SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2203.03466 · NeurIPS · 2022

Tensor Programs V: Tuning Large Neural Networks via Zero-Shot Hyperparameter Transfer

Jianfeng Gao, Edward Hu, Greg Yang, Weizhu Chen, Xiaodong Liu, Igor Babuschkin, Szymon Sidor, David Farhi, Nick Ryder, Jakub Pachocki

arXiv · PDF · Open in the Atlas

Code that ran

3 of 5 samples (functions and classes) lifted from this paper's linked repositories executed on a synthesized input in Syntology's jail (1 of them constructed an object rather than computing a result; 0 honoured the contract we drafted, 0 violated it and 1 has no contract verdict; 2 failures of our instrument; 2 no recorded run), measured 2026-11-21; repository links from the Papers with Code archive attributed under CC BY-SA, no affiliation; 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/2203.03466.json · badge: /api/ran/2203.03466.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
microsoft/mup canonical (classifier label)
classifier basis: classifier_2025
1 ran · 0 no recorded run
vita-group/principled_scaling_lr_init community
Papers with Code archive
1 ran · 1 no recorded run
mu-transformer-authors/mu_transformer community
Papers with Code archive
1 ran · 0 no recorded run
lucaslingle/mu_transformer community
Papers with Code archive
0 ran · 1 no recorded run
FunctionStatusWhere it lives
TransformerConfig Ran
ran on a synthesized input · constructed an object; no result was computed
mu-transformer-authors/mu_transformer/mu_transformer/jax_impl/model.py
code served (permissive licence) · get_code("cc51a0499aef9a80")
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 an instance of its own class, TransformerConfig: the run constructed the object and did not call it, so no result was computed. The image it ran in was not recorded. If your copy returns the same type 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 jax.numpy as jnp
from jax import numpy as jnp
import numpy as np

ARGS = {
    'param_dtype': jnp.float32,
    'dtype': jnp.float32,
    'sequence_len': 128,
    'd_model': 64,
    'd_head': 16,
    'ff_multiple': 4,
    'e_norm': True,
    'q_init': 'normal',
    'r_init': 'normal',
    'u_init': 'normal',
    'qk_scale': 1.0,
    'qk_norm': True,
    'kv_mqa': False,
    'rotary_base': 10000,
    'act_name': 'gelu',
    'act_square': False,
    'norm_eps': 1e-6,
    'norm_gains': True,
    'norm_gains_type': 'learned',
    'proj_biases': True,
    'n_layer': 2,
    'n_vocab': 256,
    'bos_token_id': 0,
    'eos_token_id': 1,
    'pad_token_id': 2,
    'is_train': True,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
flax, jax, numpy — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type TransformerConfig — the run constructed an instance of its own class and did not call it; no shape is stated — the instrument coerces any output to an array; for a TransformerConfig output the recorded ndim 0 and dtype family float describe that coercion, not the output, and are not stated
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
constant_std_init_ Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
microsoft/mup/mup/init.py
code served (permissive licence) · get_code("0c4c2c6a29fb9431")
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 Tensor, ndim 2, dtype family float, shape (4, 3) as a later run recorded it. The image it ran in was not recorded. If your copy returns the same type, ndim and shape 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
import torch.nn as nn

torch.manual_seed(0)

class InfShape:
    def __init__(self, num_inf, width):
        self._num_inf = num_inf
        self._width = width
    
    def ninf(self):
        return self._num_inf
    
    def width_mult(self):
        return self._width

class SamplerCallable:
    def __call__(self, tensor, scale=1.0):
        tensor.data.normal_(0, scale)

tensor = torch.randn(4, 3)
tensor.infshape = InfShape(num_inf=2, width=2.0)

sampler = SamplerCallable()

ARGS = {
    'tensor': tensor,
    'sampler_': sampler
}
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 Tensor, ndim 2, dtype family float, shape (4, 3) — shape (4, 3) as a later run recorded it
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
critic_init_ Ran
ran; our contract draft was wrong, not the code
vita-group/principled_scaling_lr_init/models/utils.py
code served (permissive licence) · get_code("cc4b5a2dc1f8c1e4")
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 Tensor, ndim 2, dtype family float, shape (4, 3) as a later run recorded it. The image it ran in was not recorded. If your copy returns the same type, ndim and shape 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
import torch.nn as nn
import math

torch.manual_seed(0)

# Create a small tensor to initialize
tensor = torch.empty(4, 3)

# Define all parameters
ARGS = {
    'tensor': tensor,
    'a': 0.01,
    'mode': 'fan_in',
    'nonlinearity': 'leaky_relu',
    'degree': 1,
    'std_power': 1
}
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 Tensor, ndim 2, dtype family float, shape (4, 3) — shape (4, 3) as a later run recorded it
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
_init Not yet run
unverified (harvested, no recorded run)
vita-group/principled_scaling_lr_init/models/utils.py
code served (permissive licence) · get_code("c8add3cbe90a58e8")
How it ran

no invocation recorded.

get_rel_mup_scaling Not yet run
unverified (harvested, no recorded run)
lucaslingle/mu_transformer/mu_transformer/jax_impl/launch.py
code served (permissive licence) · get_code("86cf6316062ec680")
How it ran

no invocation recorded.

Also present in other papers' code

Of 5 samples lifted from this paper's linked repositories, 5 carry a function identity and 0 do not. Of the 5 distinct functions with an identity, 0 are also present in other papers whose code we hold (0 with at least one paper dated earlier, 0 the same month, 0 later) and 5 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.

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 2026-11-21.

Repositories linked to this paper

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

Abstract

Hyperparameter (HP) tuning in deep learning is an expensive process, prohibitively so for neural networks (NNs) with billions of parameters. We show that, in the recently discovered Maximal Update Parametrization (µP), many optimal HPs remain stable even as model size changes. This leads to a new HP tuning paradigm we call µTransfer: parametrize the target model in µP, tune the HP indirectly on a smaller model, and zero-shot transfer them to the full-sized model, i.e., without directly tuning the latter at all. We verify µTransfer on Transformer and ResNet. For example, 1) by transferring pretraining HPs from a model of 13M parameters, we outperform published numbers of BERT-large (350M parameters), with a total tuning cost equivalent to pretraining BERT-large once; 2) by transferring from 40M parameters, we outperform published numbers of the 6.7B GPT-3 model, with tuning cost only 7% of total pretraining cost. A Pytorch implementation of our technique can be found at github.com/microsoft/mup and installable via pip install mup.

For agents

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

get_harvested_code_for_paper("2203.03466")
get_code_for_paper("2203.03466")
have("2203.03466")

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

curl https://syntology.ai/api/ran/2203.03466.json

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

[![Syntology run record](https://syntology.ai/api/ran/2203.03466.svg)](https://syntology.ai/paper/2203.03466)

Connect an agent — have() is free.