feat(sc): sibling level rollout checkpointing at train boundary - #3923

Merged

terrykong merged 28 commits into

mainfrom

amahishi/partial-rollout-sibling-v3

Sep 8, 2026

Commits on Sep 8, 2026

feat(rollout): add partial rollout lineage ledger
macandro96
committed
6d1c0f8
fix(data-plane): make checkpoint barrier reentrant
macandro96
committed
3947302
feat(rollout): persist sibling recovery state
macandro96
committed
e09aab4
feat(rollout): add configurable recovery granularity
macandro96
committed
ee3906b
fix(rollout): reconcile sibling recovery tests
macandro96
committed
6e33523
test(rollout): cover sibling recovery across restart
macandro96
committed
b7fc4dc
test(rollout): add data-plane checkpoint override
macandro96
committed
402fc4d
test(rollout): validate sibling recovery config
macandro96
committed
d06318d
fix(rollout): propagate prompt index through finalizer
macandro96
committed
d966d3a
fix(rollout): harden sibling recovery checkpointing
macandro96
committed
079f6f6
fix(rollout): remove replay groups by stable id
macandro96
committed
43871ff
fix(rollout): apply prompt-group recovery to live retries
macandro96
committed
ff7fb98
fix(rollout): preserve missing receipts across restart
macandro96
committed
e852891
test(rollout): forward recovery granularity in sibling hook
macandro96
committed
4bfc4ae
fix(rollout): address sibling recovery review findings
macandro96
committed
472ccd4
fix(rollout): address token capture review findings
macandro96
committed
ff36267
refactor(rollout): rename resolved recovery granularity
macandro96
committed
4d31868
fix(sc): enforce data-plane barrier ownership
macandro96
committed
2755a26
fix(rollout): preserve mask state across recovery
macandro96
committed
839f38d
fix(rollout): repair recovery test integration
macandro96
committed
284d404
fix(rollout): convert streamed capture receipts
macandro96
committed
cb78d63
fix(rollout): harden recovery schema coverage
macandro96
committed
9d2b3f6
fix: lint errors
macandro96
committed
6549009
fix(rollout): harden sibling recovery invariants
macandro96
committed
fa7f2bd
refactor(rollout): rename recovery granularity result
macandro96
committed
efcf43b
fix(rollout): align recovery policy with Gym routing
macandro96
committed
34bcf9b
fix(nemo-gym): count resolved rollout agents
macandro96
committed
5a74a6a
test(sc): update rollout recovery fixtures
macandro96
committed
e7447f7

Navigation Menu

feat(sc): sibling level rollout checkpointing at train boundary - #3923

feat(sc): sibling level rollout checkpointing at train boundary#3923
terrykong merged 28 commits into
mainfrom
amahishi/partial-rollout-sibling-v3

Commits on Sep 8, 2026

feat(rollout): add partial rollout lineage ledger

fix(data-plane): make checkpoint barrier reentrant

feat(rollout): persist sibling recovery state

feat(rollout): add configurable recovery granularity

fix(rollout): reconcile sibling recovery tests

test(rollout): cover sibling recovery across restart

test(rollout): add data-plane checkpoint override

test(rollout): validate sibling recovery config

fix(rollout): propagate prompt index through finalizer

fix(rollout): harden sibling recovery checkpointing

fix(rollout): remove replay groups by stable id

fix(rollout): apply prompt-group recovery to live retries

fix(rollout): preserve missing receipts across restart

test(rollout): forward recovery granularity in sibling hook

fix(rollout): address sibling recovery review findings

fix(rollout): address token capture review findings

refactor(rollout): rename resolved recovery granularity

fix(sc): enforce data-plane barrier ownership

fix(rollout): preserve mask state across recovery

fix(rollout): repair recovery test integration

fix(rollout): convert streamed capture receipts

fix(rollout): harden recovery schema coverage

fix: lint errors

fix(rollout): harden sibling recovery invariants

refactor(rollout): rename recovery granularity result

fix(rollout): align recovery policy with Gym routing

fix(nemo-gym): count resolved rollout agents

test(sc): update rollout recovery fixtures