FAILURE MAP
← Case archive

FA-99281 / Distributed training aggregation / Open access

A late gradient from an older collective step is applied · case 01

A late gradient from an older collective step is applied.

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

ROOT CAUSE

Buffered contributions can arrive after a newer step has started.

THE FAILURE

Buffered contributions can arrive after a newer step has started.

Unsuccessful approach: Choosing the maximum observed step discards valid current-step workers when one future item arrives.

Case contract

Given [step,worker,gradient] contributions and target step, return sum for exactly that step.

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, step):
    return sum(g for s,w,g in rows)
def check(label, actual, expected):
    observations.append({"check": label, "actual": actual, "expected": expected, "passed": actual == expected})
rows0=[[N,'a',N],[N,'b',2*N],[N-1,'late',100*N]]
check('old step excluded',solve(rows0,N),3*N)
check('empty',solve([],N),0)
check('no target entries',solve([[N+1,'x',N]],N),0)
check('one current entry',solve([[N,'x',4]],N),4)
check('future does not replace target',solve([[N,'a',N],[N+1,'b',20*N]],N),N)
check('negative contributions',solve([[N,'a',-N],[N,'b',2*N]],N),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 fixtureActualExpectedOutcome
old step excluded1033Failed
empty00Passed
no target entries10Failed
one current entry44Passed
future does not replace target211Failed
negative contributions11Passed

SHA-256 / 568b3cc54f4dfadc1718771687c0cd5a954cfe7941c1a4440c44783e29827311

2 / The unsuccessful fix

Exit 1
"""Failure Map reference implementation. Python standard library only."""
import json

N = 1
observations = []
def solve(rows, step):
    newest=max((s for s,w,g in rows),default=step); return sum(g for s,w,g in rows if s==newest)
def check(label, actual, expected):
    observations.append({"check": label, "actual": actual, "expected": expected, "passed": actual == expected})
rows0=[[N,'a',N],[N,'b',2*N],[N-1,'late',100*N]]
check('old step excluded',solve(rows0,N),3*N)
check('empty',solve([],N),0)
check('no target entries',solve([[N+1,'x',N]],N),0)
check('one current entry',solve([[N,'x',4]],N),4)
check('future does not replace target',solve([[N,'a',N],[N+1,'b',20*N]],N),N)
check('negative contributions',solve([[N,'a',-N],[N,'b',2*N]],N),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 fixtureActualExpectedOutcome
old step excluded33Passed
empty00Passed
no target entries10Failed
one current entry44Passed
future does not replace target201Failed
negative contributions11Passed

SHA-256 / d8ef020d45c0e49cf3ff5a7c6d45e88e7f19363baccf88f43859aea462599d35

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.

Sign in to the archive ↗

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 2026-09-30T16:48:22.633661+00:00.

Case digest / ee18f317dcdbf8a8082668a55fee1a9803f39b15a4bf627438d579a856789c97