Repository navigation
Bind deployment recipes to declared frames and token timing - #158
Merged
ElmoPA merged 1 commit intoOct 7, 2026
Merged
Conversation
This was referenced Oct 2, 2026
Collaborator
This was referenced Oct 2, 2026
ElmoPA
marked this pull request as ready for review
October 2, 2026 19:05
This was referenced Oct 7, 2026
ElmoPA
changed the base branch from
codex/graph-integration-20260924/08-parity-hardening
to
graphite-base/158
October 7, 2026 21:20
ElmoPA
force-pushed
the
codex/graph-integration-20260924/09-deployment-contracts
branch
from
October 7, 2026 21:22
1118d53 to
32a40f7
Compare
ElmoPA
added a commit
that referenced
this pull request
Oct 7, 2026
Preserve the legacy HPT trunk initialization capability with an executable source reference. Generate CPU parameters using the original pinned `HPTModel` factory and loader, then load its exact `trunk.` namespace through generic graph initialization. Compare masked outputs, both block outputs, loss, input gradients and all 28 parameter gradients. The local and immutable-Hub transport paths use the same strict tensor/hash validation; only Hub transport is substituted in the fixture test. Complete dispositions for all 83 legacy-only paths: 56 ported, 25 replaced and two without in-repository consumers at the audited source. Each entry has an owner and evidence, with GPU and assembled-destination gates still explicit. No legacy files are deleted by this layer. Fresh read-only destination preflight records unresolved merge paths without modifying either main branch. Validation: isolated snapshot `8db359eabdfd2b1cd903f49cd8e66da3df6e8a03`, **1,454 CPU tests passed**; lock and pinned Ruff pass; recursive tests cover **254/254** YAML/component contexts. Fixture tensors are generated CPU data, not pretrained weights or historical ARC checkpoints. This draft is layer 10, based on #158. OSMO still returns HTTP 403 at the service/login endpoints; no integration GPU workflow has been submitted. Destination assembly and cutover remain incomplete, and DQC stays stopped.
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Nested Yam, ARC, PI and E1 recipes could pair a valid tensor shape with the wrong coordinate frame or tokenizer timing. Extend model-owned data requirements to every shipped ready model, including EgoBridge, and give native/canonical outputs explicit frames. PI selects its deployment adapter frame from the model declaration; E1 binds its experiment-selected source identity. Make the existing Aria stride and E1 human defaults explicit in data YAML.
Add independent model-versus-data composition checks, including negative frame-rate and same-shape timing cases. Add a migration guide connecting the locked environment, retained training recipes, data tools and strict checkpoint loading.
Validation: isolated snapshot
fb9b57cebb7b8c7816df2611f6078a86166e5466, 1,452 CPU tests passed;uv lock --checkand pinned Ruff pass; 254/254 YAML and component contexts pass. This is layer 09, based on #157.Real-weight/data OSMO validation remains blocked by HTTP 403 at the service/login endpoints. No integration GPU workflow has been submitted. The destination merge and legacy cutover remain gated; no historical work is removed and DQC remains stopped.