FA-99156 / Checkpoint tensor mapping / Open access

Checkpoint mapping: arrival order changes reconstructed shard order · case 01

arrival order changes reconstructed shard order.

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

ROOT CAUSE

Arrival order is mistaken for the declared shard order.

THE FAILURE

Arrival order is mistaken for the declared shard order.

Unsuccessful approach: Sorting by payload length does not recover shard identity.

Case contract

Shards are [index, values]; concatenate values in ascending index order. Indices are unique.

Why this case matters

A bounded offline model of checkpoint tensor bookkeeping in a training workflow.

1 / The failure

Exit 1

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

N = 1
observations = []
def solve(shards):
    return sum((v for i,v in shards), [])
def check(label, actual, expected):
    observations.append({"check": label, "actual": actual, "expected": expected, "passed": actual == expected})
check('reverse', solve(*([[1, [3]], [0, [0]]],)), [0, 3])
check('three', solve(*([[2, [2]], [0, [0]], [1, [1]]],)), [0, 1, 2])
check('single', solve(*([[0, [3]]],)), [3])
check('empty', solve(*([],)), [])
check('unequal', solve(*([[1, [8, 9]], [0, [6]]],)), [6, 8, 9])
check('zero value', solve(*([[1, [0]], [0, [3]]],)), [3, 0])
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
reverse	`[3, 0]`	`[0, 3]`	Failed
three	`[2, 0, 1]`	`[0, 1, 2]`	Failed
single	`[3]`	`[3]`	Passed
empty	`[]`	`[]`	Passed
unequal	`[8, 9, 6]`	`[6, 8, 9]`	Failed
zero value	`[0, 3]`	`[3, 0]`	Failed

SHA-256 / 592d7f12c7688282a03c94d2fba7d61b7469e57ac320f089e3309c31639c40ea

2 / The unsuccessful fix

Exit 1

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

N = 1
observations = []
def solve(shards):
    return sum((v for i,v in sorted(shards, key=lambda x: len(x[1]))), [])
def check(label, actual, expected):
    observations.append({"check": label, "actual": actual, "expected": expected, "passed": actual == expected})
check('reverse', solve(*([[1, [3]], [0, [0]]],)), [0, 3])
check('three', solve(*([[2, [2]], [0, [0]], [1, [1]]],)), [0, 1, 2])
check('single', solve(*([[0, [3]]],)), [3])
check('empty', solve(*([],)), [])
check('unequal', solve(*([[1, [8, 9]], [0, [6]]],)), [6, 8, 9])
check('zero value', solve(*([[1, [0]], [0, [3]]],)), [3, 0])
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
reverse	`[3, 0]`	`[0, 3]`	Failed
three	`[2, 0, 1]`	`[0, 1, 2]`	Failed
single	`[3]`	`[3]`	Passed
empty	`[]`	`[]`	Passed
unequal	`[6, 8, 9]`	`[6, 8, 9]`	Passed
zero value	`[0, 3]`	`[3, 0]`	Failed

SHA-256 / cfd9e51bd76f44418cfa9cd341a8b87ee99d7ad30369a9c161d1bcfc3d9e672a

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 decoded Python values; this toy contract makes no claim about a framework, file format, or production checkpoint. 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 2026-09-30T16:48:20.799687+00:00.

Case digest / 054ac08ca9a060347944a8d4a70269461e5b793040027d0a3579cf5ce666653d