refactor: isolate model families end to end - #1093
Conversation
|
Caution CodeRabbit couldn't post its review summary. Error details |
5a8e111 to
8c77059
Compare
|
Caution CodeRabbit couldn't post its review summary. Error details |
8c77059 to
8253660
Compare
|
Caution CodeRabbit couldn't post its review summary. Error details |
8253660 to
e0be104
Compare
|
Caution CodeRabbit couldn't post its review summary. Error details |
e0be104 to
3daa7c0
Compare
|
Caution CodeRabbit couldn't post its review summary. Error details |
3daa7c0 to
3d07318
Compare
|
Caution CodeRabbit couldn't post its review summary. Error details |
3d07318 to
474cbe1
Compare
|
Caution CodeRabbit couldn't post its review summary. Error details |
474cbe1 to
89bf793
Compare
|
Caution CodeRabbit couldn't post its review summary. Error details |
10241b6 to
202a7f9
Compare
|
Caution CodeRabbit couldn't post its review summary. Error details |
2 similar comments
|
Caution CodeRabbit couldn't post its review summary. Error details |
|
Caution CodeRabbit couldn't post its review summary. Error details |
505f7c6 to
de21139
Compare
|
Caution CodeRabbit couldn't post its review summary. Error details |
1 similar comment
|
Caution CodeRabbit couldn't post its review summary. Error details |
ce3bd64 to
d7fbc38
Compare
|
Caution CodeRabbit couldn't post its review summary. Error details |
1 similar comment
|
Caution CodeRabbit couldn't post its review summary. Error details |
|
This is an automated Internal CI result; no review from an individual maintainer is requested. Open the public Source Actions run from the automated status link above. |
|
Caution CodeRabbit couldn't post its review summary. Error details |
6 similar comments
|
Caution CodeRabbit couldn't post its review summary. Error details |
|
Caution CodeRabbit couldn't post its review summary. Error details |
|
Caution CodeRabbit couldn't post its review summary. Error details |
|
Caution CodeRabbit couldn't post its review summary. Error details |
|
Caution CodeRabbit couldn't post its review summary. Error details |
|
Caution CodeRabbit couldn't post its review summary. Error details |
|
Important Review skippedToo many files! This PR contains 2966 files, which is 2666 over the limit of 300. To get a review, reduce the PR to 300 files or fewer by splitting it into smaller PRs or changing its base branch. ⚙️ Run configurationConfiguration used: Path: .coderabbit.yaml Review profile: CHILL Plan: Enterprise Run ID: ⛔ Files ignored due to path filters (34)
📒 Files selected for processing (2966)
You can disable this status message by setting the Use the checkbox below for a quick retry:
|
f591438 to
0b63f0c
Compare
7016cea to
6415529
Compare
Move each model into a self-owned build, runtime, dependency, and test slice behind the shared bundle and abstract Task contracts. Keep BYOK, benchmarks, examples, DevToolkit lifecycle, K2-Horizon, and current MoGe behavior as one-way consumers or family-owned implementations. Remove legacy registries, compatibility paths, and custom content digests. BREAKING CHANGE: existing bundles and integrations must be rebuilt against the new public contracts. Signed-off-by: yifeif-nv <277870278+yifeif-nv@users.noreply.github.com>
6415529 to
9777fef
Compare
Background
The previous layout split each model across central builder, runtime,
registration, validation, dependency, and documentation systems. Adding or
changing one family therefore required shared model edits and increased
cross-family coupling.
This PR atomically replaces that layout with the architecture documented in
website/docs/architecture/ai-native-horizontal-scaling.md. It is anintentional breaking cutover: no compatibility layer, migration path, legacy
fallback, or dual implementation remains.
Result
under
families/<family>/: build code, runtime DSO, dependency declaration,manifests, thresholds, and direct tests.
dispatch through sibling families.
model.pyexposes one plainbuild(request, writer)function. Builderinheritance is forbidden.
shared Engine/Bundle contracts.
core/builder/is Python-only andcore/runtime/is C++-only. Sharedinfrastructure contains no central family registry or optional model hooks.
family directories; normal family work stays inside its owner directory.
Preserved Functionality
trtmc-bench, the native benchmark worker, dataset benchmark, andrelease performance matrix as one-way consumers of public APIs.
FoundationPose, LeRobot ACT, VoiceChat full-duplex, and BYOK.
public library boundary.
timm families merged while this PR was open: DenseNet, EfficientNet,
Inception-v3, MNASNet, and RepVGG.
explicit family-owned local command. It intentionally stays outside the
automated manifest inventory because its bundle, frames, and golden evidence
are not repository assets.
Deliberate Simplifications
python/,src/,include/,tests/,benchmarks/, andscripts/architecture roots after moving active functionality tocore/,families/,apps/, andexamples/.fallback reports.
requirements.txt; there is no central dependency catalog.content-digest validation. Runtime checks retain only semantic format, shape,
ABI, ownership, and numerical correctness checks.
isolated.
Coverage Preservation
The dedicated matrix replays the protected behavior of the exact current-main
base and fails closed on missing or reassigned family/case ownership.
SAM2 L4 is additionally retained as a local-only qualification and is not
counted as an automated pass without caller-provisioned evidence.
Validation
Exact head:
9777fef79702c3d511a7e626617913d02101f45fpassed. Eight explicitly GPU-only CTests were skipped in the hardened CPU
lane rather than counted as passes.
tools/model_ci.py validate: valid for all 98 families.trtmc,trtmc-bench, and bundle inspection passed;RTLD_NOW.expected unselected-GPU skips; five timm DSOs and five preprocessing tests
built and passed in both focused and full native builds.
symlinked, incomplete, and accuracy-regressed inputs fail closed.
checks, and
git diff --check: passed.Remaining Gate
The dedicated protected pre-merge and nightly lanes must pass on this exact
head before merge. They provide real checkpoint/native/reference coverage that
is intentionally unavailable in the public CPU lane. No merge is requested by
this update.
Review Notes
structurally different family slices; Task/Engine interfaces; apps/examples;
closed-set architecture tests; CI/package selection.
concrete stable cross-family contract.
Change categories
Risk level
This deliberately replaces every family build/runtime/test path, packaging
layout, and core API/ABI boundary in one incompatible change.