Skip to content

[DataFlow runtime 3/7] Data plane: FeatureStore, queue, offline reader, loader#596

Merged
jiapingW merged 1 commit into
sgl-project:dataflow-up-2-contractsfrom
maocheng23:dataflow-up-3-data-plane
Jun 25, 2026
Merged

[DataFlow runtime 3/7] Data plane: FeatureStore, queue, offline reader, loader#596
jiapingW merged 1 commit into
sgl-project:dataflow-up-2-contractsfrom
maocheng23:dataflow-up-3-data-plane

Conversation

@maocheng23

@maocheng23 maocheng23 commented Jun 24, 2026

Copy link
Copy Markdown
Collaborator

DataFlow runtime — stacked PR. Stacked on #595true-stacked: this PR's base is the previous PR's branch, so the diff below shows only this layer.

Part 3/7 — data plane (FeatureStore, queue, offline reader, loader).

Adds specforge/runtime/data_plane/. FeatureStore/LocalFeatureStore is the only tensor-holding component; it serves mem:// (online, RAM, freed on last lease) and file:// (offline .ckpt, mmap) behind one API. Carries lease/generation primitives (consume-once free), a max_resident_bytes backpressure cap, generation-guarded refs (a stale ref after reclaim+republish is rejected, not silently aliased), atomic lease registration, and a best-effort disk dump tap. Plus SampleRefQueue (metadata-only lease/ack/fail), OfflineManifestReader (.ckptfile:// refs), and FeatureDataLoader (SampleRefTrainBatch; queue/refs modes; clone-on-fetch; injected transform/collate so it carries no model knowledge). Tests: test_feature_store, test_feature_dataloader, test_sample_ref_queue. Additive.

Part of a 7-PR series adding the DataFlow runtime (specforge/runtime/, milestones M1–M4). Verified on current upstream main: all subpackages import and 65 component tests pass. The integration launcher (launch.py + train_eagle3_dataflow.py) and the end-to-end equivalence gates are a deliberate follow-up, not in this series.

@gemini-code-assist

Copy link
Copy Markdown
Contributor

Warning

You have reached your daily quota limit. Please wait up to 24 hours and I will start processing your requests again!

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
@maocheng23 maocheng23 force-pushed the dataflow-up-3-data-plane branch from 032df71 to 9b9d84b Compare June 24, 2026 17:49
@maocheng23 maocheng23 changed the base branch from main to dataflow-up-2-contracts June 25, 2026 00:14
@jiapingW jiapingW self-requested a review June 25, 2026 08:41
@jiapingW jiapingW merged commit 7c00b54 into sgl-project:dataflow-up-2-contracts Jun 25, 2026
2 checks passed
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants