SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 1909.08053 · 2019

Megatron-LM: Training Multi-Billion Parameter Language Models Using Model Parallelism

Bryan Catanzaro, Mohammad Shoeybi, Mostofa Patwary, Raul Puri, Jared Casper, Patrick Legresley

arXiv · PDF · Open in the Atlas

Code that ran

12 of 47 samples (functions and classes) lifted from this paper's linked repositories executed on a synthesized input in Syntology's jail (2 of them constructed an object rather than computing a result; 4 honoured the contract we drafted, 0 violated it and 3 have no contract verdict; 5 failures of our instrument; 35 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/1909.08053.json · badge: /api/ran/1909.08053.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
facebookresearch/fairscale community
Papers with Code archive
5 ran · 10 no recorded run
qhduan/CPM-LM-TF2 community
Papers with Code archive
4 ran · 15 no recorded run
kingoflolz/mesh-transformer-jax community
Papers with Code archive
3 ran · 10 no recorded run
FunctionStatusWhere it lives
_CopyToModelParallelRegion Ran
ran on a synthesized input · constructed an object; no result was computed
facebookresearch/fairscale/fairscale/nn/model_parallel/layers.py
pointer only (licence: NOASSERTION) · get_code("cf02ea13122d66ef")
How it ran

The body of this sample is withheld: GitHub could not classify the licence of this paper's repository, facebookresearch/fairscale (NOASSERTION). The setup block that called it is withheld with it, because it is written against that body. source_path and origin say where the code lives, to read under the repository's own terms. A run is recorded.

_GatherFromModelParallelRegion Ran
ran on a synthesized input · constructed an object; no result was computed
facebookresearch/fairscale/fairscale/nn/model_parallel/layers.py
pointer only (licence: NOASSERTION) · get_code("0b452ebdf5b90909")
How it ran

The body of this sample is withheld: GitHub could not classify the licence of this paper's repository, facebookresearch/fairscale (NOASSERTION). The setup block that called it is withheld with it, because it is written against that body. source_path and origin say where the code lives, to read under the repository's own terms. A run is recorded.

apply_rotary_pos_emb Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
kingoflolz/mesh-transformer-jax/mesh_transformer/transformer_shard.py
code served (permissive licence) · get_code("3e9321ad8928f17d")
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 ArrayImpl, ndim 3, dtype family float, shape (2, 3, 4) 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 jax.numpy as jnp
import numpy as np

# Create small test tensors
# x: shape (batch, seq_len, dim) = (2, 3, 4)
x = jnp.array(np.random.randn(2, 3, 4), dtype=jnp.float32)

# sincos: tuple of (sin, cos) tensors
# Each should have shape (seq_len, dim//2) = (3, 2) based on how they're used in repeat
sin = jnp.array(np.random.randn(3, 2), dtype=jnp.float32)
cos = jnp.array(np.random.randn(3, 2), dtype=jnp.float32)
sincos = (sin, cos)

ARGS = {
    'x': x,
    'sincos': sincos
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
einops, jax, numpy — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type ArrayImpl, ndim 3, dtype family float, shape (2, 3, 4) — shape (2, 3, 4) 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
copy_to_model_parallel_region Ran
ran, honoured the contract we drafted
facebookresearch/fairscale/fairscale/nn/model_parallel/layers.py
pointer only (licence: NOASSERTION) · get_code("b333830ef6a60a60")
How it ran

The body of this sample is withheld: GitHub could not classify the licence of this paper's repository, facebookresearch/fairscale (NOASSERTION). The setup block that called it is withheld with it, because it is written against that body. source_path and origin say where the code lives, to read under the repository's own terms. A run is recorded.

copy_to_model_parallel_region Ran
ran, honoured the contract we drafted
qhduan/CPM-LM-TF2/CPM-Generate/mpu/transformer.py
code served (permissive licence) · get_code("04bccd9db93a140d")
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) and returned Tensor, ndim 2, 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 torch
import torch.distributed as dist
from unittest.mock import Mock, patch

torch.manual_seed(0)

# Mock the distributed environment since we're CPU-only and not actually running distributed training
def mock_get_world_size(group=None):
    return 1

def mock_all_reduce(tensor, group=None):
    pass

# Patch torch.distributed functions
with patch.object(dist, 'get_world_size', side_effect=mock_get_world_size):
    with patch.object(dist, 'all_reduce', side_effect=mock_all_reduce):
        # Set up the model parallel group
        import sys
        current_module = sys.modules[__name__]
        
        # Create a mock group object
        mock_group = Mock()
        
        # Inject the global variable into the module where the function is defined
        # For this setup, we'll create the necessary context
        _MODEL_PARALLEL_GROUP = mock_group

# Create a small input tensor (batch_size=2, dim=4)
input_tensor = torch.randn(2, 4)

ARGS = {
    'input_': input_tensor
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1
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 — no output shape is stated: observe_signature.py could not observe it (fixture:KeyError)
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
divide Ran
ran, honoured the contract we drafted
qhduan/CPM-LM-TF2/CPM-Generate/mpu/transformer.py
code served (permissive licence) · get_code("33de97c702d46371")
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) and returned int, ndim 0, dtype family int, shape () 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 numpy as np

torch.manual_seed(0)

# The function divide() takes two parameters: numerator and denominator
# It performs modulo and integer division operations on them
# Both should be integers

ARGS = {
    "numerator": 12,
    "denominator": 3
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1
Imports
numpy, torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type int, ndim 0, dtype family int, shape () — shape () 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
divide_and_check_no_remainder Ran
ran, honoured the contract we drafted
facebookresearch/fairscale/fairscale/nn/model_parallel/layers.py
pointer only (licence: NOASSERTION) · get_code("53dcb3e03bcd8ac6")
How it ran

The body of this sample is withheld: GitHub could not classify the licence of this paper's repository, facebookresearch/fairscale (NOASSERTION). The setup block that called it is withheld with it, because it is written against that body. source_path and origin say where the code lives, to read under the repository's own terms. A run is recorded.

fixed_pos_embedding Ran
ran; our contract draft was wrong, not the code
kingoflolz/mesh-transformer-jax/mesh_transformer/transformer_shard.py
code served (permissive licence) · get_code("3d74408951f86d6b")
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 3, 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

# Create a small numpy array with shape that has last dimension 4
# seq_dim=0 means the sequence dimension is axis 0
x = np.random.randn(5, 4)

ARGS = {
    'x': x,
    'seq_dim': 0
}
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 tuple, ndim 3, 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
get_cuda_rng_tracker Ran
ran on a synthesized input
qhduan/CPM-LM-TF2/CPM-Generate/mpu/transformer.py
code served (permissive licence) · get_code("c3a8d35e5b99ad30")
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) and returned CudaRNGStatesTracker. 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 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 get_cuda_rng_tracker() takes no parameters
# It simply returns the global _CUDA_RNG_STATE_TRACKER instance

ARGS = {}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1
Imports
torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type CudaRNGStatesTracker — a CudaRNGStatesTracker output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a CudaRNGStatesTracker 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
rotate_every_two Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
kingoflolz/mesh-transformer-jax/mesh_transformer/transformer_shard.py
code served (permissive licence) · get_code("5efa020fabe4b20b")
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 ArrayImpl, ndim 3, dtype family float, shape (2, 3, 4) 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 jax.numpy as jnp
from einops import rearrange, repeat

# The function rotate_every_two takes a single parameter x
# Looking at the function body:
# - x is indexed as x[:, :, ::2] and x[:, :, 1::2], so it needs at least 3 dimensions
# - The third dimension is sliced with ::2 and 1::2, so it should have even size
# - Output of jnp.stack is rearranged from '... d j -> ... (d j)' format

# Create a small 3D JAX array
# Shape: (batch=2, sequence=3, features=4)
# This satisfies the requirement of 3 dimensions with the last dimension being even
x = jnp.array([
    [[1.0, 2.0, 3.0, 4.0],
     [5.0, 6.0, 7.0, 8.0],
     [9.0, 10.0, 11.0, 12.0]],
    [[13.0, 14.0, 15.0, 16.0],
     [17.0, 18.0, 19.0, 20.0],
     [21.0, 22.0, 23.0, 24.0]]
])

ARGS = {
    'x': x
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
einops, jax — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type ArrayImpl, ndim 3, dtype family float, shape (2, 3, 4) — shape (2, 3, 4) 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
split_tensor_along_last_dim Ran
ran; our contract draft was wrong, not the code
facebookresearch/fairscale/fairscale/nn/model_parallel/layers.py
pointer only (licence: NOASSERTION) · get_code("787ffed12fc28a46")
How it ran

The body of this sample is withheld: GitHub could not classify the licence of this paper's repository, facebookresearch/fairscale (NOASSERTION). The setup block that called it is withheld with it, because it is written against that body. source_path and origin say where the code lives, to read under the repository's own terms. A run is recorded.

split_tensor_along_last_dim Ran
ran; our contract draft was wrong, not the code
qhduan/CPM-LM-TF2/CPM-Generate/mpu/transformer.py
code served (permissive licence) · get_code("c7a00a98e71fad14")
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) and returned tuple, ndim 4, 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 torch

torch.manual_seed(0)

# Create a small tensor with shape (2, 4, 8)
# Last dimension is 8, which is divisible by 2 and 4
tensor = torch.randn(2, 4, 8)

ARGS = {
    "tensor": tensor,
    "num_partitions": 2,
    "contiguous_split_chunks": False,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1
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 4, 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
BertParallelSelfAttention Not yet run
unverified (harvested, no recorded run)
qhduan/CPM-LM-TF2/CPM-Generate/mpu/transformer.py
code served (permissive licence) · get_code("3a3e6cd309d80ca5")
How it ran

no invocation recorded.

CausalTransformerShard Not yet run
unverified (harvested, no recorded run)
kingoflolz/mesh-transformer-jax/mesh_transformer/transformer_shard.py
code served (permissive licence) · get_code("9fafe439d9f687d9")
How it ran

no invocation recorded.

ColumnParallelLinear Not yet run
unverified (harvested, no recorded run)
facebookresearch/fairscale/fairscale/nn/model_parallel/layers.py
pointer only (licence: NOASSERTION) · get_code("764045860ccf534a")
How it ran

The body of this sample is withheld: GitHub could not classify the licence of this paper's repository, facebookresearch/fairscale (NOASSERTION). The setup block that called it is withheld with it, because it is written against that body. source_path and origin say where the code lives, to read under the repository's own terms. No invocation recorded.

ColumnParallelLinear Not yet run
unverified (harvested, no recorded run)
qhduan/CPM-LM-TF2/CPM-Generate/mpu/transformer.py
code served (permissive licence) · get_code("288f7749514cc2bd")
How it ran

no invocation recorded.

CudaRNGStatesTracker Not yet run
unverified (harvested, no recorded run)
qhduan/CPM-LM-TF2/CPM-Generate/mpu/transformer.py
code served (permissive licence) · get_code("fc4bdb2e28201578")
How it ran

no invocation recorded.

EmbeddingShard Not yet run
unverified (harvested, no recorded run)
kingoflolz/mesh-transformer-jax/mesh_transformer/transformer_shard.py
code served (permissive licence) · get_code("176d216b0475ad82")
How it ran

no invocation recorded.

ProjectionShard Not yet run
unverified (harvested, no recorded run)
kingoflolz/mesh-transformer-jax/mesh_transformer/transformer_shard.py
code served (permissive licence) · get_code("2a11a065886e73ae")
How it ran

no invocation recorded.

RMSNorm Not yet run
unverified (harvested, no recorded run)
kingoflolz/mesh-transformer-jax/mesh_transformer/transformer_shard.py
code served (permissive licence) · get_code("ae952da8d1f3ecfc")
How it ran

no invocation recorded.

RelativePositionEmbs Not yet run
unverified (harvested, no recorded run)
kingoflolz/mesh-transformer-jax/mesh_transformer/transformer_shard.py
code served (permissive licence) · get_code("356e9a6604faf993")
How it ran

no invocation recorded.

ReplicatedLayerNorm Not yet run
unverified (harvested, no recorded run)
kingoflolz/mesh-transformer-jax/mesh_transformer/transformer_shard.py
code served (permissive licence) · get_code("4e4deb5427de1770")
How it ran

no invocation recorded.

TransformerLayerShard Not yet run
unverified (harvested, no recorded run)
kingoflolz/mesh-transformer-jax/mesh_transformer/transformer_shard.py
code served (permissive licence) · get_code("f1f29dd1feac61bc")
How it ran

no invocation recorded.

_CopyToModelParallelRegion Not yet run
unverified (harvested, no recorded run)
qhduan/CPM-LM-TF2/CPM-Generate/mpu/transformer.py
code served (permissive licence) · get_code("daaecc67e19dee51")
How it ran

no invocation recorded.

_GatherFromModelParallelRegion Not yet run
unverified (harvested, no recorded run)
qhduan/CPM-LM-TF2/CPM-Generate/mpu/transformer.py
code served (permissive licence) · get_code("d074b0cdd29844e9")
How it ran

no invocation recorded.

_gather Not yet run
unverified (harvested, no recorded run)
facebookresearch/fairscale/fairscale/nn/model_parallel/layers.py
pointer only (licence: NOASSERTION) · get_code("b63c4a93c8fbaec8")
How it ran

The body of this sample is withheld: GitHub could not classify the licence of this paper's repository, facebookresearch/fairscale (NOASSERTION). The setup block that called it is withheld with it, because it is written against that body. source_path and origin say where the code lives, to read under the repository's own terms. No invocation recorded.

_gather Not yet run
unverified (harvested, no recorded run)
qhduan/CPM-LM-TF2/CPM-Generate/mpu/transformer.py
code served (permissive licence) · get_code("e042d40151c8e5b5")
How it ran

no invocation recorded.

_initialize_affine_weight Not yet run
unverified (harvested, no recorded run)
facebookresearch/fairscale/fairscale/nn/model_parallel/layers.py
pointer only (licence: NOASSERTION) · get_code("6c2eb620fc5d8483")
How it ran

The body of this sample is withheld: GitHub could not classify the licence of this paper's repository, facebookresearch/fairscale (NOASSERTION). The setup block that called it is withheld with it, because it is written against that body. source_path and origin say where the code lives, to read under the repository's own terms. No invocation recorded.

_initialize_affine_weight Not yet run
unverified (harvested, no recorded run)
qhduan/CPM-LM-TF2/CPM-Generate/mpu/transformer.py
code served (permissive licence) · get_code("5bdf5f6af94d14be")
How it ran

no invocation recorded.

_reduce Not yet run
unverified (harvested, no recorded run)
facebookresearch/fairscale/fairscale/nn/model_parallel/layers.py
pointer only (licence: NOASSERTION) · get_code("2bae7a1462d53b78")
How it ran

The body of this sample is withheld: GitHub could not classify the licence of this paper's repository, facebookresearch/fairscale (NOASSERTION). The setup block that called it is withheld with it, because it is written against that body. source_path and origin say where the code lives, to read under the repository's own terms. No invocation recorded.

_reduce Not yet run
unverified (harvested, no recorded run)
qhduan/CPM-LM-TF2/CPM-Generate/mpu/transformer.py
code served (permissive licence) · get_code("a8bcff2e094dfd4a")
How it ran

no invocation recorded.

_set_cuda_rng_state Not yet run
unverified (harvested, no recorded run)
qhduan/CPM-LM-TF2/CPM-Generate/mpu/transformer.py
code served (permissive licence) · get_code("26c4d45293254bda")
How it ran

no invocation recorded.

_split Not yet run
unverified (harvested, no recorded run)
facebookresearch/fairscale/fairscale/nn/model_parallel/layers.py
pointer only (licence: NOASSERTION) · get_code("e798817bd7481060")
How it ran

The body of this sample is withheld: GitHub could not classify the licence of this paper's repository, facebookresearch/fairscale (NOASSERTION). The setup block that called it is withheld with it, because it is written against that body. source_path and origin say where the code lives, to read under the repository's own terms. No invocation recorded.

_split Not yet run
unverified (harvested, no recorded run)
qhduan/CPM-LM-TF2/CPM-Generate/mpu/transformer.py
code served (permissive licence) · get_code("93b792270121fe82")
How it ran

no invocation recorded.

ensure_divisibility Not yet run
unverified (harvested, no recorded run)
facebookresearch/fairscale/fairscale/nn/model_parallel/layers.py
pointer only (licence: NOASSERTION) · get_code("3bd26074ee7c9889")
How it ran

The body of this sample is withheld: GitHub could not classify the licence of this paper's repository, facebookresearch/fairscale (NOASSERTION). The setup block that called it is withheld with it, because it is written against that body. source_path and origin say where the code lives, to read under the repository's own terms. No invocation recorded.

ensure_divisibility Not yet run
unverified (harvested, no recorded run)
qhduan/CPM-LM-TF2/CPM-Generate/mpu/transformer.py
code served (permissive licence) · get_code("158e905c29f23472")
How it ran

no invocation recorded.

f_psum Not yet run
unverified (harvested, no recorded run)
kingoflolz/mesh-transformer-jax/mesh_transformer/transformer_shard.py
code served (permissive licence) · get_code("b5f626934ffb18b4")
How it ran

no invocation recorded.

g_psum Not yet run
unverified (harvested, no recorded run)
kingoflolz/mesh-transformer-jax/mesh_transformer/transformer_shard.py
code served (permissive licence) · get_code("0687254bf74e83b3")
How it ran

no invocation recorded.

gather_from_model_parallel_region Not yet run
unverified (harvested, no recorded run)
facebookresearch/fairscale/fairscale/nn/model_parallel/layers.py
pointer only (licence: NOASSERTION) · get_code("3e7b5da4deb87f39")
How it ran

The body of this sample is withheld: GitHub could not classify the licence of this paper's repository, facebookresearch/fairscale (NOASSERTION). The setup block that called it is withheld with it, because it is written against that body. source_path and origin say where the code lives, to read under the repository's own terms. No invocation recorded.

gather_from_model_parallel_region Not yet run
unverified (harvested, no recorded run)
qhduan/CPM-LM-TF2/CPM-Generate/mpu/transformer.py
code served (permissive licence) · get_code("f9991e714de95be5")
How it ran

no invocation recorded.

get_model_parallel_group Not yet run
unverified (harvested, no recorded run)
facebookresearch/fairscale/fairscale/nn/model_parallel/layers.py
pointer only (licence: NOASSERTION) · get_code("2929645262f955ef")
How it ran

The body of this sample is withheld: GitHub could not classify the licence of this paper's repository, facebookresearch/fairscale (NOASSERTION). The setup block that called it is withheld with it, because it is written against that body. source_path and origin say where the code lives, to read under the repository's own terms. No invocation recorded.

get_model_parallel_group Not yet run
unverified (harvested, no recorded run)
qhduan/CPM-LM-TF2/CPM-Generate/mpu/transformer.py
code served (permissive licence) · get_code("6bb3803691d991da")
How it ran

no invocation recorded.

get_model_parallel_rank Not yet run
unverified (harvested, no recorded run)
facebookresearch/fairscale/fairscale/nn/model_parallel/layers.py
pointer only (licence: NOASSERTION) · get_code("cc03593755c3ba09")
How it ran

The body of this sample is withheld: GitHub could not classify the licence of this paper's repository, facebookresearch/fairscale (NOASSERTION). The setup block that called it is withheld with it, because it is written against that body. source_path and origin say where the code lives, to read under the repository's own terms. No invocation recorded.

get_model_parallel_rank Not yet run
unverified (harvested, no recorded run)
qhduan/CPM-LM-TF2/CPM-Generate/mpu/transformer.py
code served (permissive licence) · get_code("f6251a861544b65e")
How it ran

no invocation recorded.

get_model_parallel_world_size Not yet run
unverified (harvested, no recorded run)
facebookresearch/fairscale/fairscale/nn/model_parallel/layers.py
pointer only (licence: NOASSERTION) · get_code("e2db563887482b11")
How it ran

The body of this sample is withheld: GitHub could not classify the licence of this paper's repository, facebookresearch/fairscale (NOASSERTION). The setup block that called it is withheld with it, because it is written against that body. source_path and origin say where the code lives, to read under the repository's own terms. No invocation recorded.

get_model_parallel_world_size Not yet run
unverified (harvested, no recorded run)
qhduan/CPM-LM-TF2/CPM-Generate/mpu/transformer.py
code served (permissive licence) · get_code("e40c745e214477b2")
How it ran

no invocation recorded.

getnorm Not yet run
unverified (harvested, no recorded run)
kingoflolz/mesh-transformer-jax/mesh_transformer/transformer_shard.py
code served (permissive licence) · get_code("e905d7fbdd56818f")
How it ran

no invocation recorded.

Also present in other papers' code

Of 47 samples lifted from this paper's linked repositories, 47 carry a function identity and 0 do not. Of the 46 distinct functions with an identity, 7 are also present in other papers whose code we hold (0 with at least one paper dated earlier, 0 the same month, 7 later) and 39 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
divide
brick of this paper's assembly: helper
4 other papers
4 later
2209.02970 (2022-09)
Fengshenbang 1.0: Being the Foundation of Chinese Cognitive Intelligence
2404.03608 (2024-04)
Sailor: Open Language Models for South-East Asia
2406.11257 (2024-06)
ExCP: Extreme LLM Checkpoint Compression via Weight-Momentum Joint Shrinking
and 1 more not named here
CudaRNGStatesTracker
brick of this paper's assembly: helper
1 other paper
1 later
2409.17481 (2024-09)
MaskLLM: Learnable Semi-Structured Sparsity for Large Language Models
_set_cuda_rng_state
brick of this paper's assembly: helper
1 other paper
1 later
2409.17481 (2024-09)
MaskLLM: Learnable Semi-Structured Sparsity for Large Language Models
apply_rotary_pos_emb
brick of this paper's assembly: helper
1 other paper
1 later · 1 only as an assembly brick
2104.09864 (2021-04)
RoFormer: Enhanced Transformer with Rotary Position Embedding
get_cuda_rng_tracker
brick of this paper's assembly: helper
1 other paper
1 later
2409.17481 (2024-09)
MaskLLM: Learnable Semi-Structured Sparsity for Large Language Models
rotate_every_two
brick of this paper's assembly: helper
1 other paper
1 later · 1 only as an assembly brick
2104.09864 (2021-04)
RoFormer: Enhanced Transformer with Rotary Position Embedding
split_tensor_along_last_dim
brick of this paper's assembly: helper
1 other paper
1 later
2406.11257 (2024-06)
ExCP: Extreme LLM Checkpoint Compression via Weight-Momentum Joint Shrinking

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

Recent work in language modeling demonstrates that training large transformer models advances the state of the art in Natural Language Processing applications. However, very large models can be quite difficult to train due to memory constraints. In this work, we present our techniques for training very large transformer models and implement a simple, efficient intra-layer model parallel approach that enables training transformer models with billions of parameters. Our approach does not require a new compiler or library changes, is orthogonal and complimentary to pipeline model parallelism, and can be fully implemented with the insertion of a few communication operations in native PyTorch. We illustrate this approach by converging transformer based models up to 8.3 billion parameters using 512 GPUs. We sustain 15.1 PetaFLOPs across the entire application with 76% scaling efficiency when compared to a strong single GPU baseline that sustains 39 TeraFLOPs, which is 30% of peak FLOPs. To demonstrate that large language models can further advance the state of the art (SOTA), we train an 8.3 billion parameter transformer language model similar to GPT-2 and a 3.9 billion parameter model similar to BERT. We show that careful attention to the placement of layer normalization in BERT-like models is critical to achieving increased performance as the model size grows. Using the GPT-2 model we achieve SOTA results on the WikiText103 (10.8 compared to SOTA perplexity of 15.8) and LAMBADA (66.5% compared to SOTA accuracy of 63.2%) datasets. Our BERT model achieves SOTA results on the RACE dataset (90.9% compared to SOTA accuracy of 89.4%).

For agents

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

get_harvested_code_for_paper("1909.08053")
get_code_for_paper("1909.08053")
have("1909.08053")

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

curl https://syntology.ai/api/ran/1909.08053.json

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

[![Syntology run record](https://syntology.ai/api/ran/1909.08053.svg)](https://syntology.ai/paper/1909.08053)

Connect an agent — have() is free.