SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2602.11686 · 2026

LAER-MoE: Load-Adaptive Expert Re-layout for Efficient Mixture-of-Experts Training

Bin Cui, Huixia Li, Xuefeng Xiao, Fangcheng Fu, Xinyi Liu, Jiashi Li, Yujie Wang

arXiv · PDF · Open in the Atlas

Code that ran

2 of 4 samples (functions and classes) lifted from this paper's linked repositories executed on a synthesized input in Syntology's jail (0 of them constructed an object rather than computing a result; 1 honoured the contract we drafted, 0 violated it and 0 have no contract verdict; 1 failure of our instrument; 2 no recorded run), measured 2026-11-22; bodies inline only under a permissive licence as GitHub detects it. Not a reproduction of the paper's results.

"Ran" means the function executed on a synthesized input and returned a value. It is not a reproduction of the paper's results. This record, dated, as JSON: /api/ran/2602.11686.json · badge: /api/ran/2602.11686.svg

Each repository's role is a label: a classifier's output recorded on the link, or, where no role is recorded, what the Papers with Code archive marked. The line under it names its basis. Neither says who wrote the code.

RepositoryRoleRan
PKU-DAIR/Hetu-Galvatron canonical (classifier label)
classifier basis: census
2 ran · 2 no recorded run
FunctionStatusWhere it lives
get_time_cost_all_stages Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
PKU-DAIR/Hetu-Galvatron/galvatron/core/cost_model/cost_model_handler.py
code served (permissive licence) · get_code("e150d0d2237a0eea")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned list, ndim 1, dtype family float; no output shape is stated. The image it ran in was not recorded. If your copy returns the same type and ndim on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import numpy as np

# Analyze the function:
# - layer_timecosts: used with len() and indexing [layer_start_id:layer_end_id]
#   Should be an array-like of numeric values representing time costs per layer
# - pp_stage_division: used with len(), np.sum() on slices [:stage_id] and [:stage_id + 1]
#   Should be an array of integers that sum to len(layer_timecosts)

# Create a simple scenario:
# Let's say we have 6 layers divided into 3 stages with 2 layers each

layer_timecosts = np.array([1.0, 2.0, 3.0, 4.0, 5.0, 6.0])  # 6 layers with different costs

pp_stage_division = np.array([2, 2, 2])  # 3 stages: 2 layers each

# Verify the assertion will pass
assert np.sum(pp_stage_division) == len(layer_timecosts)

ARGS = {
    "layer_timecosts": layer_timecosts,
    "pp_stage_division": pp_stage_division,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type list, ndim 1, dtype family float — a list output has no array shape of its own; none is stated
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
print_peak_memory Ran
ran, honoured the contract we drafted
PKU-DAIR/Hetu-Galvatron/galvatron/core/profiler/utils.py
code served (permissive licence) · get_code("5cefabd8436714e8")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned tuple, ndim 1, dtype family float; no output shape is stated. The image it ran in was not recorded. The code mentions cuda, and the harness aliases cuda to the CPU for every sample; whether this run needed that is not recorded. If your copy returns the same type and ndim on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import torch

torch.manual_seed(0)

# The function uses torch.cuda.max_memory_allocated, torch.cuda.memory_allocated,
# torch.cuda.max_memory_reserved, and torch.cuda.memory_reserved.
# These require a device argument (can be CPU or CUDA device).
# The function prints memory statistics and returns a tuple of (max_mem, cur_mem).

# Since we're CPU-only, we'll use a CPU device
device = 'cpu'

# The prefix is just a string that gets printed
prefix = "Memory Check"

# The type parameter should be one of 'allocated' or 'reserved'
memory_type = 'allocated'

ARGS = {
    'prefix': prefix,
    'device': device,
    'type': memory_type
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type tuple, ndim 1, dtype family float — a tuple output has no array shape of its own; none is stated
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
galvatron_profile_args Not yet run
unverified (harvested, no recorded run)
PKU-DAIR/Hetu-Galvatron/galvatron/core/profiler/arguments.py
code served (permissive licence) · get_code("c432b2807aeec9e0")
How it ran

no invocation recorded.

galvatron_profile_hardware_args Not yet run
unverified (harvested, no recorded run)
PKU-DAIR/Hetu-Galvatron/galvatron/core/profiler/arguments.py
code served (permissive licence) · get_code("01be37157b6581bc")
How it ran

no invocation recorded.

Also present in other papers' code

Of 4 samples lifted from this paper's linked repositories, 4 carry a function identity and 0 do not. Of the 4 distinct functions with an identity, 4 are also present in other papers whose code we hold (4 with at least one paper dated earlier, 0 the same month, 0 later) and 0 are only here among the 52,977 papers whose code Syntology holds (as of 2026-09-26).

Only here means: among the 52,977 papers whose code Syntology holds (51,887 with lifted samples, 1,090 through their assemblies only), as of 2026-09-26, no other paper's lifted samples or current assembly bricks contain an identical function body after normalisation.

FunctionAlso present inEarliest-datedOthers named
galvatron_profile_args 1 other paper
1 dated earlier
2412.01523 (2024-12)
FlexSP: Accelerating Large Language Model Training via Flexible Sequence Parallelism
galvatron_profile_hardware_args 1 other paper
1 dated earlier
2412.01523 (2024-12)
FlexSP: Accelerating Large Language Model Training via Flexible Sequence Parallelism
get_time_cost_all_stages 1 other paper
1 dated earlier
2412.01523 (2024-12)
FlexSP: Accelerating Large Language Model Training via Flexible Sequence Parallelism
print_peak_memory 1 other paper
1 dated earlier
2412.01523 (2024-12)
FlexSP: Accelerating Large Language Model Training via Flexible Sequence Parallelism

Presence, not direction: an identical function body in two papers' repositories does not say who wrote it, which came from which, or that either paper's method rests on it. Matching is exact after normalisation, so a renamed or edited function is not matched. Dates are when papers appeared, not when code was written. Only .py files the harvester lifted are covered; notebooks are not read. A new-style arXiv id YYMM.NNNN dates a paper to the month 20YY-MM; otherwise Paper.year dates it to the year; otherwise it is undated. Dates are when the paper appeared, not when the code was written. Read 2026-11-22.

Repositories linked to this paper

Some links come from the archived Papers with Code dataset (CC BY-SA 4.0): attribution and licence.

Abstract

Expert parallelism is vital for effectively training Mixtureof-Experts (MoE) models, enabling different devices to host distinct experts, with each device processing different input data. However, during expert parallel training, dynamic routing results in significant load imbalance among experts: a handful of overloaded experts hinder overall iteration, emerging as a training bottleneck. In this paper, we introduce LAER-MoE, an efficient MoE training framework. The core of LAER-MoE is a novel parallel paradigm, Fully Sharded Expert Parallel (FSEP), which fully partitions each expert parameter by the number of devices and restores partial experts at expert granularity through All-to-All communication during training. This allows for flexible re-layout of expert parameters during training to enhance load balancing. In particular, we perform fine-grained scheduling of communication operations to minimize communication overhead. Additionally, we develop a load balancing planner to formulate re-layout strategies of experts and routing schemes for tokens during training. We perform experiments on an A100 cluster, and the results indicate that our system achieves up to 1.69x acceleration compared to the current state-of-the-art training systems. Source code available at https://github.com/PKU-DAIR/Hetu-Galvatron/tree/laer-moe.

For agents

The same record, over MCP at https://syntology.ai/mcp:

get_harvested_code_for_paper("2602.11686")
get_code_for_paper("2602.11686")
have("2602.11686")

The run record, dated, one paper per request, free:

curl https://syntology.ai/api/ran/2602.11686.json

A badge for a README (the split and the date, never a ratio):

[![Syntology run record](https://syntology.ai/api/ran/2602.11686.svg)](https://syntology.ai/paper/2602.11686)

Connect an agent — have() is free.