Repository navigation

S26 E3: faster multi-island weight sync (pipelined sweeps, WAN stream robustness) - #284

Closed

michaellchung wants to merge 9 commits into

Closed

S26 E3: faster multi-island weight sync (pipelined sweeps, WAN stream robustness)#284
michaellchung wants to merge 9 commits into
s25-finalfrom
s26-e3-sync

Commits on Oct 20, 2026
- syncer: pipeline policy-sweep fragments and async per-round checkpoints
  
  chowhsu
  and
  claude
  committed
  8358c1b
- syncer: bf16 sweep pushes and sweep-aligned checkpoint intervals
  
  chowhsu
  and
  claude
  committed
  eb9e229
- syncer: int8 sweep pushes, bf16 downlink deltas, sync/train overlap
  
  chowhsu
  and
  claude
  committed
  1a783d0
- overlap: keep rollout policy age 0, relabel rebased cuts, push whole sweep
  
  chowhsu
  and
  claude
  committed
  8af78fa
- test: overlap island with bf16 downlink deltas against the real syncer
  
  chowhsu
  and
  claude
  committed
  152374b
Commits on Oct 21, 2026
- syncer: send each fragment's FINAL version as an exact FULL broadcast
  
  chowhsu
  and
  claude
  committed
  bc8f53b
- wan: per-stream stall timeout, rate-aware chunk striping, sync diagnostics
  
  chowhsu
  and
  claude
  committed
  3dcbec8
- wan: replay INIT after a mid-init redial, link probe with container relaunch
  
  chowhsu
  and
  claude
  committed
  2509fde
- wan: refuse the link probe with the elastic syncer; test a lost socket mid-INIT
  
  chowhsu
  and
  claude
  committed
  823b2e8