Add topology-portable trainer checkpoints - #784

Open

bradhilton wants to merge 11 commits into

mainfrom

agent/topology-portable-checkpoints

Commits on Aug 11, 2026

Implement topology-portable trainer checkpoints
bradhilton
committed
6392319
Test concurrent checkpoint prefetches
bradhilton
committed
118e34b
Harden checkpoint restoration invariants
bradhilton
committed
98eecd8
Bound checkpoint save idempotency cache
bradhilton
committed
330b16f
Pin trainer base model revisions
bradhilton
committed
cd92267
Validate checkpoint collective payloads
bradhilton
committed
a9304ed
Fix mixed precision checkpoint exchange
bradhilton
committed
3b18695
Honor checkpoint model revision offline
bradhilton
committed
26b4a40
Consolidate checkpoint snapshot serialization
bradhilton
committed
7eb8a33
Initialize checkpoint queue in GPU fixture
bradhilton
committed
f72a2a1
Validate replicated checkpoint state
bradhilton
committed
b422d33