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

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

17 Sep 2019arXiv:1909.08053archive 2025-07-28

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

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%).

PaperPDFCodeCode Syntology ranMethod as code

In Syntology View this paper on Syntology, its page in Syntology's graph. That page adds which of its functions are also present in other papers' code Syntology holds, with dates, and its dated run record as JSON and a README badge. It also lists the repositories linked to the paper, the abstract and the calls for agents.

Open this paper in Syntology's Atlas, the map of the papers in Syntology's graph and their citations.

For agents, on Syntology's MCP service (how to connect):

Code

No sample from the archive's official repository (NVIDIA/Megatron-LM) is linked to this paper; the 47 samples below come from 3 community repositories. Syntology harvested code from NVIDIA/Megatron-LM for 2 other papers.

NVIDIA/Megatron-LM: Syntology's graph links these samples to Efficient Large-Scale Language Model Training on GPU Clusters Using Megatron-LM (arXiv:2104.04473; that paper's own run record: 0 ran · 1 unverified) and EnergyLens: Predictive Energy-Aware Exploration for Multi-GPU LLM Inference Optimization (arXiv:2605.14249; that paper's own run record: 6 ran (of which 1 constructed an object rather than computing a result; 3 with no instrument failure: 0 honoured, 1 violated, 2 with no contract checked; 3 where Syntology's instrument failed) · 0 unverified); nothing checks that they implement this paper's method.

Syntology 47 code samples harvested from 3 repositories linked to this paper: 12 ran (of which 2 constructed an object rather than computing a result; 7 with no instrument failure: 4 honoured, 0 violated, 3 with no contract checked; 5 where Syntology's instrument failed) · 35 unverified. Where Syntology's instrument failed: 3 where our contract draft was wrong; 2 where the contract check could not run; failures of Syntology's instrument, not of the code.

By repository: community (archive-listed): 47 samples from 3 repositories, 12 ran (7 with no instrument failure; 5 where Syntology's instrument failed). The run record, sample by sample. “Ran” means executed on a synthesized input, not that the code is correct or reproduces the paper.

No sample is linked to this paper from 6 of its repositories that Syntology harvested for other papers:

Levels written on 2026-09-22 (UTC): the dates Syntology's graph wrote each sample's verification level, not when the code executed.

NVIDIA/Megatron-LM officialmentioned in papermentioned on GitHubpytorchSyntology: no sample linked to this paper (harvested for another paper). Syntology's graph links these samples to Efficient Large-Scale Language Model Training on GPU Clusters Using Megatron-LM (arXiv:2104.04473; that paper's own run record: 0 ran · 1 unverified) and EnergyLens: Predictive Energy-Aware Exploration for Multi-GPU LLM Inference Optimization (arXiv:2605.14249; that paper's own run record: 6 ran (of which 1 constructed an object rather than computing a result; 3 with no instrument failure: 0 honoured, 1 violated, 2 with no contract checked; 3 where Syntology's instrument failed) · 0 unverified); nothing checks that they implement this paper's method. report
THUDM/ProteinLM mentioned on GitHubpytorchApache-2.0Syntology: no sample linked to this paper (harvested for another paper). Syntology's graph links these samples to Modeling Protein Using Large-scale Pretrain Language Model (arXiv:2108.07435; that paper's own run record: 3 ran (of which 0 constructed an object rather than computing a result; 3 with no instrument failure: 0 honoured, 0 violated, 3 with no contract checked; 0 where Syntology's instrument failed) · 1 unverified); nothing checks that they implement this paper's method. report
ezelikman/STaR mentioned on GitHubjaxSyntology: no sample linked to this paper (harvested for another paper). Syntology's graph links these samples to STaR: Bootstrapping Reasoning With Reasoning (arXiv:2203.14465; that paper's own run record: 0 ran · 2 unverified); nothing checks that they implement this paper's method. report
facebookresearch/fairscale mentioned on GitHubpytorchSyntology: 15 samples, 5 ran (1 instrument-failed) · 10 unverified report
kingoflolz/mesh-transformer-jax mentioned on GitHubjaxSyntology: 13 samples, 3 ran (3 instrument-failed) · 10 unverified report
nvidia/transformerengine mentioned on GitHubjaxApache-2.0Syntology: no sample linked to this paper (harvested for another paper). Syntology's graph links these samples to FP8 Formats for Deep Learning (arXiv:2209.05433; that paper's own run record: 6 ran (of which 0 constructed an object rather than computing a result; 6 with no instrument failure: 0 honoured, 0 violated, 6 with no contract checked; 0 where Syntology's instrument failed) · 1 unverified); nothing checks that they implement this paper's method. report
qhduan/CPM-LM-TF2 mentioned on GitHubpytorchSyntology: 19 samples, 4 ran (1 instrument-failed) · 15 unverified report
Beacontownfc/MegatronBert_paddle paddleSyntology: not harvested report
PaddlePaddle/FleetX paddleSyntology: no sample linked to this paper (harvested for another paper). Syntology's graph links these samples to MoESys: A Distributed and Efficient Mixture-of-Experts Training and Inference System... (arXiv:2205.10034; that paper's own run record: 1 ran (of which 0 constructed an object rather than computing a result; 1 with no instrument failure: 0 honoured, 0 violated, 1 with no contract checked; 0 where Syntology's instrument failed) · 0 unverified); nothing checks that they implement this paper's method. report
PaddlePaddle/PaddleNLP paddleSyntology: no sample linked to this paper (harvested for another paper). Syntology's graph links these samples to ERNIE 3.0 Tiny: Frustratingly Simple Method to Improve Task-Agnostic Distillation... (arXiv:2301.03416; that paper's own run record: 2 ran (of which 0 constructed an object rather than computing a result; 2 with no instrument failure: 0 honoured, 0 violated, 2 with no contract checked; 0 where Syntology's instrument failed) · 1 unverified); nothing checks that they implement this paper's method. report

Repository list and official/mentioned flags are the archive's, frozen 2025-07-28. Reachability, where shown, is from one Syntology probe window (2026-09-16 to 2026-09-18); repositories not probed show nothing. GitHub stars are not tracked.

A paper named beside a repository with no sample linked to this paper is shown with that paper's own run record, not this paper's: “ran” means executed on a synthesized input, not that the code is correct, and “where Syntology's instrument failed” counts failures of Syntology's instrument, not of the code. Papers are listed in arXiv-id order, at most three per repository.

Code Syntology ran Syntology

47 samples harvested: 12 ran (of which 2 constructed an object rather than computing a result; 7 with no instrument failure: 4 honoured, 0 violated, 3 with no contract checked; 5 where Syntology's instrument failed) · 35 unverified. “Where Syntology's instrument failed” counts failures of Syntology's instrument, not of the code. Syntology's record for this page has not changed since , the first build that kept a record date for it; when this build read Syntology's graph is in the build record. That date is not when the samples ran.

4ran · honoured contract
3ran · our draft was wrong
2ran; the contract check could not run
3ran
35unverified

Licence: 15 of the 47 samples are pointer only, meaning Syntology does not serve that copy's text. This page shows no code text for any sample; each one links to its file in the repository.

Harvested from 3 repositories linked to this paper, official or community; the list below is grouped by repository, each group headed by its name and what it is to the paper. “Ran” means the sample executed on a synthesized input. It does not mean the output is correct, and nothing here reproduces the paper's results. A line beginning “fixture:” under a sample that ran gives the arguments Syntology's fixture called the sample with, each as name:type/rank/element kind (? where none was recorded), and the type of what the fixture's run returned, exactly as the harness recorded them; the arguments and the returned type were recorded by two separate runs of the same fixture. No output value is shown, and a returned type says nothing about whether the output was right. “Level recorded” gives the UTC date when Syntology's graph wrote the sample's verification level, not when the code executed; where the level was written before Syntology's graph kept that date, the row says “level recorded, date not recorded”. “Honoured” and “violated” refer to a contract Syntology drafted from the code itself. “Our draft was wrong” and “the contract check could not run” are failures of Syntology's instrument, not of the code: the sample ran on Syntology's synthesized input, and then the drafted contract did not describe it, or the check itself raised an error (RAISES), lacked a dependency (DEP_MISSING) or ran out of time (TIMEOUT); hover the label for which. A metamorphic tier, where shown, is evidence of the output's shape, not that the sample implements the method.

Each sample ends with its code_sha256. A sample's code_sha256 is Syntology's identity for that exact code. It is printed in full (16 hexadecimal characters for every sample in this build), and get_code(code_sha256="…") takes it exactly as printed. An agent calls it on Syntology's MCP service (how to connect); click an identity to copy that call.

Repository labels, per sample. official repository: The archive marks this repository official for the paper. named in the paper: The archive records that the paper mentions this repository; it is not marked official. community (archive-listed): In the archive's code links for this paper, not marked official and not recorded as mentioned in the paper. found in paper text by Syntology: Syntology found this repository in the paper's own text; whether it is the authors' implementation is not asserted. community: Not in the archive's code links for this paper; a community repository Syntology harvested. Samples are grouped by repository: a repository marked official first, then by how directly the repository is tied to the paper, then in the archive's order; never by how many ran. Licence labels name the repository's licence as recorded at harvest. “Pointer only” means Syntology does not serve that copy's text, for one of six reasons: no licence file was found; the licence was not identified; the licence is recorded as permissive but that copy's record is not marked cleared; the licence is outside the permissive list Syntology serves text under (MIT, Apache-2.0, BSD and similar); this paper's copy was not located, so its licence is not recorded; or this paper's copy is cleared, but the stored copy of the same code that Syntology serves is not. Some licences outside that list permit redistribution, such as WTFPL, and GPL-3.0 under its conditions; they are simply not on the list. Hover a licence label for the reason. File links open the file on GitHub at the default branch, which may have changed since the harvest.

A function or class name links to its page when the name occurs in the harvested code of at least 5 papers; the name is a string, not an identity. A census tag appears when Syntology's census ran implementations with that name on one shared input per bucket: it gives how many were compared, from how many papers, and how many distinct outputs they produced, and names this sample's bucket when this sample was one of them.

facebookresearch/fairscale community (archive-listed) 15 samples, 5 ran (1 instrument-failed) · 10 unverified
_CopyToModelParallelRegion fairscale/nn/model_parallel/layers.py ran licence not identified · pointer only · cf02ea13122d66ef · report fixture: returned _CopyToModelParallelRegion · level recorded 2026-09-22
_GatherFromModelParallelRegion fairscale/nn/model_parallel/layers.py ran licence not identified · pointer only · 0b452ebdf5b90909 · report fixture: returned _GatherFromModelParallelRegion · level recorded 2026-09-22
copy_to_model_parallel_region fairscale/nn/model_parallel/layers.py ran · honoured contract licence not identified · pointer only · b333830ef6a60a60 · report fixture: returned Tensor · level recorded 2026-09-22
divide_and_check_no_remainder fairscale/nn/model_parallel/layers.py ran · honoured contract fingerprinted licence not identified · pointer only · 53dcb3e03bcd8ac6 · report fixture: signature numerator:int/0/int, denominator:int/0/int · returned int · level recorded 2026-09-22
split_tensor_along_last_dim in 12 papers fairscale/nn/model_parallel/layers.py ran · our draft was wrong fingerprinted licence not identified · pointer only · 787ffed12fc28a46 · report fixture: signature tensor:Tensor/2/float, num_partitions:int/0/int, contiguous_split_chunks:bool/0/bool · returned tuple · level recorded 2026-09-22
ColumnParallelLinear fairscale/nn/model_parallel/layers.py unverified licence not identified · pointer only · 764045860ccf534a · report
_gather fairscale/nn/model_parallel/layers.py unverified licence not identified · pointer only · b63c4a93c8fbaec8 · report
_initialize_affine_weight fairscale/nn/model_parallel/layers.py unverified licence not identified · pointer only · 6c2eb620fc5d8483 · report
_reduce fairscale/nn/model_parallel/layers.py unverified licence not identified · pointer only · 2bae7a1462d53b78 · report
_split fairscale/nn/model_parallel/layers.py unverified licence not identified · pointer only · e798817bd7481060 · report
ensure_divisibility fairscale/nn/model_parallel/layers.py unverified licence not identified · pointer only · 3bd26074ee7c9889 · report
gather_from_model_parallel_region fairscale/nn/model_parallel/layers.py unverified licence not identified · pointer only · 3e7b5da4deb87f39 · report
get_model_parallel_group fairscale/nn/model_parallel/layers.py unverified licence not identified · pointer only · 2929645262f955ef · report
get_model_parallel_rank fairscale/nn/model_parallel/layers.py unverified licence not identified · pointer only · cc03593755c3ba09 · report
get_model_parallel_world_size fairscale/nn/model_parallel/layers.py unverified licence not identified · pointer only · e2db563887482b11 · report
kingoflolz/mesh-transformer-jax community (archive-listed) 13 samples, 3 ran (3 instrument-failed) · 10 unverified
apply_rotary_pos_emb in 393 papers
census: 14 implementations from 15 papers, 8 distinct outputs across 7 buckets, one shared input per bucket
mesh_transformer/transformer_shard.py ran; the contract check could not run (RAISES) · metamorphic tier: invariant Apache-2.0 (permissive) · 3e9321ad8928f17d · report fixture: signature x:ArrayImpl/3/float, sincos:tuple/1/ArrayImpl · returned ArrayImpl · level recorded 2026-09-22
fixed_pos_embedding in 16 papers mesh_transformer/transformer_shard.py ran · our draft was wrong fingerprinted Apache-2.0 (permissive) · 3d74408951f86d6b · report fixture: signature x:ndarray/2/float, seq_dim:int/0/int · returned tuple · level recorded 2026-09-22
rotate_every_two in 33 papers
census: 5 implementations from 8 papers, 2 distinct outputs across 2 buckets, one shared input per bucket
mesh_transformer/transformer_shard.py ran; the contract check could not run (RAISES) · metamorphic tier: invariant fingerprinted Apache-2.0 (permissive) · 5efa020fabe4b20b · report fixture: signature x:ArrayImpl/3/float · returned ArrayImpl · level recorded 2026-09-22
CausalTransformerShard mesh_transformer/transformer_shard.py unverified Apache-2.0 (permissive) · 9fafe439d9f687d9 · report
EmbeddingShard mesh_transformer/transformer_shard.py unverified Apache-2.0 (permissive) · 176d216b0475ad82 · report
ProjectionShard mesh_transformer/transformer_shard.py unverified Apache-2.0 (permissive) · 2a11a065886e73ae · report
RMSNorm in 61 papers mesh_transformer/transformer_shard.py unverified Apache-2.0 (permissive) · ae952da8d1f3ecfc · report
RelativePositionEmbs mesh_transformer/transformer_shard.py unverified Apache-2.0 (permissive) · 356e9a6604faf993 · report
ReplicatedLayerNorm mesh_transformer/transformer_shard.py unverified Apache-2.0 (permissive) · 4e4deb5427de1770 · report
TransformerLayerShard mesh_transformer/transformer_shard.py unverified Apache-2.0 (permissive) · f1f29dd1feac61bc · report
f_psum mesh_transformer/transformer_shard.py unverified Apache-2.0 (permissive) · b5f626934ffb18b4 · report
g_psum mesh_transformer/transformer_shard.py unverified Apache-2.0 (permissive) · 0687254bf74e83b3 · report
getnorm mesh_transformer/transformer_shard.py unverified Apache-2.0 (permissive) · e905d7fbdd56818f · report
qhduan/CPM-LM-TF2 community (archive-listed) 19 samples, 4 ran (1 instrument-failed) · 15 unverified
copy_to_model_parallel_region CPM-Generate/mpu/transformer.py ran · honoured contract MIT (permissive) · 04bccd9db93a140d · report fixture: returned Tensor · level recorded 2026-09-22
divide in 14 papers CPM-Generate/mpu/transformer.py ran · honoured contract fingerprinted MIT (permissive) · 33de97c702d46371 · report fixture: signature numerator:int/0/int, denominator:int/0/int · returned int · level recorded 2026-09-22
get_cuda_rng_tracker CPM-Generate/mpu/transformer.py ran MIT (permissive) · c3a8d35e5b99ad30 · report fixture: returned CudaRNGStatesTracker · level recorded 2026-09-22
split_tensor_along_last_dim in 12 papers CPM-Generate/mpu/transformer.py ran · our draft was wrong MIT (permissive) · c7a00a98e71fad14 · report fixture: signature tensor:Tensor/3/float, num_partitions:int/0/int, contiguous_split_chunks:bool/0/bool · returned tuple · level recorded 2026-09-22
BertParallelSelfAttention CPM-Generate/mpu/transformer.py unverified MIT (permissive) · 3a3e6cd309d80ca5 · report
ColumnParallelLinear CPM-Generate/mpu/transformer.py unverified MIT (permissive) · 288f7749514cc2bd · report
CudaRNGStatesTracker CPM-Generate/mpu/transformer.py unverified MIT (permissive) · fc4bdb2e28201578 · report
_CopyToModelParallelRegion CPM-Generate/mpu/transformer.py unverified MIT (permissive) · daaecc67e19dee51 · report
_GatherFromModelParallelRegion CPM-Generate/mpu/transformer.py unverified MIT (permissive) · d074b0cdd29844e9 · report
_gather CPM-Generate/mpu/transformer.py unverified MIT (permissive) · e042d40151c8e5b5 · report
_initialize_affine_weight CPM-Generate/mpu/transformer.py unverified MIT (permissive) · 5bdf5f6af94d14be · report
_reduce CPM-Generate/mpu/transformer.py unverified MIT (permissive) · a8bcff2e094dfd4a · report
_set_cuda_rng_state CPM-Generate/mpu/transformer.py unverified MIT (permissive) · 26c4d45293254bda · report
_split CPM-Generate/mpu/transformer.py unverified MIT (permissive) · 93b792270121fe82 · report
ensure_divisibility CPM-Generate/mpu/transformer.py unverified MIT (permissive) · 158e905c29f23472 · report
gather_from_model_parallel_region CPM-Generate/mpu/transformer.py unverified MIT (permissive) · f9991e714de95be5 · report
get_model_parallel_group CPM-Generate/mpu/transformer.py unverified MIT (permissive) · 6bb3803691d991da · report
get_model_parallel_rank CPM-Generate/mpu/transformer.py unverified MIT (permissive) · f6251a861544b65e · report
get_model_parallel_world_size CPM-Generate/mpu/transformer.py unverified MIT (permissive) · e40c745e214477b2 · report

Method as code Syntology

What Syntology's graph records as this paper's claimed method, lifted into code: each assembly names its bricks (functions and classes), each with its own role and its own verification level. An assembly is not a check that the code implements the method correctly or reproduces the paper's results. Levels are each brick's own verification level in Syntology's graph (level 0: no recorded run), listed brick by brick and never combined into one level; “ran” means the brick executed on a synthesized input, not that its output is correct, and the instrument figure counts failures of Syntology's instrument, not of the code. Where a model read the code against the paper's claim, each assembly's label verdict is one model's judgement, kept in this page's JSON twin; it is not calibrated, so no card prints it. Siblings are assemblies lifted from this paper in the same batch for the same method: alternative readings kept side by side; none replaces or supersedes another.

3 assemblies on this paper: 3 harvested, 0 generated, listed in 1 groups (siblings lifted in one batch form one group).

3 sibling assemblies for Megatron-LM (harvested), lifted in the same batch on 2026-09-09: alternative readings kept side by side; none replaces or supersedes another.

harvested Assembly for Megatron-LM (graph method: Megatron-LM). 13 bricks; individually on a fixture: 3 ran (of which 0 constructed an object rather than computing a result; 0 with no instrument failure: 0 honoured, 0 violated, 0 with no contract checked; 3 where Syntology's instrument failed) · 10 unverified. The assembly has not been run end-to-end.

one of 3 siblings

  1. CausalTransformerShard entry · level 0 · unverified · 9fafe439d9f687d9
  2. EmbeddingShard model · level 0 · unverified · 176d216b0475ad82
  3. ProjectionShard model · level 0 · unverified · 2a11a065886e73ae
  4. RelativePositionEmbs helper · level 0 · unverified · 356e9a6604faf993
  5. TransformerLayerShard model · level 0 · unverified · f1f29dd1feac61bc
  6. g_psum helper · level 0 · unverified · 0687254bf74e83b3
  7. f_psum helper · level 0 · unverified · b5f626934ffb18b4
  8. getnorm helper · level 0 · unverified · e905d7fbdd56818f
  9. apply_rotary_pos_emb helper · level 1 · ran; the contract check could not run (RAISES) · 3e9321ad8928f17d
  10. fixed_pos_embedding helper · level 1 · ran · our draft was wrong · 3d74408951f86d6b
  11. RMSNorm helper · level 0 · unverified · ae952da8d1f3ecfc
  12. ReplicatedLayerNorm helper · level 0 · unverified · 4e4deb5427de1770
  13. rotate_every_two helper · level 1 · ran; the contract check could not run (RAISES) · 5efa020fabe4b20b

harvested Assembly for Megatron-LM (graph method: Megatron-LM). 15 bricks; individually on a fixture: 5 ran (of which 2 constructed an object rather than computing a result; 4 with no instrument failure: 2 honoured, 0 violated, 2 with no contract checked; 1 where Syntology's instrument failed) · 10 unverified. The assembly has not been run end-to-end.

one of 3 siblings

  1. ColumnParallelLinear entry · level 0 · unverified · 764045860ccf534a
  2. _initialize_affine_weight helper · level 0 · unverified · 6c2eb620fc5d8483
  3. copy_to_model_parallel_region helper · level 1 · ran · honoured contract · b333830ef6a60a60
  4. divide_and_check_no_remainder helper · level 1 · ran · honoured contract · 53dcb3e03bcd8ac6
  5. gather_from_model_parallel_region helper · level 0 · unverified · 3e7b5da4deb87f39
  6. get_model_parallel_world_size helper · level 0 · unverified · e2db563887482b11
  7. get_model_parallel_rank helper · level 0 · unverified · cc03593755c3ba09
  8. _CopyToModelParallelRegion helper · level 1 · ran · cf02ea13122d66ef
  9. ensure_divisibility helper · level 0 · unverified · 3bd26074ee7c9889
  10. _GatherFromModelParallelRegion helper · level 1 · ran · 0b452ebdf5b90909
  11. get_model_parallel_group helper · level 0 · unverified · 2929645262f955ef
  12. _reduce helper · level 0 · unverified · 2bae7a1462d53b78
  13. _gather helper · level 0 · unverified · b63c4a93c8fbaec8
  14. _split helper · level 0 · unverified · e798817bd7481060
  15. split_tensor_along_last_dim helper · level 1 · ran · our draft was wrong · 787ffed12fc28a46

harvested Assembly for Megatron-LM (graph method: Megatron-LM). 19 bricks; individually on a fixture: 4 ran (of which 0 constructed an object rather than computing a result; 3 with no instrument failure: 2 honoured, 0 violated, 1 with no contract checked; 1 where Syntology's instrument failed) · 15 unverified. The assembly has not been run end-to-end.

one of 3 siblings

  1. BertParallelSelfAttention entry · level 0 · unverified · 3a3e6cd309d80ca5
  2. ColumnParallelLinear helper · level 0 · unverified · 288f7749514cc2bd
  3. divide helper · level 1 · ran · honoured contract · 33de97c702d46371
  4. gather_from_model_parallel_region helper · level 0 · unverified · f9991e714de95be5
  5. get_cuda_rng_tracker helper · level 1 · ran · c3a8d35e5b99ad30
  6. get_model_parallel_world_size helper · level 0 · unverified · e40c745e214477b2
  7. split_tensor_along_last_dim helper · level 1 · ran · our draft was wrong · c7a00a98e71fad14
  8. _initialize_affine_weight helper · level 0 · unverified · 5bdf5f6af94d14be
  9. copy_to_model_parallel_region helper · level 1 · ran · honoured contract · 04bccd9db93a140d
  10. ensure_divisibility helper · level 0 · unverified · 158e905c29f23472
  11. _GatherFromModelParallelRegion helper · level 0 · unverified · d074b0cdd29844e9
  12. get_model_parallel_group helper · level 0 · unverified · 6bb3803691d991da
  13. get_model_parallel_rank helper · level 0 · unverified · f6251a861544b65e
  14. _CopyToModelParallelRegion helper · level 0 · unverified · daaecc67e19dee51
  15. _gather helper · level 0 · unverified · e042d40151c8e5b5
  16. _split helper · level 0 · unverified · 93b792270121fe82
  17. _reduce helper · level 0 · unverified · a8bcff2e094dfd4a
  18. _set_cuda_rng_state helper · level 0 · unverified · 26c4d45293254bda
  19. CudaRNGStatesTracker helper · level 0 · unverified · fc4bdb2e28201578

Each brick ends with its code_sha256, Syntology's identity for that exact code; click one to copy the get_code call that returns it. For agents, on Syntology's MCP service: assembly(arxiv_id="1909.08053") returns these assemblies with each brick's role and wiring (how to connect).

Tasks

LAMBADALanguage ModelingLanguage ModellingQuestion AnsweringReading Comprehension

1 archive task tag without a task page not shown.

Results from the paper archive 2025-07-28

TaskDatasetModelMetricValueRank at snapshotLeaderboardReport
Language Modelling WikiText-103 Megatron-LM Number of params 8300M #3 of 89 Archive leaderboard report
Language Modelling WikiText-103 Megatron-LM Test perplexity 10.81 #3 of 89 Archive leaderboard report
Question Answering PIQA MT-NLG 530B (0-shot) Accuracy 82.0 #23 of 67 Archive leaderboard report
Reading Comprehension RACE Megatron-BERT (ensemble) Accuracy 90.9 #2 of 24 Archive leaderboard report
Reading Comprehension RACE Megatron-BERT (ensemble) Accuracy (High) 90.0 #2 of 24 Archive leaderboard report
Reading Comprehension RACE Megatron-BERT (ensemble) Accuracy (Middle) 93.1 #2 of 24 Archive leaderboard report
Reading Comprehension RACE Megatron-BERT Accuracy 89.5 #4 of 24 Archive leaderboard report
Reading Comprehension RACE Megatron-BERT Accuracy (High) 88.6 #4 of 24 Archive leaderboard report
Reading Comprehension RACE Megatron-BERT Accuracy (Middle) 91.8 #4 of 24 Archive leaderboard report

Ranks are positions in the archive's leaderboards as they stood at the 2025-07-28 snapshot. Results published since then are not among these rows, so a rank here is not a current standing.

Methods

Absolute Position EncodingsAdamAttentionAttention DropoutBERTBPECosine AnnealingDense ConnectionsDiscriminative Fine-TuningDropoutGPT-2Label SmoothingLayer NormalizationLinear LayerLinear Warmup With Cosine AnnealingLinear Warmup With Linear DecayMulti-Head AttentionPosition-Wise Feed-Forward LayerReLUResidual ConnectionSoftmaxTransformerWeight DecayWordPiece

Report a problem or propose a change · a person checks every report against the paper or source before anything changes; decisions are listed on /corrections