version: "lab:latest"
230+
driver: cloud
231+
endpoint: "docker/make-product-smarter"
232+
install: true
215233

216-
- name: Build CANN image
217-
if: ${{ inputs.buildMusaCann }}
234+
- name: Build and push ${{ matrix.name }} image
218235
uses: docker/build-push-action@263435318d21b8e681c14492fe198d362a7d2c83
219236
with:
220237
file: Dockerfile
221-
target: final-llamacpp
222-
platforms: linux/arm64, linux/amd64
238+
target: ${{ matrix.target }}
239+
platforms: ${{ matrix.platforms }}
223240
build-args: |
224-
"LLAMA_SERVER_VERSION=${{ inputs.llamaServerVersion }}"
225-
"LLAMA_SERVER_VARIANT=cann"
226-
"BASE_IMAGE=ascendai/cann:8.2.rc2-910b-ubuntu22.04-py3.11"
241+
LLAMA_SERVER_VERSION=${{ inputs.llamaServerVersion }}
242+
LLAMA_SERVER_VARIANT=${{ matrix.variant }}
243+
BASE_IMAGE=${{ matrix.base_image }}
227244
push: true
228245
sbom: true
229246
provenance: mode=max
230-
tags: ${{ steps.tags.outputs.cann }}
247+
tags: ${{ steps.tags.outputs.tags }}

0 commit comments

Comments
 (0)