Skip to content

simplification of CI: matrix strategy #3882

simplification of CI: matrix strategy

simplification of CI: matrix strategy #3882

Triggered via pull request February 18, 2025 14:49
Status Failure
Total duration 10h 49m 39s
Artifacts 42

ci.yaml

on: pull_request
metadata
0s
metadata
bump-manifest
36s
bump-manifest
Matrix: build-amd64-arm64
make-publish-configs
0s
make-publish-configs
merge-new-manifest
0s
merge-new-manifest
Matrix: publish-containers
Waiting for pending jobs
finalize  /  workflow-badge
3s
finalize / workflow-badge
finalize  /  report
8s
finalize / report
finalize  /  upload-badge
5s
finalize / upload-badge
finalize  /  publish-badge
4s
finalize / publish-badge
Fit to window
Zoom out
Zoom in

Annotations

9 errors and 2 warnings
build-amd64-arm64 (amd64) / test-jax / jax-A100-unit-test
The self-hosted runner: A100-197018b130b2 lost communication with the server. Verify the machine is running and has a healthy network connection. Anything in your workflow that terminates the runner process, starves it for CPU/Memory, or blocks its network access can cause this error.
build-amd64-arm64 (amd64) / test-nsys-jax / nsys-jax-A100-unit-test
Process completed with exit code 1.
build-amd64-arm64 (amd64) / test-nsys-jax / nsys-jax-A100-unit-test
Process completed with exit code 4.
build-amd64-arm64 (amd64) / test-triton / triton-A100-unit-test
Process completed with exit code 1.
build-amd64-arm64 (amd64) / test-upstream-t5x / test-upstream-t5x-outcome
Process completed with exit code 1.
build-amd64-arm64 (amd64) / test-rosetta-t5x / test-t5x-rosetta-outcome
Process completed with exit code 1.
build-amd64-arm64 (amd64) / test-maxtext / test-maxtext-outcome
Process completed with exit code 1.
make-publish-configs
The template is not valid. .github/workflows/ci.yaml (Line: 332, Col: 14): Error reading JToken from JsonReader. Path '', line 0, position 0.
CI
Error when evaluating 'strategy' for job 'publish-containers'. .github/workflows/ci.yaml (Line: 414, Col: 15): Error parsing fromJson,.github/workflows/ci.yaml (Line: 414, Col: 15): Error reading JToken from JsonReader. Path '', line 0, position 0.,.github/workflows/ci.yaml (Line: 414, Col: 15): Unexpected value ''
make-publish-configs
Encountered an error when evaluating display name ${{ format('declare -a FLAVORS=( base jax triton equinox maxtext levanter upstream-t5x t5x gemma ) declare -a STAGES=( mealkit final ) ## create JSON specs for a 1D matrix of container publication jobs ALL_TAGS=$( echo ''{0}'' \ ''{1}'' |\ jq -s ''add'' ) PUBLISH_CONFIGS=''[]'' for stage in "${{STAGES[@]}}"; do for flavor in "${{FLAVORS[@]}}";do # collect images for different platforms, e.g. amd64 and arm64 matching_tags=$( echo "$ALL_TAGS" |\ jq -c ".[] | select(.stage == \"${{stage}}\" and .flavor == \"${{flavor}}\" and .tag != \"\")" ) # source_image is a list of all platform-specific tags source_image=$(echo "${{matching_tags}}" | jq -c "[.tag]" | jq -s ''add'') # if the build job failed without producing any images, skip this flavor n_source_images=$(echo "$source_image" | jq ''length'') if [[ $n_source_images -gt 0 ]]; then echo "PUBLISH image $flavor with $n_source_images $stage containers" # tag priority is the highest priority of all platform-specific tags priority=$(echo "${{matching_tags}}" | jq -r ".priority" | jq -s ''max'') # put all final images in the `ghcr.io/nvidia/jax` namespace # and mealkit images in `ghcr.io/nvidia/jax-toolbox-mealkit` namespace case ${{stage}} in mealkit) target_image=${{MEALKIT_IMAGE_REPO}} ;; final) target_image=${{FINAL_IMAGE_REPO}} ;; esac PUBLISH_CONFIGS=$( echo ${{PUBLISH_CONFIGS}} | jq -c ". + [{{ \"flavor\": \"${{flavor}}\", \"target_image\": \"${{target_image}}\", \"priority\": \"${{priority}}\", \"source_image\": ${{source_image}}, \"stage\": \"${{stage}}\" }}]" ) else echo "SKIPPED image $flavor with 0 $stage containers" fi done done PUBLISH_CONFIGS=$(echo "$PUBLISH_CONFIGS" | jq -c ''{{"config": .}}'') echo ${{PUBLISH_CONFIGS}} | jq echo "PUBLISH_CONFIGS=${{PUBLISH_CONFIGS}}" >> $GITHUB_OUTPUT ', fromJson(needs.build.outputs['0'].DOCKER_TAGS), fromJson(needs.build.outputs['1'].DOCKER_TAGS)) }}. The template is not valid. .github/workflows/ci.yaml (Line: 332, Col: 14): Error reading JToken from JsonReader. Path '', line 0, position 0.
make-publish-configs
Encountered an error when evaluating display name ${{ format('declare -a FLAVORS=( base jax triton equinox maxtext levanter upstream-t5x t5x gemma ) declare -a STAGES=( mealkit final ) ## create JSON specs for a 1D matrix of container publication jobs ALL_TAGS=$( echo ''{0}'' \ ''{1}'' |\ jq -s ''add'' ) PUBLISH_CONFIGS=''[]'' for stage in "${{STAGES[@]}}"; do for flavor in "${{FLAVORS[@]}}";do # collect images for different platforms, e.g. amd64 and arm64 matching_tags=$( echo "$ALL_TAGS" |\ jq -c ".[] | select(.stage == \"${{stage}}\" and .flavor == \"${{flavor}}\" and .tag != \"\")" ) # source_image is a list of all platform-specific tags source_image=$(echo "${{matching_tags}}" | jq -c "[.tag]" | jq -s ''add'') # if the build job failed without producing any images, skip this flavor n_source_images=$(echo "$source_image" | jq ''length'') if [[ $n_source_images -gt 0 ]]; then echo "PUBLISH image $flavor with $n_source_images $stage containers" # tag priority is the highest priority of all platform-specific tags priority=$(echo "${{matching_tags}}" | jq -r ".priority" | jq -s ''max'') # put all final images in the `ghcr.io/nvidia/jax` namespace # and mealkit images in `ghcr.io/nvidia/jax-toolbox-mealkit` namespace case ${{stage}} in mealkit) target_image=${{MEALKIT_IMAGE_REPO}} ;; final) target_image=${{FINAL_IMAGE_REPO}} ;; esac PUBLISH_CONFIGS=$( echo ${{PUBLISH_CONFIGS}} | jq -c ". + [{{ \"flavor\": \"${{flavor}}\", \"target_image\": \"${{target_image}}\", \"priority\": \"${{priority}}\", \"source_image\": ${{source_image}}, \"stage\": \"${{stage}}\" }}]" ) else echo "SKIPPED image $flavor with 0 $stage containers" fi done done PUBLISH_CONFIGS=$(echo "$PUBLISH_CONFIGS" | jq -c ''{{"config": .}}'') echo ${{PUBLISH_CONFIGS}} | jq echo "PUBLISH_CONFIGS=${{PUBLISH_CONFIGS}}" >> $GITHUB_OUTPUT ', fromJson(needs.build.outputs['0'].DOCKER_TAGS), fromJson(needs.build.outputs['1'].DOCKER_TAGS)) }}. The template is not valid. .github/workflows/ci.yaml (Line: 332, Col: 14): Error reading JToken from JsonReader. Path '', line 0, position 0.

Artifacts

Produced during runtime
Name Size
artifact-base-build-amd64
566 Bytes
artifact-base-build-arm64
566 Bytes
artifact-equinox-build-amd64
570 Bytes
artifact-equinox-build-arm64
568 Bytes
artifact-final-report
5.02 KB
artifact-gemma-build-amd64
560 Bytes
artifact-jax-build-amd64
553 Bytes
artifact-jax-build-arm64
552 Bytes
artifact-levanter-build-amd64
571 Bytes
artifact-levanter-build-arm64
572 Bytes
artifact-maxtext-build-amd64
567 Bytes
artifact-maxtext-build-arm64
567 Bytes
artifact-maxtext-test
661 Bytes
artifact-rosetta-build-t5x-amd64
584 Bytes
artifact-rosetta-build-t5x-arm64
585 Bytes
artifact-rosetta-t5x-mgmn-test
2.31 KB
artifact-t5x-build-amd64
568 Bytes
artifact-t5x-build-arm64
568 Bytes
artifact-triton-build-amd64
565 Bytes
artifact-upstream-t5x-mgmn-test
1.62 KB
artifact-workflow-metadata
277 Bytes
bumped-manifest
46.7 KB
gemma-unit-test-A100
7.75 KB
levanter-unit-test-A100
24.7 KB
nsys-jax-unit-test-A100
30.5 MB
rosetta-t5x-13393331622-1P1G_te-0
6.02 MB
rosetta-t5x-13393331622-1P8G_te-1
5.58 MB
rosetta-t5x-metrics-test-log
5.75 KB
rosetta-t5x-vit-13393331622-VIT1G1N
21.6 KB
rosetta-t5x-vit-13393331622-VIT1P8G
21.7 KB
rosetta-t5x-vit-13393331622-VIT8G1N
33 KB
triton-unit-test-A100
3.18 KB
upstream-maxtext-13393331622-1DP1FSDP1TP1PP
16.6 KB
upstream-maxtext-13393331622-1DP1FSDP8TP1PP
22.8 KB
upstream-maxtext-13393331622-1DP2FSDP4TP1PP_single_process
17 KB
upstream-maxtext-13393331622-1DP4FSDP2TP1PP
22.7 KB
upstream-maxtext-13393331622-1DP8FSDP1TP1PP
22.3 KB
upstream-maxtext-13393331622-2DP2FSDP2TP1PP
22.9 KB
upstream-maxtext-metrics-test-log
2.95 KB
upstream-t5x-13393331622-1P2G_fmha
7.24 MB
upstream-t5x-13393331622-1P8G
7.24 MB
upstream-t5x-metrics-test-log
7.79 KB