FA-99336 / Distributed training aggregation / Open access

Unknown worker IDs inflate a replica average · case 01

Unknown worker IDs inflate a replica average.

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

ROOT CAUSE

A stale or malformed sender must not participate in the configured collective.

THE FAILURE

A stale or malformed sender must not participate in the configured collective.

Unsuccessful approach: Filtering membership but counting retries still inflates the average.

Case contract

Given configured workers and [worker,value] rows, last listed value per eligible worker contributes; return mean or None.

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(workers, rows):
    return sum(v for w,v in rows)/len(rows) if rows else None
def check(label, actual, expected):
    observations.append({"check": label, "actual": actual, "expected": expected, "passed": actual == expected})
check('foreign ignored',solve(['a','b'],[['a',N],['x',100*N]]),N)
check('retry latest',solve(['a','b'],[['a',N],['a',3*N],['b',2*N]]),2.5*N)
check('empty workers',solve([], [['x',N]]),None)
check('empty rows',solve(['a'],[]),None)
check('single worker',solve(['a'],[['a',N]]),N)
check('unknown-only',solve(['a'],[['x',N],['y',2*N]]),None)
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
foreign ignored	`50.5`	`1`	Failed
retry latest	`2.0`	`2.5`	Failed
empty workers	`1.0`	`None`	Failed
empty rows	`None`	`None`	Passed
single worker	`1.0`	`1`	Passed
unknown-only	`1.5`	`None`	Failed

SHA-256 / 338a7cdcb42e0927430c8e786d533412e79c9210d2748afaf67fd29ce787840d

2 / The unsuccessful fix

Exit 1

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

N = 1
observations = []
def solve(workers, rows):
    valid=[(w,v) for w,v in rows if w in workers]; return sum(v for w,v in valid)/len(valid) if valid else None
def check(label, actual, expected):
    observations.append({"check": label, "actual": actual, "expected": expected, "passed": actual == expected})
check('foreign ignored',solve(['a','b'],[['a',N],['x',100*N]]),N)
check('retry latest',solve(['a','b'],[['a',N],['a',3*N],['b',2*N]]),2.5*N)
check('empty workers',solve([], [['x',N]]),None)
check('empty rows',solve(['a'],[]),None)
check('single worker',solve(['a'],[['a',N]]),N)
check('unknown-only',solve(['a'],[['x',N],['y',2*N]]),None)
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
foreign ignored	`1.0`	`1`	Passed
retry latest	`2.0`	`2.5`	Failed
empty workers	`None`	`None`	Passed
empty rows	`None`	`None`	Passed
single worker	`1.0`	`1`	Passed
unknown-only	`None`	`None`	Passed

SHA-256 / 561439b62eec91b92833bfa1f294e364449d815d39fa6220c374e427d8e07663

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 2026-09-30T16:48:23.079151+00:00.

Case digest / 039f66b9d7b67f17e0560ab1cf36115c5e8153ac7aed430961d815c47e874507