[Victsm]

Summary

Use DG-owned anchor batches and native model results directly instead of rebuilding per-anchor Python requests/results in the supported fixed-size model route. Sequence loading keeps native component merging inside iteration; graph loading preserves topology/features together. Preserve split/shuffle/padding policies, contract validation, and existing Windows/reference fallbacks. Encoder fitting remains unchanged.

Stack slice 4/5. Targets the preceding stack branch; merge in dependency order.

Change map

#282 (comment)

Testing

64 focused native-anchor/model-batch tests passed at this checkpoint; package Ruff check and formatting pass. The complete stack additionally passed 1,951 public tests (3 CUDA skips; 47 notebook deselections).

Dependency order: #279 → #280 → #281 → #282 → #283.

[Victsm]

Change map

User-requested aggregation consolidation, slice 4/5.

Observed requirement and smallest sufficient change: Use DG-owned anchor batches and native model results directly instead of rebuilding per-anchor Python requests/results in the supported fixed-size model route. Sequence loading keeps native component merging inside iteration; graph loading preserves topology/features together. Preserve split/shuffle/padding policies, contract validation, and existing Windows/reference fallbacks. Encoder fitting remains unchanged.

Focused verification / stopping condition: 64 focused native-anchor/model-batch tests passed at this checkpoint; package Ruff check and formatting pass. The complete stack additionally passed 1,951 public tests (3 CUDA skips; 47 notebook deselections). Stop at this slice’s verified behavior; fitting redesign, new temporal/LP capabilities and live sweeps are separate.

File	Function / scope	What changed	Why
packages/experimental/kumo-pytorch/conformance/properties.py	Native work and split invariants	Instrument batch builds; verify deterministic shuffle and exact coverage	Apply the documented engine-order contract to native loaders
packages/experimental/kumo-pytorch/src/kumo/pytorch/assembly.py	PreparedBatchPlan model-column assembly	Accept batch-level native values and enforce model contracts	Keep tensor encoding/layout policy without reconstructing records
packages/experimental/kumo-pytorch/src/kumo/pytorch/data/model.py	ModelSourceBatch	Carry native ordered-event/model value metadata	Adapt native batches while preserving existing model-facing structure
packages/experimental/kumo-pytorch/src/kumo/pytorch/engines/diskgraph.py	DiskGraphSource / native model and scalar adapters	Connect declared contracts to prepared execution and native column batches	Reuse the common backend while preserving model layout and validation
packages/experimental/kumo-pytorch/src/kumo/pytorch/engines/native_anchors.py	NativeAnchors / fitting and batch iteration	Own native seed views and consume native model batches	Preserve identity/order/partitioning without a second Python anchor store
packages/experimental/kumo-pytorch/src/kumo/pytorch/preparation.py	Plan.materialize / materialize_values	Bind native anchors and discover persisted scalar bindings	Preserve explicit fitting/split policy while exposing precompute workflow
packages/experimental/kumo-pytorch/src/kumo/pytorch/task.py	Task batch iteration	Delegate eligible selections to native anchor iteration	Preserve caller ordering, shuffle, rank and padding policy
packages/experimental/kumo-pytorch/tests/conformance/test_functional_harness.py	Conformance negative control	Inject growing work on build_model_batch	Ensure native work regressions still fail the property
packages/experimental/kumo-pytorch/tests/test_model_column_hotpath.py	Focused regression coverage	Verify numerical, clock, identity, lifecycle or route behavior of this slice	Catch wrong answers and unintended fallback/reconstruction, not merely successful calls
packages/experimental/kumo-pytorch/tests/test_native_anchor_handoff.py	Focused regression coverage	Verify numerical, clock, identity, lifecycle or route behavior of this slice	Catch wrong answers and unintended fallback/reconstruction, not merely successful calls

Verified repair — 2026-11-11

Head: 98fa94ecadccc81b4f377df600c54e116a6522d9. Native-only tests skip when DiskGraph is absent. Conformance counts native batch work and checks the documented shuffle contract: determinism and exact coverage, with unshuffled order preserved.

Validation: repaired combined stack passed 146 Rust TT integration tests and 579 Python native/conformance tests with a freshly rebuilt extension. Public Python suite: 1,951 passed; optional-native import checks: two modules skipped as expected without DiskGraph. Generated artifact manifest and 403-component notice checks passed. No live sweeps or merges.

[copy-pr-bot]

This pull request requires additional validation before any workflows can run on NVIDIA's runners.

Pull request vetters can view their responsibilities here.

Contributors can view more details about this message here.

[coderabbitai]

Important

Review skipped

Auto reviews are disabled on base/target branches other than the default branch.

Please check the settings in the CodeRabbit UI or the .coderabbit.yaml file in this repository. To trigger a single review, invoke the @coderabbitai review command.

Run configuration

• Source mode: Auto
• Repository: NVIDIA/kumo-engine
• Branch: codex/anchor-history-native-20261110
• File: .coderabbit.yaml
• Commit: 98fa94e
• Layers: Repository YAML
• Review profile: QUIET
• Plan: Enterprise
• Run ID: 14a213b1-86a2-49d9-9618-e028ed04eeee

You can disable this status message by setting the reviews.review_status to false in the CodeRabbit configuration file.

Use the checkbox below for a quick retry:

• Trigger review

• Autofix · Keep fixing CodeRabbit findings and required CI, and resolving merge conflicts