Skip to content

podman-test-gpu

podman-test-gpu #3

name: podman-test-gpu
on:
workflow_dispatch:
permissions: read-all
jobs:
test:
runs-on: self-hosted
timeout-minutes: 240
steps:
- name: cleanup workspace
run: |
rm -rf ./* || true
rm -rf ./.??* || true
- uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683 # v4.2.2
# use default docker driver builder with containerd image store for local aikit image
# these must be setup before running this test
- run: docker buildx use desktop-linux
- name: build aikit
run: |
docker buildx build . -t aikit:test \
--load --provenance=false --progress plain
- name: build test model
run: |
docker buildx build . -t sozercan/testmodel:test \
-f test/aikitfile-llama.yaml \
--load --provenance=false --progress plain \
--build-arg="runtime=applesilicon" \
--push
- name: list images
run: docker images
- name: run test model
run: |
podman run --name testmodel -d --rm -p 8080:8080 \
--device /dev/dri \
--pull always \
sozercan/testmodel:test
- name: run test (gguf)
run: |
result=$(curl --fail --retry 10 --retry-all-errors http://127.0.0.1:8080/v1/chat/completions -H "Content-Type: application/json" -d '{
"model": "llama-3.2-1b-instruct",
"messages": [{"role": "user", "content": "explain kubernetes in a sentence"}]
}')
echo $result
choices=$(echo "$result" | jq '.choices')
if [ -z "$choices" ]; then
exit 1
fi
- name: save logs
if: always()
run: podman logs testmodel > /tmp/podman-gpu.log
- run: podman stop testmodel
if: always()
- run: podman system prune -a -f --volumes || true
if: always()
- name: publish test artifacts
if: always()
uses: actions/upload-artifact@b4b15b8c7c6ac21ea08fcf65892d2ee8f75cf882 # v4.4.3
with:
name: test-podman-gpu
path: |
/tmp/*.log