Add: host_build_graph copies of the paged_attention examples - #1667
Add: host_build_graph copies of the paged_attention examples#1667noabauma wants to merge 1 commit into
Conversation
paged_attention and paged_attention_manual_scope existed only for tensormap_and_ringbuffer. pto_orchestration_api.h is byte-identical between the two runtimes and compile_orchestration() selects the include dirs of whichever runtime @scene_test names, so the kernels and orchestration C++ are unchanged copies; the delta is the runtime= decorator, the class name, and ring sizing. host_build_graph builds the whole graph before the device schedules anything, so nothing is reclaimed mid-orchestration and the entire graph must be resident. Case1 (65792 tasks) and Case2 (32832) therefore carry ring_task_window/ring_heap in their own CASES[*]["config"]["runtime_env"], needing no PTO2_RING_* env var. Case3 (head_dim 256) is omitted: it fails on both runtimes, giving max_diff=nan here and max_diff=0.1255 on the unmodified tensormap_and_ringbuffer example, with a normal 16 ms device_wall and no allocator fatal. That is a pre-existing kernel numerics defect at head_dim > 128, not a property of this runtime. Validated on a2a3 with golden checking: 12/12 cases pass across both examples.
|
Warning Review limit reached
Next review available in: 59 minutes You've used all free OSS reviews for now. Wait for the free limit to reset to keep reviewing this public repository. How can I continue?After more reviews become available, a review can be triggered using the To avoid repeated limits, reduce automatic review volume by pausing incremental auto-reviews earlier, using label-based review opt-in, excluding WIP or generated PR titles, or requesting reviews manually when the PR is ready. If your team needs uninterrupted high-volume reviews, an organization admin can enable usage-based reviews. How do review limits work?CodeRabbit enforces per-developer PR review limits for each organization. Most developers receive the normal plan review availability. For paid Pro and Pro+ PR reviews, CodeRabbit uses adaptive limits for sustained high-volume activity. When a developer's recent PR review activity reaches the 95th percentile or higher among CodeRabbit users, additional reviews become available more gradually as earlier reviews age out of the rolling window. Please refer docs for additional details. Review details⚙️ Run configurationConfiguration used: Organization UI Review profile: CHILL Plan: Pro Plus Run ID: 📒 Files selected for processing (14)
Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out. Comment |
1 participant
paged_attention and paged_attention_manual_scope existed only for tensormap_and_ringbuffer. pto_orchestration_api.h is byte-identical between the two runtimes and compile_orchestration() selects the include dirs of whichever runtime @scene_test names, so the kernels and orchestration C++ are unchanged copies; the delta is the runtime= decorator, the class name, and ring sizing.
host_build_graph builds the whole graph before the device schedules anything, so nothing is reclaimed mid-orchestration and the entire graph must be resident. Case1 (65792 tasks) and Case2 (32832) therefore carry ring_task_window/ring_heap in their own CASES[]["config"]["runtime_env"], needing no PTO2_RING_ env var.
Case3 (head_dim 256) is omitted: it fails on both runtimes, giving max_diff=nan here and max_diff=0.1255 on the unmodified tensormap_and_ringbuffer example, with a normal 16 ms device_wall and no allocator fatal. That is a pre-existing kernel numerics defect at head_dim > 128, not a property of this runtime.
Validated on a2a3 with golden checking: 12/12 cases pass across both examples.