FA-99276 / Distributed training aggregation / Open access

Non-owner replicas contribute gradients for an owned shard · case 01

Non-owner replicas contribute gradients for an owned shard.

Verified by executionVariant 1 · 6 checks per implementationDownload source bundle ↓JSON ↗

ROOT CAUSE

Replicated workers may all hold a shard, but only its designated owner contributes its gradient.

THE FAILURE

Replicated workers may all hold a shard, but only its designated owner contributes its gradient.

Unsuccessful approach: Selecting one arbitrary replica avoids duplicates but may select a stale non-owner.

Case contract

Rows [shard_id, worker_id, owner_id, gradient]; sum only rows where worker equals owner, grouped by shard. Return sorted [shard,total] pairs.

Why this case matters

A toy distributed-training contract makes replica and shard contribution rules explicit without claiming framework compatibility or performance.

1 / The failure

Exit 1

"""Failure Map reference implementation. Python standard library only."""
import json

N = 1
observations = []
def solve(rows):
    d={}
    for s,w,o,g in rows: d[s]=d.get(s,0)+g
    return [[s,d[s]] for s in sorted(d)]
def check(label, actual, expected):
    observations.append({"check": label, "actual": actual, "expected": expected, "passed": actual == expected})
rows0=[['x','a','a',N],['x','b','a',9*N],['y','c','c',2*N]]
check('ignore replica copy',solve(rows0),[['x',N],['y',2*N]])
check('empty owners',solve([]),[])
check('only nonowner',solve([['z','b','a',N]]),[])
check('owner can be late',solve([['z','b','a',N],['z','a','a',3*N]]),[['z',3*N]])
check('several owners',solve([['b','b','b',N],['a','a','a',2*N]]),[['a',2*N],['b',N]])
check('owner contributions add',solve([['q','q','q',N],['q','q','q',2*N]]),[['q',3*N]])
print(json.dumps({"observations": observations, "passed": all(x["passed"] for x in observations)}, ensure_ascii=False))
raise SystemExit(0 if all(x["passed"] for x in observations) else 1)

Boundary fixture	Actual	Expected	Outcome
ignore replica copy	`[['x', 10], ['y', 2]]`	`[['x', 1], ['y', 2]]`	Failed
empty owners	`[]`	`[]`	Passed
only nonowner	`[['z', 1]]`	`[]`	Failed
owner can be late	`[['z', 4]]`	`[['z', 3]]`	Failed
several owners	`[['a', 2], ['b', 1]]`	`[['a', 2], ['b', 1]]`	Passed
owner contributions add	`[['q', 3]]`	`[['q', 3]]`	Passed

SHA-256 / d98f27d8469992d930b705c691ccf7776affdba9ba0164844aa593a029014547

2 / The unsuccessful fix

Exit 1

"""Failure Map reference implementation. Python standard library only."""
import json

N = 1
observations = []
def solve(rows):
    d={}
    for s,w,o,g in rows:
        if s not in d: d[s]=g
    return [[s,d[s]] for s in sorted(d)]
def check(label, actual, expected):
    observations.append({"check": label, "actual": actual, "expected": expected, "passed": actual == expected})
rows0=[['x','a','a',N],['x','b','a',9*N],['y','c','c',2*N]]
check('ignore replica copy',solve(rows0),[['x',N],['y',2*N]])
check('empty owners',solve([]),[])
check('only nonowner',solve([['z','b','a',N]]),[])
check('owner can be late',solve([['z','b','a',N],['z','a','a',3*N]]),[['z',3*N]])
check('several owners',solve([['b','b','b',N],['a','a','a',2*N]]),[['a',2*N],['b',N]])
check('owner contributions add',solve([['q','q','q',N],['q','q','q',2*N]]),[['q',3*N]])
print(json.dumps({"observations": observations, "passed": all(x["passed"] for x in observations)}, ensure_ascii=False))
raise SystemExit(0 if all(x["passed"] for x in observations) else 1)

Boundary fixture	Actual	Expected	Outcome
ignore replica copy	`[['x', 1], ['y', 2]]`	`[['x', 1], ['y', 2]]`	Passed
empty owners	`[]`	`[]`	Passed
only nonowner	`[['z', 1]]`	`[]`	Failed
owner can be late	`[['z', 1]]`	`[['z', 3]]`	Failed
several owners	`[['a', 2], ['b', 1]]`	`[['a', 2], ['b', 1]]`	Passed
owner contributions add	`[['q', 1]]`	`[['q', 3]]`	Failed

SHA-256 / 8ad8e9064f0cc6d1419110dff7632d5e6bcfdfb828f3479ba7774d2214b58783

HELD IN THE MEMBER ARCHIVE

The verified repair and its recorded checks are member-only.

This mechanism has 6 recorded checks per implementation. The open-access tier publishes the failure and the unsuccessful fix; the repaired source that passes every check, and the observations that prove it, are available to members.

Every case sharing this mechanism uses the same contract and the same repair, so this one record is held back for all of them.

Member access is invitation-based. Sign in with your invited account to inspect the repair.

Verification & scope

Inputs are already materialized Python values from a fixed toy collective; no transport, failure recovery, optimizer state, or real model behavior is represented. This reproducer isolates one failure mechanism. Results cover the supplied fixtures. Variants within a family share a test contract and should remain grouped when constructing evaluation splits. Related mechanisms with a shared evaluation_group must also remain together; these controlled models are not independent production incidents.

Observations recorded using Python 3.12.14 at 2026-09-30T16:48:22.178741+00:00.

Case digest / 320a577bb838243d2401079c7e8f708fd5c176cfe25ab261395a0f31a206572e