Enable initialize_gigl_runtime in all components #663
Conversation
|
/all_test |
GiGL Automation@ 18:18:11UTC : 🔄 @ 19:22:41UTC : ✅ Workflow completed successfully. |
GiGL Automation@ 18:18:11UTC : 🔄 @ 18:26:23UTC : ✅ Workflow completed successfully. |
GiGL Automation@ 18:18:11UTC : 🔄 @ 18:28:23UTC : ✅ Workflow completed successfully. |
GiGL Automation@ 18:18:11UTC : 🔄 @ 18:19:58UTC : ✅ Workflow completed successfully. |
GiGL Automation@ 18:18:11UTC : 🔄 @ 19:45:32UTC : ✅ Workflow completed successfully. |
GiGL Automation@ 18:18:14UTC : 🔄 |
mkolodner-sc
left a comment
There was a problem hiding this comment.
Thanks Kyle! A few comments/questions
For SubgraphSampler and SplitGenerator, initialize_gigl_runtime now initializes metrics only and skips runtime env-var setup. This centralizes the per-component skip decision in the runtime initializer instead of the local orchestration runner, so Runner._initialize_component_runtime becomes a thin marshaller. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
|
/all_test |
GiGL Automation@ 01:45:57UTC : 🔄 @ 01:47:55UTC : ✅ Workflow completed successfully. |
GiGL Automation@ 01:45:57UTC : 🔄 @ 01:54:40UTC : ✅ Workflow completed successfully. |
GiGL Automation@ 01:45:59UTC : 🔄 @ 02:53:08UTC : ✅ Workflow completed successfully. |
GiGL Automation@ 01:46:01UTC : 🔄 @ 01:54:19UTC : ✅ Workflow completed successfully. |
GiGL Automation@ 01:46:01UTC : 🔄 @ 02:58:30UTC : ✅ Workflow completed successfully. |
GiGL Automation@ 01:46:01UTC : 🔄 @ 03:13:10UTC : ✅ Workflow completed successfully. |
Brings ConfigPopulator and ConfigValidator to parity with the other components by threading cpu_docker_uri/cuda_docker_uri (and thus the GIGL_CPU_DOCKER_URI/GIGL_CUDA_DOCKER_URI env vars) through every entrypoint: the CLI mains, KFP component specs and wiring, and the local Runner (which now routes both through _initialize_component_runtime). Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Note in the SubgraphSampler and SplitGenerator mains that these legacy Scala/Spark components only initialize metrics (not the full GiGL runtime env vars), mirroring the skip branch in initialize_gigl_runtime. Also add a TODO to expose the dataflow docker URI as a GIGL env var for parity with the CPU/CUDA docker URIs. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Scope of work done
As a follow up to #661 we should expose these env vars everywhere (this let's user code be consistent across the code base).
Where is the documentation for this feature?: N/A
Did you add automated tests or write a test plan?
Updated Changelog.md? NO
Ready for code review?: NO