Conversation
|
/all_test |
GiGL Automation@ 17:19:51UTC : 🔄 @ 18:49:56UTC : ✅ Workflow completed successfully. |
GiGL Automation@ 17:19:52UTC : 🔄 @ 17:28:36UTC : ✅ Workflow completed successfully. |
GiGL Automation@ 17:19:52UTC : 🔄 @ 18:35:29UTC : ❌ Workflow failed. |
GiGL Automation@ 17:19:56UTC : 🔄 @ 18:43:49UTC : ✅ Workflow completed successfully. |
GiGL Automation@ 17:19:57UTC : 🔄 @ 17:28:24UTC : ✅ Workflow completed successfully. |
…e types Move _create_dist_sampler() and _prepare_degree_tensors() from dist_sampling_producer.py into a new gigl/distributed/utils/dist_sampler.py module as public functions, along with SamplerInput and SamplerRuntime type aliases. Update dist_sampling_producer.py to import from the new module. Add InitSamplingBackendRequest and RegisterBackendRequest frozen dataclasses to messages.py for the upcoming shared sampling backend. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
152cfaa to
f87e4f3
Compare
Introduce SharedDistSamplingBackend, a shared sampling backend that multiplexes multiple compute-rank channels over a single pool of sampling workers using fair-queued round-robin scheduling. This replaces the per-producer model with a per-channel model for graph-store mode. Includes unit tests for helper functions (batch computation, epoch indices, worker seed ranges) and basic lifecycle (init, epoch start, describe). Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
f87e4f3 to
868773e
Compare
|
/all_test |
GiGL Automation@ 18:49:27UTC : 🔄 @ 18:56:26UTC : ✅ Workflow completed successfully. |
GiGL Automation@ 18:49:27UTC : 🔄 @ 18:57:30UTC : ✅ Workflow completed successfully. |
GiGL Automation@ 18:49:28UTC : 🔄 @ 19:59:40UTC : ✅ Workflow completed successfully. |
GiGL Automation@ 18:49:30UTC : 🔄 @ 20:15:48UTC : ✅ Workflow completed successfully. |
GiGL Automation@ 18:49:30UTC : 🔄 @ 20:15:54UTC : ✅ Workflow completed successfully. |
Scope of work done
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