FA-99311 / Distributed training aggregation / Open access

Shard values are placed by list position instead of declared offsets · case 01

Shard values are placed by list position instead of declared offsets.

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

ROOT CAUSE

A worker can send disjoint ranges out of order.

THE FAILURE

A worker can send disjoint ranges out of order.

Unsuccessful approach: Sorting values alone ignores explicit offset gaps and ownership.

Case contract

Given [offset,value] pairs and output length, return dense list with None for unowned positions; duplicate offset last wins.

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(pairs, length):
    out=[None]*length
    for i,(offset,v) in enumerate(pairs):
        if i<length: out[i]=v
    return out
def check(label, actual, expected):
    observations.append({"check": label, "actual": actual, "expected": expected, "passed": actual == expected})
check('unordered ownership',solve([[2,'c'],[0,'a'],[1,'b']],3),['a','b','c'])
check('gap preserved',solve([[2,'c'],[0,'a']],3),['a',None,'c'])
check('empty',solve([],N),[None]*N)
check('length zero',solve([[0,'x']],0),[])
check('duplicate last wins',solve([[1,'old'],[1,'new']],2),[None,'new'])
check('out of range ignored',solve([[-1,'x'],[0,'a']],1),['a'])
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
unordered ownership	`['c', 'a', 'b']`	`['a', 'b', 'c']`	Failed
gap preserved	`['c', 'a', None]`	`['a', None, 'c']`	Failed
empty	`[None]`	`[None]`	Passed
length zero	`[]`	`[]`	Passed
duplicate last wins	`['old', 'new']`	`[None, 'new']`	Failed
out of range ignored	`['x']`	`['a']`	Failed

SHA-256 / d95bcdaf6d797702acca1e2c5083e1f29a58a424768249f52e15ef033383830f

2 / The unsuccessful fix

Exit 1

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

N = 1
observations = []
def solve(pairs, length):
    return [v for i,v in sorted(pairs,key=lambda x:x[1])][:length]
def check(label, actual, expected):
    observations.append({"check": label, "actual": actual, "expected": expected, "passed": actual == expected})
check('unordered ownership',solve([[2,'c'],[0,'a'],[1,'b']],3),['a','b','c'])
check('gap preserved',solve([[2,'c'],[0,'a']],3),['a',None,'c'])
check('empty',solve([],N),[None]*N)
check('length zero',solve([[0,'x']],0),[])
check('duplicate last wins',solve([[1,'old'],[1,'new']],2),[None,'new'])
check('out of range ignored',solve([[-1,'x'],[0,'a']],1),['a'])
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
unordered ownership	`['a', 'b', 'c']`	`['a', 'b', 'c']`	Passed
gap preserved	`['a', 'c']`	`['a', None, 'c']`	Failed
empty	`[]`	`[None]`	Failed
length zero	`[]`	`[]`	Passed
duplicate last wins	`['new', 'old']`	`[None, 'new']`	Failed
out of range ignored	`['a']`	`['a']`	Passed

SHA-256 / 5ce2772bad398206e32aa3b4e7d369734b84e2f1b2f3622880eeeb58b58f3ab0

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 2026-09-30T16:48:22.633482+00:00.

Case digest / 507f49886375420e56a21ac19634a77f9d6e3b7cca546016d49fdd517bcf36ab